ZavaMesh Revenue Deal Room

Current release: solution 1.0.0.9 / authored agent 1.0.9. Built with SPFx 1.24.0 GA and React 18.3.1. Current package versions, test results and screenshot hashes are recorded in release evidence. Local validation is not tenant deployment or certification of every Copilot host.

ZavaMesh experience

The sample uses the supplied ZavaMesh woven identity, green/lime visual language and locally bundled brand imagery. The original 21 Copilot tool names and package identities remain stable. Six additional components bring the catalog to 27: attention drill-down, personal priorities, guided deal submission, independent approval, pipeline value flow and approval progress.

The keynote connects My Deals -> Deal Room -> Commercial Desk -> reviewed submission and independent decision -> Revenue Command. Each workspace has its own useful landing; a deal, request or scenario appears only through component context or explicit selection. Country opportunity values come from the scoped offline catalog with account-headquarters attribution. Recorded spend comes from a separate $3.4M expense ledger, not requested investment, rollout revenue or ROI.

The redesign is informed by public enterprise revenue patterns from Salesforce Revenue Cloud (guided quoting and human governance), DealHub (commercial orchestration), Clari / Salesloft (risk-to-action forecasting) and Dynamics 365 Sales (connected seller workflow). These are category benchmarks, not a worldwide market ranking.

Global competitor benchmark and design choices

Public product descriptions were reviewed on October 26, 2026. The comparison considers globally available offerings; it is not a hands-on certification, exhaustive competitive audit or market ranking.

Offering / primary source	Relevant pattern	ZavaMesh response
Salesforce Revenue Cloud	Guided quoting, pricing rules and human-controlled revenue governance	Live commercial math followed by visible policy routing and explicit confirmation
DealHub	Context-led, traceable commercial orchestration	One canonical request inbox, independent role stages and matching audit receipts
Clari / Salesloft Forecast	Period/region forecasting and risk-to-action visibility	Linked portfolio filters, seller-weighted outlook and evidence-aware interventions
Dynamics 365 Sales	Enterprise seller-workflow category benchmark	Separate seller, pursuit, commercial and leadership destinations rather than one universal dashboard
Zoho CRM	Team-specific spaces, collaboration, forecasting and BI	Role-owned workspaces with a shared confirmed-request overlay, not duplicated workflow records
Pipedrive reporting	Clear KPI dashboards, pipeline visibility and actionable reporting	Readable value hierarchy, real filter-linked chart geometry, geographic focus and exact-value alternatives

Sweet spot: a focused conversational entry expands into an evidence-aware system of action: inspect the opportunity, shape the economics, submit the actual request, make an independent human decision, and inspect the result. The sample borrows workflow principles, not competitor UI or assets.

The implementation contract and workspace boundaries are tracked in todo.md.

Brand redistribution: ZavaMesh logos and imagery were supplied for this sample. Final public redistribution approval remains pending, alongside the existing persona-media review. Optimized runtime images and source hashes are recorded in brand-provenance.json. To reimport approved originals, run .\scripts\import-zavamesh-brand.ps1 -Source '<local ZavaMesh asset folder>'.

This release targets SPFx 1.24.0 and React 18.3.1. Current local validation and package hashes are recorded in the generated release evidence. Starter ownership is defined in conversation-starters.json.

[!IMPORTANT] Ready-made package: deploy zava-revenue-deal-room.sppkg without building the project. The experience is a deterministic offline showcase; confirmations and receipts do not write to CRM, contact customers, approve terms, or change a production forecast. The deployment package is 1.0.0.9; its embedded agent matches the standalone Teams agent archive, authored as 1.0.9. SPFx may append a generated fourth agent-version segment; see release evidence for the packaged value.

Sample type: self-contained, offline-first SPFx Copilot Components showcase
Source catalog: 26 operational inline components + 1 capability explorer
Full-screen model: 4 role-aware lenses in one shared revenue application
Reference quality bar: Zava Innovation Hub and its agentic creation rules

Summary

ZavaMesh Revenue Deal Room demonstrates how an agent can become a precise system of action rather than a chat response or sales dashboard. Twenty-six independently routed operational tools and one capability explorer share a React 18 application, one connected Contoso revenue graph, four full-screen workspaces, and human-controlled review and confirmation patterns.

The keynote follows Contoso Global Expansion / ZDR-2042 from an owned priority through buyer proof, commercial choice and an independently reviewed request. The generated ZM-... request ID remains distinct from the opportunity. Facts, seller judgment, stale records, contrary evidence, deterministic calculations and inference remain visibly different throughout the journey.

Screenshots

Four purposeful workspace landings

Left-menu navigation opens these starting points, not an unrelated task appended below a dashboard. Select an item to work on it; Resume restores explicitly selected work and its mounted draft.

Selected work, country maps and value flows

Focused chat and responsive layouts

Applies to

• SharePoint Framework 1.24.0
• Microsoft Copilot declarative agents and Copilot Components preview
• React 18.3.1, Fluent UI v9 9.74.6, and Griffel 1.7.7
• Node.js 22 (22.14 or later), 24, or 26

Minimal Path to Awesome

1. Download and deploy zava-revenue-deal-room.sppkg to the tenant app catalog.
2. Approve deployment for all sites when prompted.
3. Add the packaged ZavaMesh Revenue Deal Room agent in Microsoft Copilot.
4. Start with Pipeline value flow, World opportunity map, Approval value flow, Submit a deal or Deal approvals. Explore capabilities is always the last starter.

High-impact conversation starters

Order	Starter	Opens
1	Pipeline value flow	Fictional regional value -> buyer readiness -> next-action Sankey.
2	World opportunity map	Current-quarter, all-region country opportunity choropleth; no default account filter.
3	Approval value flow	Fictional intake -> independent reviewer -> stage-outcome Sankey, distinct from stored requests.
4	Submit a deal	Guided story, economics, evidence and explicit reviewed-value submission.
5	Deal approvals	Actual stored demo requests, independent role stages and explicit decisions.
6	My revenue priorities	Owned-action cockpit with exact deal links and pipeline status.
7	Map the buying committee	Recorded buyer-role network, portraits and access gaps for Contoso.
8	Model a commercial offer	Editable 1,000-user / 24-month offer, live economics and computed margin curve.
9	World recorded-spend map	Country spend bubbles from the separate current-quarter demo expense ledger.
10	Build a mutual action plan	Buyer/seller ownership and an editable dependency plan for Contoso.
11	Build the customer value case	Editable benefit/cost assumptions and customer payback timeline.
12	Explore capabilities	The complete scenario explorer; never a substitute for one of the operational starters above.

The agent uses all 12 supported starters: 11 distinct operational component targets and the capability explorer in the final position. The exact prompts and component targets are maintained in conversation-starters.json and generated into the agent. Starter copy uses natural business questions rather than demonstration labels or chart-engine names; the component and agent context retain the data's actual provenance and decision boundaries. The routing matrix records the same ordered prompts. Fictional approval-flow values are not stored-request totals; neither opening approvals nor selecting a starter submits or approves anything.

To build locally:

npm ci
npm run build

Stop any Heft watcher before the production build and validate the package before restarting npm exec -- heft start --nobrowser. Watch mode regenerates development assets and the Teams ZIP; the validated .sppkg remains the production artifact. When restarting watch without source changes and retaining a byte-identical release, restore its embedded ClientSideAssets/zavamesh-revenue-deal-room.zip to teams/zavamesh-revenue-deal-room.zip after startup. This keeps the standalone agent archive and release-evidence hashes aligned with the solution.

Use npm run capture:visual to regenerate tenant-free publication screenshots and evidence. Use npm run capture:matrix for the supplementary all-intent/business-state matrix, and npx tsc --project scripts/visual-harness/tsconfig.json to typecheck the persistent-root harness. Use npm run capture:journeys to exercise intake, approvals, storage reload/reset, keyboard actions, 200% page-scale emulation and all four workspaces at 340/760/1440/1920px in both themes. Current results for all three visual suites are included in release-evidence.json.

Compact component-by-component dialogue

Inline cards show one decision stage at a time, rather than a miniature full-screen dashboard. The complete card, including its header and Copilot continuation, is limited to 650 px at a 760 px desktop viewport and 800 px at a 390 px narrow viewport. The business matrix measures rendered height and rejects clipped content or nested vertical scrolling as substitutes for staging.

