ParamountIntelligence
Discuss a project
Client case study · Fortune 500 industrial automation company · North AmericaWork in progress · ongoing engagement

AI governance engineers barely notice.

An enterprise AI intake and governance platform, built natively on Microsoft 365, so about 80 engineers can log AI ideas in under two minutes and leadership can see every one from submission to deployment.

  • Two-minute intake: a short form, identity from sign-in, and no approval gate before building.
  • One register: every idea, owner, stage and support need in one SharePoint list, with full history.
  • Told in Teams: every stage change reaches the engineer where they already work.
Submit a sample idea See the story
ClientFortune 500 industrial automation company
EngagementParamount engagement
Built onMicrosoft 365 · Power Automate · Power BI
One idea's journeysanitised data
Design Review CheckerEngineering · Vertex AI / GeminiSubmitted

Meet our client

Client
Fortune 500 industrial automation company, the client's innovation hub
Industry
Industrial automation & engineering technology
Market
North America · about 80 engineers

In a nutshell

The hub's engineers originate and own their AI projects. In an ongoing engagement, we are building and advising on the intake and governance platform that lets them register ideas in under two minutes, gives the governance team one place to triage and support them, and shows leadership the whole portfolio, all inside the client's own Microsoft 365 tenant.

Client data protectedNames, ideas, counts and screens on this page are sanitised or changed. The workflow is real; the data is not.
01 · The problem

Eighty engineers. Ideas everywhere. No single view.

The hub wanted engineers to own their AI projects. That left leadership unable to answer four basic questions. Tap one to see the gap.

The client's innovation hub
About 80 engineers across seven business areas · positions and counts are sanitised
What was missing
  • Visibility: no single place captured every AI idea, so leadership could not see what was built, on which platform, or by whom.
  • A support path: a stuck engineer had no defined way to ask for help, and nobody could spot them.
  • Sensitive ideas: some touch confidential R&D, customer or export-controlled material and needed review without slowing everyone else.
  • Ownership kept: the answer could not be a central team building tools and pushing them down.
02 · The rules

The sponsor set the rules before any design

Every later decision follows from six rules. Tap a rule to see what it ruled out and what we did instead.

The success testAn engineer must never experience the process as bureaucracy.
Registering an idea is not authorising it
  • A record, not a permission: registering makes an idea visible and lets the engineer ask for support. It does not authorise the work.
  • Stages track progress: Triaged, Approved and the later stages show where an idea stands. An engineer does not wait for any of them before starting to build.
  • Review where it is needed: ideas that may touch confidential, R&D, customer or export-controlled material are flagged at intake for the governance owner to review.
  • The minimum layer: it is not an enterprise AI governance suite. It adds intake, tracking, support and reporting inside the client's existing security boundaries, licensing and administration.
03 · Steps 1 and 2 · Intake

Two minutes in. A few seconds behind the screen.

Fill in the form like an engineer would. Follow-up questions appear only when they apply. Submit it and watch the intake flow run.

AI Idea SubmissionMicrosoft FormsSigned in as Ryan Brooks · name and email recorded
"Not sure" is treated as Yes and sent for review.
Real form: 7 of 9 questions showing · target under 2 minutes
AI Intake · Form to Registrywaiting for a response
    Why intake is built this way
    • Nine questions, mostly hidden: four core questions plus identity; the "other tool" and "what kind of help" questions appear only when they apply.
    • A standalone form on purpose: a list-bound form would skip the flow, losing the identity lookup, the sensitivity check, the messages and the failure alert.
    • Ownership in one variable: the governance owner is read from one setting, so it can change without rebuilding any step.
    • Never lost silently: a parallel branch alerts the governance owner if the SharePoint write fails.
    04 · Steps 3 to 5 · Triage, pipeline and Teams

    Seven stages. Drag a card, and everyone hears.

    Drag a card to another column, or tap a card and then a column, as the governance owner would. Engineers can see the board but cannot move cards. The move is validated, saved to SharePoint and announced in Teams.

    Write-back onSanitised data
    Write-back trace
      What the engineer sees in Teams
      How a move is saved and announced
      • A board that writes back: Power BI is normally read-only, so a dropped card goes to an Azure Function that rejects unknown IDs and any status outside the seven.
      • One timestamp everywhere: a Power Automate flow stamps one UTC time and saves it to SharePoint, so the card and the record always agree.
      • Teams, not email: a status-change flow posts the matching message; it ignores items at Submitted, so the intake write never fires it.
      • Honest failures: if a save fails, the board reports an error and asks the user to refresh, because a timed-out save may still have completed.
      Following one idea that asks for help
      Step 1 · IntakeThe engineer asks for helpEngineerAnswers the support question, for example "Need help to get started", then picks