Short lists, focused evidence details, individual charts and form steps keep the conversation moving. The primary continuation requests the next relevant component with the exact deal or request ID; alternative questions discuss the currently visible stage. Copilot receives the current stage, selection, entered values and confirmation state before the chat message, within the complete 10,000-byte UTF-8 context budget. Switching stages does not submit, approve or reset a draft. Expand remains available for the existing full-screen workspaces.

Try the connected deal journey

1. Expand a component into its matching workspace. Use the left menu to open a workspace landing, then select an actual opportunity or request; no unrelated task is appended to another view.
2. In My Deals, work a priority or open an opportunity. Connect buyer proof in Deal Room, then use Continue to Commercial scenario and Continue to Submit a deal to carry actual values. Enter the outcome and explicitly attest the evidence.
3. Review the exact values. Submission stays disabled until the confirmation is checked.
4. In Commercial Desk, select the same request, choose its assigned simulated reviewer role, enter a 15-1,000-character rationale, and explicitly confirm the stage decision.
5. The route always starts with Sales manager; discount above 10% adds Commercial director; discount above 18% or margin below 68% adds Executive committee. Each stage requires independent authority. Missing security or sponsor evidence blocks approval but permits a reasoned return/rejection.
6. Inspect the matching receipt, decision trail and updated status filter. Returned requests can seed a revised submission. Reset session data explicitly restores the two canonical baseline requests.
7. In Revenue Command, change quarter, region and Opportunity / Recorded spend. Country geometry, values and record drill-down share the visible scope; approved demo-request status remains distinct from the unchanged CRM-shaped catalog.

Only confirmed demo requests and decisions are stored in browser-session storage. Drafts and review consent stay with the current invocation. If storage is denied, a visible warning explains that confirmed actions remain memory-only. The shared request overlay survives route changes/reloads in the same browser session, but does not mutate the separate specialist workflows or CRM catalog. Role switching is a demonstration affordance, not authentication or real delegated approval. The agent ZIP is zavamesh-revenue-deal-room.zip; the .sppkg filename and component GUIDs are unchanged.

Current specialist experiences

All 27 intents have business-specific content: 14 read-only analyses, 6 draft submissions, 6 governed reviews and 1 capability explorer. There is no generic analysis/operation fallback. Only the following optional arguments are advertised; decisions and approvals are never prompt fields.

Tool	Consumed optional arguments	Working business UX
BuildAccountBrief	accountId, selectedId	Account growth-value footprint from recorded opportunity values, with source changes and whitespace kept distinct.
QualifyOpportunity	opportunityId	Evidence-based pursuit gate, discovery owner and missing criteria; return/decline remain available.
MapBuyingCommittee	opportunityId, selectedId	Contact portrait network with recorded roles/influence and unproven economic access; inspect the actual contact.
GetDealRisk	opportunityId, selectedId	Evidence-to-signature blocker network linking contrary/stale/inferred signals to relevant close checkpoints.
PrepareCustomerMeeting	meetingId, opportunityId, selectedId	Distinct steering/discovery agenda clocks, known attendees and proposed 30-minute allocations; not accepted outcomes.
BuildMutualActionPlan	opportunityId	Editable dependency DAG, buyer/seller owners and proposed signature target; the 21-day guard remains enforced.
ReviewMeetingCommitments	meetingId, opportunityId	Proposed-claim versus transcript/email reconciliation; unsupported funding/security cannot become accepted proof.
ShapeSolutionProposal	opportunityId, proposalId	Editable outcome, requirement-to-proof traceability, adoption, exclusions and solution owner.
SimulateCommercialOffer	opportunityId, quantity, termMonths, discount, services, probability, ramp	Five numeric inputs and annual licensing ramp recalculate TCV, full-run-rate ARR, margin, authority and weighted forecast.
ReviewDealException	exceptionId, opportunityId	Cash-timing runway and calculated incremental receivables exposure; finance/legal authority remains absent.
InspectForecastCommit	opportunityId	Seller-weighted outlook funnel versus sponsor/control/timing proof; draft category changes never change CRM.
ExplorePipelineQuality	period, region, selectedId, countryId, mapMetric	Filtered stage/proof/value health matrix, country recorded-spend analysis and exact-country continuations.
DiscoverAccountOpportunity	accountId, period, region, selectedId, countryId, mapMetric	Country opportunity choropleth or separate expense bubbles; exact country/deal drill-down. Omitted account scope means portfolio discovery.
ResearchCompetitivePosition	opportunityId, selectedId	Customer criteria, explicitly missing competitor proof and a scored-workshop proof plan.
CoachDealStrategy	opportunityId, selectedId	Constraint-led strategy choice canvas; access, proof and value plays do not claim win-rate uplift.
CreateExecutiveEngagementPlan	opportunityId	Proposed executive introduction path with actual demo identities, edited purpose and explicit contact guardrails.
TrackMeetingFollowUp	opportunityId, meetingId, selectedId	Buyer/seller commitment ownership board with recorded status and due offsets; elapsed time does not mean completion.
BuildValueCase	opportunityId	Customer investment/payback timeline using actual edited hours, rate, adoption and costs; no guaranteed ROI.
ReviewProposalReadiness	opportunityId, proposalId	Release-control board sourced from security, claims and terms flags; readiness drafts do not authorize release.
PlanCustomerSuccessHandoff	opportunityId	Outcome-to-measure-to-owner success canvas with an edited checkpoint and explicit unsigned-scope constraint.
ExploreAgentCapabilities	query	Search all 26 operational scenarios, inspect safe previews, or open the clicked component in Copilot with one explicit chat action.
SubmitDealRequest	accountId, opportunityId, quantity, termMonths, discount, services, probability	Inline story, exact full-license scenario prefills, evidence attestations, reviewed submission and receipt.
ReviewDealApproval	requestId	Current independent-review workload from stored requests, focused request list, exact values, authority/evidence/consent and audit receipts.
ExploreDealAttention	period, region, selectedId	Actual deal value versus observed attention flags, short ranked lists and focused economic/buyer/evidence facets.
GetMyRevenuePriorities	period, region	Owned-action cockpit with recorded due/source cues, not editorial score bars; scoped pipeline/status graphics remain separate.
ExplorePipelineFlow	period, region, selectedId	Regional value -> buyer readiness -> next action; inspect stages/connections, exact USD and conserved allocations.
ExploreApprovalFlow	period, selectedId	Intake route -> current reviewer -> stage outcome; distinguish stage versus final approval and inspect a named request's own terms.

Business value-flow dialogue

Ask Show how pipeline value moves from regions through buyer readiness to next actions. The current all-region scenario allocates $29.4M. Select any ribbon or node to replace the diagram with exact-value details; connections are paginated two at a time. Quarter and region filters change the planning model, not the separate catalog pipeline. Planning assumptions explains the 20% next scenario multiplier, conserved integer-USD allocation and provenance. Pointer, Enter and Space are supported; Back to value flow restores the chart. Its business title is Pipeline value & actions.

The overview continuation opens Approval progress, whose current planning scenario allocates $30.0M from intake routes to current independent reviewer stages and stage outcomes. Requests at later stages are assumed to have passed earlier reviews; Stage completion is not final approval. Its Copilot continuation opens ZM-1001 in Deal approvals, using that request's own values, evidence and decision trail. No chart volume is attributed to that request, and the chart never submits or approves anything. Both components publish the selected flow, visible page, scope, full model and assumptions within the existing 10,000-byte context budget. They obey the 650/800px complete-card budgets.

Task-specific qualification and competitive charts

Qualification gate uses a five-dimension buyer-proof ring: scope, budget, authority, security and timing. The Contoso example has 1 of 5 evidenced; the discovery example has 0 of 5. Contrary sponsor evidence, inferred security concerns and stale timing remain distinct. Editing a seller draft cannot establish proof or enable the blocked pursuit decision.

Competitive position uses a different lollipop chart of proposed workshop importance: security 35%, rollout 30%, value 25% and commercial fit 10%. The discovery scenario instead proposes budget 35%, integration 30%, decision process 20% and pilot proof 15%. These are explicit planning assumptions, not buyer-validated weights or invented competitor-performance scores. Selecting a driver opens its exact proposed weighting and supporting/missing evidence in the same component. Chart geometry and Copilot context use the same scoped business models.

Keynote-quality portfolio contract

All 27 registered components keep stable identities and answer distinct business questions. Operational charts no longer fall back to shared source-classification, age or generic value bars. An unregistered business visualization fails explicitly rather than rendering a plausible generic chart. Each primary experience has a task-specific model, meaningful source/proposal semantics, material selection or editing, and a relevant contextual continuation.

Information experiences start with the business graphic; Evidence records opens the short source list. Draft workflows separate Review impact from Edit business fields, then review the exact values. Long values are disclosed in 240-code-point parts without losing characters; all review pages must be visited before confirmation. Submission remains a dedicated initial form, and approval remains an independently assigned human decision with explicit rationale/consent.

Country opportunity value sums the provided catalog records and attributes them to explicit account headquarters, never international rollout revenue. Recorded spend uses a separate recorded expense ledger: current totals reconcile to $3.4M. Country details show exact opportunities or one expense per compact page. The next-quarter expense ledger is genuinely empty: the component shows an empty-spend stage, not a decorative zero map or invented future actuals. Quarter/region recovery updates the parent scope before any Copilot handoff. Maps use the offline Natural Earth atlas; no external tiles, data calls or customer records are fetched.

Commercial curve samples now recalculate full licensing term, ramp and services cost at every discount. The former arbitrary adoption index and approximated margin curve are removed.

Keynote: one conversation, one decision journey

Use the 6-minute keynote, with an optional five-minute cut: pipeline value -> world opportunity -> buyer proof -> inline commercial offer -> reviewed submission -> independent decision -> wider workspace context. Keep ZDR-2042 throughout the deal discussion; the new ZM-... receipt identifies the submitted request, not a replacement opportunity.

Keep the main run to two meaningful Copilot handoffs: offer -> submission -> independent approval. Inspect and edit directly; use Expand when more context and detail are useful, not as a required step between tasks. Keep the working submission outcome: "Roll out validated capabilities across 18 countries with an accountable customer owner and independent commercial review."

Inline Continue with Copilot: [next business step] requests the next interactive component. The optional full-screen Continue with Copilot: Summary & next steps requests a text-only chat summary of the active view, edited values, evidence gaps and confirmed status. It has no tool handoff and never submits or approves. Continue to [business step] remains local workspace navigation. More-discussion questions remain text; explicitly selected Open actions can still launch a component. An unfinished form's discussion cannot bypass its required review.

The personal brief uses Megan Bowen's illustrative eight-deal book, including named Northwind, Fabrikam, Alpine, Litware, AdventureWorks, Tailspin and Proseware examples. Current-quarter catalog math gives $18.54M open / $11.72M seller-weighted, with 1 of 5 recorded close milestones complete. Quarter/region filters change the actual data. This is not the signed-in user's real workload. Each priority has a direct scoped Copilot action, including exact request IDs for approval evidence.

The value bridge reconciles open value to seller-weighted value; it is not invented history or a loss prediction. Approved requests update the work overlay, not the CRM-shaped pipeline. This distinction is the final trust moment in the keynote.

For the complete pricing-to-submission path use 1,000 users / 24 months / 8% discount / $150,000 services / 65% seller probability, without a ramp. Full-screen Continue to Submit a deal opens the matching seller workflow with the exact inputs, preserving $2,818,000 TCV. Inline Continue with Copilot: Submit a deal requests that component through chat instead. In either path, the user still enters the outcome, attests evidence, reviews and explicitly confirms. Ramp scenarios cannot silently enter a non-ramped intake.

Inline deal-to-decision and Copilot dialogue

1. Ask for Deals needing attention. Filter the quarter and region, rank by open value or seller-unweighted value, then select a chart bar or Details for a deal. Buyer coverage, exact economic facts and recorded evidence have separate focused detail stages in the same inline component, without expanding. Focus moves to the detail heading; Back to attention catalog restores the filters and ranking and returns focus to the catalog.
2. Use Continue with Copilot to discuss the visible detail, or request a focused risk, buying committee, commercial or forecast component for that exact deal. The original view remains intact.
3. Ask to Submit a deal. Enter the deal story and economics, explicitly attest the evidence, visit all three exact-value review pages, and check confirmation before Submit deal request becomes available.
4. Continue with Copilot: Deal approvals sends the exact ZM-... receipt to ReviewDealApproval. The approval component shows the same stored request; required stages, independent reviewer authority, rationale and explicit confirmation remain mandatory. Other components in the same browser tab receive confirmed submission/decision updates through shared session storage.
5. After a confirmed approval, Continue with Copilot: My priorities opens the refreshed work overlay. Pending additional approval stages remain pending; the pipeline catalog is unchanged.

Each inline stage has one primary continuation; More discussions holds alternative questions and component launches. The keynote chain is priorities -> buying committee -> deal risk -> commercial scenario -> submission -> approval -> refreshed priorities. Capability previews never execute automatically. Each capability card has Open [title] in Copilot: one click sends a chat request for that card's registered component, without first selecting a preview or scrolling to a separate continuation control. The host/model, not the UI, decides tool routing. Unknown or filtered-out selections cannot hand off to a substituted deal.

The bridge publishes the click-time context before sending the follow-up, as one serialized operation. Context includes the discussion topic, exact scope and selection, source provenance, visible filters/records, edited draft, calculations, validation, confirmation/receipt, visible workspace state and requested next-tool properties. The complete content + structuredContent payload is capped at 10,000 UTF-8 bytes. Large views retain separate selected detail and report sampled/truncated paths and original counts; omitted evidence is unknown. Context or message failures are visible and retryable, and a failed context update never sends a context-free follow-up.

Attention flags are deterministic source/evidence observations, not an AI risk score. Seller-weighted and seller-unweighted values are arithmetic, not forecast commit or predicted loss. Submitted demo requests remain a separate session overlay; no CRM, customer communication or production approval is changed. The production package, publication screenshots and release evidence have been rebuilt for all 27 tools, and the full production build passes.

Current source validation: npm run build covers the full inline submission/approval journey, in-place detail/focus restoration and the exact UTF-8 budget. npm run check:intents, npm run check:routing-matrix, npm run check:generated-plugin and npm run check:package-output validate 27 registered tools, twelve starters, compiled schemas and matching production archives.

For production packaging, stop the development watcher and use npm exec -- heft build --clean --production followed by npm exec -- heft package-solution --production. Do not package watch output: its unminified bundle and hot-update files are development assets. Run package-output checks before restarting local validation with npm exec -- heft start --nobrowser: watch mode rewrites release/assets with development output, while the already-built production package remains unchanged.

Known example scopes: CONTOSO / ZDR-2042, ACC-002 / ZDR-2043, MTG-2042 / MTG-2043, PROP-2042 / PROP-2043, EX-2042 / EX-2043, and submitted demo request IDs such as ZM-1001. Omitting deal scope intentionally opens the Contoso specialist showcase; attention defaults to the current quarter/all regions, and approvals open the seeded request inbox. Explicit unknown, empty or conflicting IDs produce visible no-match and publish no substituted hero facts. Known alternate accounts never inherit Contoso evidence. Missing source records remain evidence gaps, not positive findings. Unknown or filtered selectedId values remain visible until cleared. period is current (quarter offset 0), next (1), or all; region is all or an actual account region.

Every submission/review has domain fields, runtime validation, a read-only review of actual chosen values, an initially unchecked explicit confirmation, a scope/value-bearing session receipt, Edit and Reset. Unsupported positive qualification/commit/exception/readiness decisions are blocked; valid return/decline paths remain usable. Prompt values cannot grant authority. Receipt and draft state survives passive rerenders, inline/full-screen continuation and theme changes. Changed invocation arguments or a new mounted tool instance reset the draft; Reset restores invocation defaults. State is mounted-session only: receipts are not propagated to other tool instances, persisted across reloads, or written to CRM. Bridge failures are visible, not silently acknowledged.

Commercial math charges recurring delivery cost for every licensed year, prorates partial years and annual licensed ramps, and charges services cost once. An omitted ramp means 100% licensing each year; [60,80,100] models a three-year ramp. Full-run-rate ARR is distinct from ramped subscription revenue. Weighted forecast is TCV × probability. The 68% floor and 10%/18% discount thresholds route authority; they never approve an offer. Commercial curves sample the actual entered economics. Value-case savings are explicit deterministic assumptions, not customer-validated benefits. The four workspaces have separate landing and selected-work states. Each component expands into its matching business view with relevant scope and controls, without an extra origin banner. Left-menu navigation opens the target workspace's landing; selecting a record or choosing Resume explicitly opens work. Mounted drafts survive navigation, and only the visible workspace/task publishes model context. Full-width country maps and value flows remain in their appropriate workspaces.

Features

• 27 Copilot Component identities with unique GUIDs and tool schemas; original identities are unchanged.
• One shared production bundle for React, Fluent, Griffel, domain data, and all component entries.
• Host-authoritative inline-to-full-screen expansion across My Deals, Deal Room, Commercial Desk, and Revenue Command.
• Buying-committee influence orbit, mutual-action close runway, evidence ledger, commercial outcome contour, and evidence-led forecast review.
• D3-derived commercial curves, pipeline scales, signed forecast-waterfall geometry, and an offline Natural Earth/TopoJSON opportunity map with selectable regional signals and exact-value tables.
• Coordinated visualization-first dashboards across My Deals, Deal Room, Commercial Desk, and Revenue Command; chart-appropriate inline tools reuse the same D3 data products at conversational width.
• Four purpose-built full-screen operating models aligned with enterprise sales-platform expectations: My Deals prioritizes seller actions and customer moments; Deal Room coordinates buyer influence, evidence, commitments, and win strategy; Commercial Desk joins live offer economics to exception authority and policy; Revenue Command connects aggregate pipeline, forecast movement, geography, and named leadership interventions.
• Seven bundled Microsoft 365 demo persona portraits across deal-team identity, stakeholder nodes, evidence sources, and commitment owners, with initials retained as the runtime fallback.
• Material commercial controls that recalculate contract value, annual recurring revenue, gross margin, approval authority, and weighted forecast.
• Review and submit workflows with Draft/Evidence, Review/Decision, Confirm, Receipt, Edit, and Reset.
• 10,000-byte bounded model-context snapshots, context-first suggested dialogue and exact-scope component handoffs.
• Searchable capability explorer covering all 26 operational scenarios with direct scoped chat launches.
• D3 Sankey experiences for imaginary pipeline and approval-stage value, with conserved allocations, accessible node/ribbon inspection, paginated details and explicit separation from live records.
• Deterministic offline graph with 80 accounts, 260 contacts, 120 opportunities, connected Contoso evidence, and no runtime data or profile-photo requests.
• Owner-document Griffel rendering, light/dark themes, reduced motion, responsive navigation, semantic status colors, SVG descriptions, and text equivalents.

Validation status

The production gate validates source tools, compiled schemas, React root lifecycle, current visuals, and both deployment archives. See generated release evidence for the current test total, sizes, and hashes.

Current completion evidence includes the passing Jest suite, 32 curated publication captures, and the expanded business matrix: all 27 inline tools at standard/narrow/dark widths/themes (81), every Sankey node and ribbon, country opportunity/spend maps and recovery scopes, focused detail/chart stages, paginated fields and records, inline submission/approval receipts, four dashboards, ten reviews, ten full-screen receipts, and changed/filtered/drilled/no-match/gated states. Every inline stage is measured at desktop and narrow widths, including pages that are not saved as screenshots. Another 42 journey captures cover connected workflows and 32 workspace width/theme variants. The generated release evidence records current test/capture totals and maximum measured heights. Validation rejects runtime, console, broken-image, clipping, nested vertical scrolling and horizontal-overflow failures. Inline Copilot-send success is acknowledged by an accessible checkmark in the existing continuation, without adding card height; changing the visible scope clears that acknowledgement. All three visual sets are SHA-256-bound to current runtime sources; npm run build rejects stale screenshots. The DOM suite uses React 18 act, persistent roots and IS_REACT_ACT_ENVIRONMENT, and tests content, parameters, calculations, gates and lifecycle rather than layout labels alone. Every host redraw and committed subtree change schedules an SDK requestSizeChangeAsync notification after layout. A content-root ResizeObserver also covers delayed image/font layout and responsive changes. Requests are coalesced per animation frame and measure the actual app rather than the iframe or theme provider, so shorter stages can shrink correctly. Observers and queued measurements are cancelled on teardown or disposal.

The production .sppkg contains one shared hashed JavaScript asset, 11 unique runtime media assets, zero stale bundles and matching embedded/Teams agent archives. Package validation reads the actual solution and feature versions from the archive XML and the embedded agent version from its manifest, instead of relying only on source configuration. Generated release evidence records the current sizes and hashes. Local validation does not deploy these archives to the tenant.

Historical baseline (31 August 2026, before RC0):

• 21 components, 21 unique GUIDs, 21 unique tools, and one shared bundle.
• 6 deterministic domain/D3 geometry tests passed; 0 failed.
• 13 Playwright visual captures passed with 0 runtime, broken-image, or horizontal-overflow failures.
• 7 runtime portraits and 7 source copies passed SHA-256 provenance checks; the final package contains 7 unique hashed media assets and 0 duplicate media binaries.
• Generated API plugin v2.4 contains 21 functions; name_for_human length is 17.
• Final .sppkg: 341,417 bytes; one 398,994-byte hashed production JavaScript asset; 0 stale assets; one embedded current agent package.
• .sppkg SHA-256: 2B9898398850D0F0C7821469D79C0D3B33FF2F949E4A4305852CD1EA0F032E52.
• Agent ZIP SHA-256: 680744C396D7133CAA697D29B64FB7E37F86CDF9976D7B1CDC1B514618354B5A.

Persona files are byte-identical to the standard demo-persona media used by the reference Zava sample. Hashes, source-copy paths, and intended uses are recorded in assets/asset-provenance.json. Public redistribution remains subject to final media-rights review; this status is enforced by npm run check:media.

Tenant-authenticated Workbench CSP, host focus restoration, bridge routing, forced colors, and screen reader output remain environment-specific validation gates and are not claimed from the local harness.

Solution structure

src/copilotComponents/    27 registered identities and schemas
src/shared/               catalog, deterministic domain, React host, theme, and experiences
scripts/                  catalog, visual evidence, gallery, plugin, and package automation
assets/                   publication screenshots, evidence, and PnP sample metadata
copilot/                  declarative agent, plugin, instructions, and icons
sharepoint/solution/      ready-to-deploy production package

Demo and review assets

• 6-minute keynote - beautiful interactive UX, contextual owned priority and buyer proof through exact commercial values and an accountable decision.
• 10-minute business demo - the broader cross-role revenue journey with guardrails and current implementation boundaries.
• 5-minute technical demo - immutable catalog, React 18 host, D3 geometry, offline map, evidence, and package gates.
• Designer and rehearsal review - visual language, interaction/accessibility checks, screenshot acceptance, and fallback path.
• Routing matrix - 27 tool boundaries, twelve canonical starters, nine sibling-collision rehearsals, and tenant validation notes.
• Release evidence - generated package, bundle, media, visual, and hash evidence. Run npm run generate:release-evidence only after the final package build.

Maintenance and extension

To add or change an operational capability, update the typed catalog and the supported final-named component schema, then run npm run configure:intents, npm run generate:routing-matrix, the focused tests, npm run capture:visual, and npm run build. Capability Explorer reads the same catalog; do not maintain a second scenario list in its React UI.

The sample currently calculates commercial value, gross margin, weighted forecast, D3 commercial curves, evidence-weighted pipeline bars, forecast movement, and regional opportunity marks from deterministic local data. These are showcase calculations, not pricing, revenue-recognition, legal, or forecasting advice. Live CRM, Microsoft Graph, CPQ, pricing-policy, workflow, approval, and e-signature implementations remain deferred behind the service boundaries described below.

Outcome and telemetry plan

The offline sample emits no telemetry. A production implementation should measure bounded outcomes without capturing customer content: intended-tool routing success, property extraction completeness, time to first useful state, Expand continuation success, control use, validation recovery, explicit confirmation completion, bridge rejection, and abandonment. Review these signals by scenario and host mode; never use inferred sentiment or relationship strength as employee-performance telemetry.

Limitations and support

• Tenant-authenticated model routing, Workbench CSP, iframe focus, forced colors, and screen-reader host output require a Microsoft 365 test tenant and remain external validation gates.
• Specialist drafts/receipts remain tool-instance-local. The new full-screen request center shares only confirmed demo requests/decisions through a session overlay; it never updates specialist evidence or the opportunity baseline. Live CRM/CPQ/approval integrations are absent.
• Persona redistribution is pending final media-rights review as recorded in assets/asset-provenance.json.
• For sample issues, use the repository issue tracker and include the component name, prompt, host mode, browser, screenshot, and whether the problem reproduces in the tenant-free harness.

Version history

Version	Date	Notes
1.0.0.9 / agent 1.0.9	2026-11-02	Six-minute keynote with beautiful value flows, world maps and an accountable deal journey. Full-screen Continue with Copilot requests a text-only chat summary; inline handoffs request the next interactive UX. Explicit response-mode context and agent rules preserve scope, edited values and consent.
1.0.0.8 / agent 1.0.8	2026-11-02	Business-first continuation/discussion routes, request/approval controls and workspace copy; exact named-request handoffs, refreshed gallery and rebuilt archives. Golden and local creation rules now apply the same language policy across complete journeys while preserving provenance and consent.
1.0.0.7 / agent 1.0.7	2026-11-02	Business-first keynote starter wording, with demonstration labels and chart-engine jargon removed from all starter copy. Twelve distinct starter routes and truthful data/execution boundaries remain intact.
1.0.0.6 / agent 1.0.6	2026-11-02	Twelve conversation starters: eleven distinct operational UX components and the capability explorer last. Canonical and local automation guidance now documents the supported maximum of twelve rather than a six-starter assumption.
1.0.0.5 / agent 1.0.5	2026-11-02	High-impact pipeline/approval value-flow, world-map, submission and approval starters; capability explorer last. Reusable context-first dialogue, exact-scope continuation, bounded stages, workspace and resize lessons synchronized to the canonical golden guide and local creation rules.
1.0.0.4 / agent 1.0.4	2026-11-02	SPFx 1.24.0 GA; 27 distinct components, four explicit workspace stories, country opportunity/spend maps, fictional value flows, exact commercial/request continuity and reliable redraw sizing. Refreshed README/gallery and rebuilt production archives; current validation totals and packaged agent suffix are recorded in release evidence.
1.0.0.3 / agent 1.0.3	2026-10-26	ZavaMesh branding, linked worldwide revenue/investment analytics, guided deal intake, sequential demo approvals and session audit receipts. Current test/visual totals are generated in release evidence.
1.0.0.1 / agent 1.0.1	2026-10-20	RC0 specialist completion: 21 scoped experiences, 10 guarded workflows, 87 tests, 118 current visual captures. The SDK may append a fourth agent hotfix segment.
1.0.0	2026-08-31	Initial GitHub candidate with 21 tools, four visualization-led workspaces, D3 charts/map, persona media, demo scripts, publication evidence, and audited .sppkg.

Historical design exploration (not the shipping contract)

The remaining original product brief records broader design aspirations (including fields, comparisons, cross-tool overlays and integrations that are not advertised or implemented in this offline release). The current specialist table, runtime schemas, validation evidence and limitations above are authoritative; do not present the historical interaction matrix below as current behavior.

Business story

A seller names the account and desired outcome. The agent turns scattered relationship evidence into an inspectable deal strategy; the revenue team commits to actions, tests commercial choices, and closes with a forecast it can defend.

The current repository already demonstrates sales KPIs and filters. This sample must move from passive reporting to active opportunity execution:

1. A seller names an account, opportunity, meeting, or desired outcome in Copilot.
2. Copilot selects one bounded experience and initializes trusted CRM-shaped and Microsoft 365 context.
3. The component assembles buyer evidence, relationship signals, commitments, commercial constraints, and risks for the exact job.
4. The seller or specialist edits assumptions, accepts or rejects proposed changes, and tests scenarios.
5. Stage, quote exception, contractual exception, and forecast decisions remain visible human decisions.
6. Full screen preserves the same account, deal, stakeholders, scenario, plan, and review step.
7. Confirmed session receipts update the deal team, manager forecast, commercial desk, and revenue view.

The sample complements CRM rather than claiming to replace it. CRM-shaped records remain the system of record; Copilot becomes the intent-led workspace across CRM facts, meetings, mail, files, and people.

Market-informed product position

Public enterprise sales-platform baselines consistently cover account and contact management, activity capture, lead/opportunity qualification, pipeline, forecasting, relationship intelligence, engagement, quotes, contracts, approvals, workflow, and analytics. References include Salesforce Sales Cloud, Microsoft Dynamics 365 Sales, and the Salesforce State of Sales. The latter is vendor research and is used directionally, not as an independent benchmark.

This sample modernizes that baseline through bounded, inspectable agent UX:

Traditional sales pattern	Copilot Components modernization
Sellers navigate account, opportunity, activity, contact, quote, and forecast screens.	The seller states the outcome; one purpose-built component assembles the needed records and action.
AI creates a deal summary or score with little provenance.	Buyer evidence, seller opinion, stale activity, agent inference, and missing proof are visibly separated.
Stakeholders are a contact list.	An interactive buying-committee map reveals authority, stance, influence paths, relationship strength, and missing access.
Close plans are static task lists.	A mutual action plan joins buyer and seller commitments; dependencies redraw the credible close range.
Discount approval is detached from deal strategy.	Commercial scenarios show revenue, margin, customer outcome, approval path, and forecast consequence before review.
Forecast categories depend on seller confidence.	Managers inspect evidence, movement, critical path, risk, and scenario range before accepting commit.

Do not use Gartner branding, proprietary quadrant language, or market-leadership claims. Industry expectations define workflow completeness; the sample demonstrates a differentiated interaction model.

Showcase objectives

Objective	GA proof
Upgrade sales from dashboard to system of action	Qualification, stakeholder mapping, meeting preparation, commitments, proposal shaping, commercial simulation, exception review, and forecast inspection are actionable inline.
Make buyer evidence visible	Every forecast and recommendation distinguishes verified buyer evidence from seller opinion, stale activity, and inference.
Connect collaboration to revenue	Meeting and mail-shaped context becomes reviewed CRM updates and mutual commitments, never silent writes.
Put commercial judgment in context	Discount, term, ramp, services, and probability changes recalculate margin, authority, approval, and forecast consequences.
Preserve exact continuation	Expand retains account, opportunity, selected stakeholder, risk, plan, scenario, and decision draft.
Tell one cross-role story	A global expansion opportunity moves from risk and relationship gap to mutual plan, offer, exception approval, and defensible commit.
Deliver a distinct visual identity	Influence orbit, commitment runway, evidence ledger, commercial contour, and forecast bridge become signature visuals.

End-to-end operating model

Phase	Business job	Agentic experience	Owner
1. Understand	Assemble account priorities and relationship history.	Grounded brief with sources, freshness, open commitments, whitespace, and risks.	BuildAccountBrief
2. Qualify	Decide whether and how to pursue.	Need, authority, urgency, value, competition, fit, and evidence gaps precede human stage decision.	QualifyOpportunity
3. Mobilize	Build access to the buying group.	Influence graph shows roles, stance, authority, relationship strength, missing contacts, and internal owners.	MapBuyingCommittee
4. Diagnose	Identify what can derail the outcome.	Evidence-ranked risk combines inactivity, single-threading, criteria, competition, legal, delivery, and next steps.	GetDealRisk
5. Engage	Prepare the next customer interaction.	Briefing ties attendees, goals, commitments, objections, proof, and questions to one meeting.	PrepareCustomerMeeting
6. Commit	Convert intent into a shared path.	Buyer/seller milestones, dependencies, owners, evidence, and critical path form a mutual action plan.	BuildMutualActionPlan
7. Capture	Turn a meeting into reviewed business updates.	Each proposed CRM field and commitment can be accepted, edited, assigned, or rejected.	ReviewMeetingCommitments
8. Propose	Shape an outcome-led solution.	Requirements, proof, architecture, adoption, value, exclusions, and owners form a reviewed proposal.	ShapeSolutionProposal
9. Model	Test the commercial structure.	Quantity, term, ramp, discount, services, currency, and probability recalculate value and consequences.	SimulateCommercialOffer
10. Govern	Review a nonstandard commercial or contractual request.	Finance, legal, security, and sales inspect one evidence package and record conditions.	ReviewDealException
11. Forecast	Commit only what evidence supports.	Opportunity movement, buyer proof, critical path, risk, and range precede forecast decision.	InspectForecastCommit
12. Lead	Improve pipeline quality and intervention.	Leaders inspect coverage, aging, evidence quality, concentration, slip risk, and intervention impact.	ExplorePipelineQuality

Audiences and full-screen information architecture

Lens	Primary persona	First useful state	Work and decisions
My Deals	Megan Bowen, account executive	Prioritized deals, relationship gaps, commitments due, meeting preparation, and next best owned action.	Prepare, qualify, update, plan, propose, and request exceptions.
Deal Room	Megan plus cross-functional pursuit team	One opportunity with win strategy, buying committee, risk, mutual action plan, evidence, and team ownership.	Coordinate access, proof, commitments, solution, and close path.
Commercial Desk	Miriam Graham, finance partner, with legal/security reviewers	Exception queue, scenario comparison, margin/value, policy, terms, conditions, and authority.	Review discount, ramp, terms, security, delivery, and quote exceptions.
Revenue Command	Joni Sherman, chief revenue officer	Evidence-weighted forecast, pipeline quality, movement bridge, concentration, interventions, and accountability.	Inspect commit, challenge risk, assign intervention, and verify outcome.

My Deals starts with owned actions, not a wall of KPIs. Deal Room is the keynote workspace and coordinates one selected stakeholder/risk/commitment across every region. Commercial Desk uses a queue plus one immersive scenario review; it must not become a generic approvals page. Revenue Command emphasizes forecast evidence and intervention, not duplicate regional sales charts.

Inline component portfolio

Target: 26 operational components plus one safe capability explorer. The expanded catalog covers account executive, sales development, deal strategy, solution engineering, value engineering, executive sponsorship, finance, legal/security, sales management, revenue operations, and customer success handoff.

Model	Count	Contract
Information / interactive analysis	10	Immediate answer plus material selection, filtering, comparison, or chart changes.
Submit / create	6	Prompt-prefilled draft -> validation -> review -> explicit confirmation -> session receipt.
Review / decision	4	Record/queue -> evidence and consequence -> decision draft -> confirmation -> receipt.
Education / discovery	1	Search/filter all operations, copy prompts, and safely preview without action confirmation.

#	Component	Model	Prompt properties	First useful inline state	Material interaction and action	Full-screen continuation
1	BuildAccountBrief	Information	accountId, period, priority, sourceTypes, selectedEventId	Account goals, relationship pulse, products, opportunities, recent moments, open commitments, whitespace, and sourced risks.	Period/priority/source filters rebuild evidence; selection opens source freshness and owner.	deal-room/account-brief; preserves account, filters, event, and evidence.
2	QualifyOpportunity	Review	opportunityId, proposedStage, qualificationFramework, selectedCriterion	Need, authority, urgency, value, competition, fit, evidence completeness, gaps, and stage consequence.	Edit criterion confidence and evidence; advance/hold/disqualify requires rationale, review, confirmation, and receipt.	my-deals/qualification; preserves opportunity, criterion, evidence, and decision draft.
3	MapBuyingCommittee	Information	opportunityId, contactIds, roleFilter, stanceFilter, selectedContactId	Influence graph with buying role, stance, authority, relationship strength, last evidence, missing access, and internal owner.	Add/resolve role, filter, select, assign relationship owner, and test introduction path; graph redraws.	deal-room/stakeholders; preserves graph filters, selection, and proposed ownership.
4	GetDealRisk	Information	opportunityId, riskCategory, severity, period, selectedRiskId	Ranked risks with evidence for/against, age, consequence, owner, and recommended bounded action.	Filters rerank; selection coordinates timeline and source; assigning mitigation requires reviewed session action.	deal-room/risks; preserves filters, selected risk, and mitigation draft.
5	PrepareCustomerMeeting	Information	meetingId, opportunityId, attendeeIds, objective, duration, selectedTopicId	Attendee roles/stance, meeting objective, open commitments, likely objections, approved proof, questions, and agenda.	Reorder agenda, select proof, add question, change objective; export/copy remains reviewed.	my-deals/meeting-prep; preserves meeting, attendees, objective, agenda, and topic.
6	BuildMutualActionPlan	Submit	opportunityId, targetCloseDate, customerLaunchDate, milestones, ownerScope	Buyer and seller commitments on a dependency runway with owner, due date, evidence, status, and credible close range.	Add/edit/reorder dependencies and dates; changes recalculate critical path and slip risk. Confirm creates session plan receipt.	deal-room/mutual-plan; preserves plan draft, selected milestone, date scenario, and step.
7	ReviewMeetingCommitments	Review	meetingId, opportunityId, proposedChanges, selectedChangeId	Side-by-side meeting evidence and proposed CRM fields/commitments, each labeled fact or inference.	Accept/edit/reject/assign each change; bulk confirm prohibited until every inference is reviewed. Receipt updates mock graph.	my-deals/meeting-review; preserves evidence, decisions per change, and review step.
8	ShapeSolutionProposal	Submit	opportunityId, audience, outcomes, requirements, proofIds, currency	Editable proposal architecture: outcomes, requirements, solution, proof, value, adoption, risks, exclusions, and owners.	Reorder sections, resolve unsupported claims, choose proof, change assumptions; review precedes draft receipt.	deal-room/proposal; preserves proposal draft, evidence, validation, and step.
9	SimulateCommercialOffer	Information	opportunityId, quantity, termMonths, discount, ramp, services, currency, probability	Editable scenario beside revenue, ARR, margin, customer value, approval level, close probability, and forecast impact.	Inputs recalculate contour and compare up to three named scenarios. “Request review” creates a draft, never approval.	commercial-desk/scenario; preserves inputs, comparisons, selected metric, and scenario name.
10	ReviewDealException	Review	exceptionId, opportunityId, exceptionType, proposedDecision, conditionText	Customer need, requested exception, commercial impact, policy, precedent, risks, approvers, and alternatives.	Approve with conditions/return/decline requires authority, rationale, review, confirmation, and receipt.	commercial-desk/exception; preserves exception, scenario, evidence, conditions, and decision step.
11	InspectForecastCommit	Review	opportunityId, forecastPeriod, proposedCategory, selectedEvidenceId	Deal movement, verified buyer evidence, seller judgment, critical path, risk, range, and forecast-category consequence.	Select evidence, change scenario/category, request clarification, accept/change commit through confirmed manager decision.	revenue-command/forecast-review; preserves period, opportunity, evidence, category, and draft.
12	ExplorePipelineQuality	Information	period, region, segment, owner, stage, metric, selectedOpportunityId	Coverage, conversion, aging, evidence quality, concentration, slip risk, and forecast bridge with named interventions.	Filters rebuild models; chart/table mode and selection coordinate exact values and opportunity detail.	revenue-command/pipeline; preserves filters, metric, visual mode, and selection.
13	DiscoverAccountOpportunity	Information	accountId, period, productFamily, region, signalType, selectedSignalId	Opportunity canvas combining account priorities, product/adoption whitespace, service signals, organizational change, intent evidence, and potential outcome.	Filters and selected signals reshape opportunity themes, evidence strength, likely stakeholders, and suggested discovery question; no lead is created automatically.	my-deals/opportunity-discovery; preserves account, signals, theme, filters, and selection.
14	ResearchCompetitivePosition	Information	opportunityId, competitorIds, dimension, period, selectedClaimId	Competitive battleboard separating verified customer evidence, public facts, seller observations, strengths, gaps, landmines, and approved proof.	Selecting competitor/dimension changes comparison and countermove options; unsupported claims are excluded from customer-facing reuse.	deal-room/competition; preserves opportunity, competitor, dimension, claim, and evidence.
15	CoachDealStrategy	Information	opportunityId, coachingFocus, sellerId, period, selectedActionId	Coaching workspace with deal pattern, evidence gaps, questions, risks, prior movement, seller choices, and a small set of explainable strategy plays.	Seller tests a play and sees stakeholder, critical-path, evidence, and forecast consequence; assigning a coaching action is reviewed.	my-deals/coaching; preserves deal, focus, selected play, evidence, and action draft.
16	CreateExecutiveEngagementPlan	Submit	opportunityId, executiveIds, customerContactIds, objective, timing, talkingPoints	Sponsor plan mapping executive-to-executive relationships, objective, value message, proof, sensitivities, meeting path, owner, and follow-up.	Change sponsor/contact/objective to update fit, conflict, relationship coverage, and preparation needs. Confirm creates a reviewed engagement plan.	deal-room/executive-engagement; preserves relationships, plan draft, selected sponsor, and step.
17	TrackMeetingFollowUp	Information	opportunityId, meetingId, status, owner, selectedCommitmentId	Buyer/seller commitment runway with source, owner, due date, evidence, dependency, at-risk/overdue state, and forecast consequence.	Filters change the runway; mark received/at risk/complete requires evidence and reviewed action. Overdue commitments can be escalated with context.	deal-room/commitments; preserves deal, meeting, filters, selected commitment, and action draft.
18	BuildValueCase	Submit	opportunityId, outcomes, baseline, benefitDrivers, costs, currency, timeHorizon	Joint value model with baseline, outcome tree, benefit drivers, costs, assumptions, cash flow, payback, sensitivity, evidence, and owner.	Editing assumptions recalculates value range, payback, confidence, and evidence gaps. Review creates a value-case draft receipt.	deal-room/value-case; preserves assumptions, scenario, selected driver, evidence, and step.
19	ReviewProposalReadiness	Review	opportunityId, proposalId, reviewScope, selectedFindingId, proposedOutcome	Readiness gate covering customer outcomes, requirements traceability, approved proof, commercial alignment, legal/security dependencies, delivery capacity, and unresolved claims.	Reviewers accept/edit findings and choose ready/ready with conditions/return, with owners, due dates, rationale, confirmation, and receipt.	commercial-desk/proposal-readiness; preserves proposal, findings, conditions, and decision draft.
20	PlanCustomerSuccessHandoff	Submit	opportunityId, successOwnerId, outcomes, milestones, risks, handoffDate	Post-sale success canvas connecting sold outcomes, contract commitments, stakeholders, adoption milestones, unresolved risks, measures, and ownership.	Editing milestones/owners changes readiness, dependency, and first-value date. Confirm creates a reviewed handoff plan without closing the opportunity.	deal-room/success-handoff; preserves outcomes, owners, milestones, risks, and step.
21	ExploreAgentCapabilities	Education	query, audience, operation, featuredIntent	Searchable catalog with realistic prompts and safe previews for all 20 operations.	Search, filter, copy, and preview; review/submit previews stop before confirmation.	Isolated education/capabilities; preserves query, filters, and featured intent.

Ownership boundaries

• Account brief assembles relationship context; meeting preparation owns one upcoming interaction.
• Qualification owns pursue/stage judgment; forecast review owns period commit judgment.
• Buying committee owns people/influence; deal risk owns evidence-ranked threats and mitigations.
• Mutual action plan owns customer/seller commitments; meeting review proposes updates into that plan.
• Proposal owns outcome narrative and proof; commercial simulation owns price/term economics.
• Commercial simulation never approves; exception review owns authority, conditions, and decision.
• Pipeline Quality owns aggregate leadership analysis and deep-links to selected operational work.

Role and operation coverage

Role	Moment in the revenue process	Primary inline components	UX forms that earn separate routing
Sales development / account executive	Discover, understand, qualify, prepare, follow up	DiscoverAccountOpportunity, BuildAccountBrief, QualifyOpportunity, PrepareCustomerMeeting, ReviewMeetingCommitments, TrackMeetingFollowUp	Signal canvas, sourced brief, qualification gate, meeting brief, evidence reconciliation, commitment runway.
Deal strategist / manager	Diagnose, coach, mobilize influence, govern close	GetDealRisk, CoachDealStrategy, MapBuyingCommittee, CreateExecutiveEngagementPlan, InspectForecastCommit	Risk stack, strategy play simulator, influence graph, sponsor plan, forecast evidence review.
Solution and value engineering	Prove the outcome and solution	BuildValueCase, ShapeSolutionProposal, ReviewProposalReadiness	Editable value model, evidence-led proposal studio, cross-functional readiness gate.
Finance / legal / security	Shape economics and govern exceptions	SimulateCommercialOffer, ReviewDealException, ReviewProposalReadiness	Commercial contour, authority-aware exception review, readiness findings and conditions.
Revenue operations / executive	Inspect pipeline quality and intervention	ExplorePipelineQuality, InspectForecastCommit, CoachDealStrategy	Forecast bridge, evidence distribution, named intervention and coaching action.
Customer success	Preserve sold outcomes through handoff	PlanCustomerSuccessHandoff, BuildMutualActionPlan, TrackMeetingFollowUp	Outcome handoff canvas, shared critical path, evidence-backed commitment tracker.

Adaptive UX demo choreography

Beat	Prompt and role	Component shape	Adaptation that the audience can see
1	Seller: “Where is the next growth opportunity in Contoso?”	Account signal canvas	Product/adoption/service signals become an opportunity theme; selecting evidence changes likely outcomes and discovery questions.
2	Seller: “Should we pursue this expansion?”	Qualification review	The exploratory canvas becomes a criteria/evidence gate with advance, hold, or disqualify consequences.
3	Deal lead: “Who can move the decision?”	Buying-committee graph	Contacts become influence, stance, authority, access gaps, and owned introduction paths.
4	Manager: “What strategy improves our position?”	Coaching play simulator	Choosing executive access, technical proof, or timeline reset changes risks, stakeholders, critical path, and forecast evidence.
5	Value engineer: “Prove the customer's outcome.”	Financial value model	Conversation facts become editable baselines, benefit drivers, sensitivity, payback, confidence, and missing evidence.
6	Solution architect: “Shape the proposal.”	Proposal studio	The value model becomes requirements, solution, proof, adoption, exclusions, and owners; readiness review later changes it into findings and conditions.
7	Finance: “Protect margin while improving adoption.”	Commercial scenario contour	Quantity, ramp, term, services, and discount reshape revenue, margin, customer outcome, authority, and approval path.
8	Seller: “What did the customer actually commit to?”	Evidence reconciliation into runway	Meeting notes become item-by-item proposed changes, then verified buyer/seller commitments with due dates and forecast consequence.
9	Manager: “Can this remain in commit?”	Forecast review	All prior evidence collapses into a governed category decision with range, conditions, checkpoint, and receipt.
10	Customer success: “Carry the sold outcome into delivery.”	Success handoff canvas	Deal artifacts become adoption outcomes, risks, owners, first-value milestones, and a reviewed cross-team handoff.

Showcase story - from signal to realized customer outcome

The hero demo starts before an opportunity exists and ends after signature planning. Contoso moves from an account signal through qualification, stakeholder access, coaching, value proof, proposal readiness, commercial review, customer commitments, forecast decision, and success handoff. The component changes shape at every role boundary while the same account, opportunity, evidence, assumptions, and receipts remain connected.

Dynamic inline UX contract

• Header: Zava Revenue Deal Room + literal action + responsive Expand control.
• The first viewport answers or advances one sales job; never show prompt echo, generic chat, or a decorative “AI score.”
• Every risk, recommendation, and forecast displays buyer evidence, contrary evidence, source freshness, owner, and whether the value is fact, seller judgment, deterministic output, or inference.
• Forms and plans include validation, review, explicit confirmation, and a session receipt.
• Review components permit item-by-item correction before confirmation; no inferred meeting change is bulk-written silently.
• Commercial inputs visibly change revenue, margin, customer outcome, authority, approval path, and forecast. Formulas and assumptions are inspectable.
• Charts provide exact values and table alternatives and remain useful at 340 px.
• Design loading, stale CRM, missing account match, partial M365 evidence, permission, offline mock, conflict, and failed action states explicitly.

Signature visual and brand system

The Deal Room should feel like a precise editorial workspace for a high-performing pursuit team, not a blue CRM dashboard and not the Innovation Hub recolored.

• Brand: “Zava Revenue Deal Room”; planned mark combines an open doorway and forward revenue line.
• Palette: midnight #13263D, paper #F7F4EE, electric azure #146EF5, verdant #1A8F5A, and vermilion #E5563F; use semantic colors consistently.
• Typography: confident condensed display treatment for deal/value moments; highly legible operational sans and tabular numbers for controls and evidence.
• Composition: asymmetric editorial grid, strong vertical “close runway,” relationship orbit, and evidence ledger. Cards only for deals, people, scenarios, decisions, and bounded forms.
• Motion: stakeholder links settle, critical path reflows, and commercial contour interpolates once in 180-300 ms. Reduced motion renders final geometry immediately.

Question	Signature visual	Planned implementation
Who actually influences the decision?	Buying-committee influence orbit	Deterministic React SVG graph plus DOM people/evidence list.
Can both parties reach the target date?	Mutual-action commitment runway	DOM milestones with SVG dependency/critical-path overlay.
What supports the forecast?	Deal evidence ledger	Layered verified/judgment/inferred/stale bands with exact records.
Which offer balances customer outcome and economics?	Commercial contour and scenario table	DOM inputs plus SVG scenario geometry and exact comparison table.
Why did the forecast move?	Forecast bridge	React SVG waterfall with named deals and table alternative.

Hero scenarios

Hero 1 - Rescue the Contoso expansion deal

Prompt: “Show the risks threatening Contoso's global expansion deal this quarter.”

GetDealRisk identifies strong champion support but no economic-buyer access, an unowned security milestone, and a stale launch assumption. MapBuyingCommittee reveals the influence gap and assigns an executive relationship owner. BuildMutualActionPlan works backward from launch and recalculates a credible close range. After the next meeting, ReviewMeetingCommitments rejects one inferred customer promise and confirms the verified actions.

Proof: evidence-grounded risk, interactive people graph, plan creation, date calculation, inference correction, cross-role handoff, exact continuation, and session updates.

Hero 2 - Shape an offer without giving away the deal

Prompt: “Model a three-year ramped offer for Contoso that protects 68% gross margin.”

SimulateCommercialOffer opens inline with quantity, ramp, services, discount, term, currency, and probability. The seller compares base, accelerated, and protected-margin scenarios. The chosen ramp improves customer adoption economics but triggers one nonstandard payment-term review. ReviewDealException gives finance and legal the same evidence and records approval with a condition.

Proof: rich inline form, live KPIs and chart, deterministic calculations, comparison, policy, approval chain, conditions, confirmation, and receipt.

Hero 3 - Defend the forecast commit

Prompt: “Should Contoso remain in commit for Q4?”

InspectForecastCommit separates buyer-confirmed milestones from seller judgment and agent inference. The critical path now supports the period, but security acceptance remains a named condition. The manager accepts commit with a checkpoint; Revenue Command updates the movement bridge and confidence distribution.

Proof: review UX, evidence provenance, scenario range, human judgment, forecast consequence, and leadership rollup without a generic dashboard.

Conversation starters

#	Title	Starter	Expected component
1	Advance a deal	Show the risks threatening Contoso's global expansion deal this quarter.	GetDealRisk
2	Buying committee	Map the buying committee for the Contoso expansion.	MapBuyingCommittee
3	Commercial scenario	Model a three-year ramped offer that protects 68% margin.	SimulateCommercialOffer
4	Meeting preparation	Prepare me for tomorrow's Contoso steering meeting.	PrepareCustomerMeeting
5	Forecast decision	Should Contoso remain in commit for Q4?	InspectForecastCommit
6	Explore capabilities	Explore what the Revenue Deal Room agent can do.	ExploreAgentCapabilities

Coherent mock data and integrations

The deterministic graph must include at least 80 accounts across regions and segments; 260 contacts; 120 opportunities over 8 quarters; activities, meetings, mail-shaped evidence, documents, requirements, competitors, products, price books, currencies, stakeholder roles, relationship signals, mutual plans, proposals, scenarios, policies, exceptions, approvals, forecast snapshots, wins/losses, and outcome data. The hero record is Contoso Global Expansion / ZDR-2042 across all components.

Dates use one invocation clock; currency and numbers use Intl. Confirmed actions enter a session-only overlay; Reset restores seeds. Mock services map to typed future adapters:

Contract	Potential live systems
IRevenueDataService	Dynamics 365 Sales, Salesforce, custom CRM/CPQ
IWorkContextService	Microsoft Graph, Work IQ, Teams, SharePoint
IPricingAndPolicyService	Dynamics 365, ERP/CPQ, finance rules, contract policy
IWorkflowService	Power Automate, CRM workflow, approvals, e-signature handoff

Production authorization, territory access, price controls, revenue recognition, legal approval, records retention, data residency, and audit remain system-of-record responsibilities.

Agentic, global, and responsible requirements

• One typed catalog owns identities, schemas, prompts, negative routing boundaries, previews, and routes.
• Generate each approved component through Yeoman after plan lock; never copy or rename scaffolds.
• Copilot provides bounded intent; components own records, math, validation, review, state, and receipts.
• Never treat sentiment, inferred authority, relationship strength, or close probability as fact.
• Do not autonomously send customer communication, change CRM, set stage/forecast, approve discount, accept terms, or create a quote/contract.
• Externalize strings; validate English, German expansion, Japanese, Arabic RTL, multiple currencies, fiscal calendars, time zones, name/address structures, and regional commercial policy packs.
• Test keyboard, screen reader, forced colors, reduced motion, 200% zoom, narrow inline, mobile full screen, long localization, and chart/table equivalence.
• The shipped showcase must be offline, deterministic, and explicit that actions are session-only mocks.

Planned delivery gates

• Gate 0 - Plan lock: approve README, 25-component catalog, routes, hero scenarios, identity, market baseline, and non-goals; synchronize to GitHub.
• Gate 1 - Creation plan: copy approved creation rules and write todo.md, routing matrix, metadata/GUID plan, demos, and evidence plan before code.
• Gate 2 - Domain graph: approve typed graph, commercial math, forecast logic, policies, transitions, session overlay, reset, and unit tests.
• Gate 3 - UX proof: implement MapBuyingCommittee, BuildMutualActionPlan, SimulateCommercialOffer, ReviewDealException, and exact Deal Room continuation first.
• Gate 4 - Connected hero: run Contoso from risk through access, plan, meeting updates, offer, exception, and forecast receipt across personas.
• Gate 5 - Catalog: preserve the original 21 identities and add the three inline workflow tools; validate prompt routing collisions and capability preview safety.
• Gate 6 - Quality: accessibility, global formats/RTL, themes, responsive captures, tests, provenance, generated plugin, package audit, and clean offline rehearsal pass.
• Gate 7 - GA release: package, screenshots, five-minute keynote, business and technical demos, telemetry/outcome plan, limitations, support owner, and release evidence approved.

Explicit non-goals

• Replacing CRM, CPQ, contract lifecycle management, forecasting, e-signature, or revenue recognition.
• Autonomous prospecting, customer contact, CRM writes, stage changes, pricing, terms, or forecast commit.
• A regional sales KPI dashboard, generic account summary, contact list, or generated proposal document.
• Unsupported claims of buyer intent, close probability, legal approval, or financial accuracy.

Definition of done

The future implementation is complete only when all 26 operational inline components are independently valuable; the capability explorer safely previews them; every full-screen lens has a useful default; the three hero scenarios run on one coherent graph; buyer evidence and inference are distinguishable; commercial and forecast math is inspectable; consequential actions use human review, confirmation, and receipt; exact continuation preserves context; accessible/global evidence passes; and an audited offline package deploys without runtime data dependencies.

The locally scoped specialist implementation is validated against the current contract above, not every aspiration in this historical brief. Tenant-host checks remain explicitly separate because they require an authenticated Microsoft 365 environment.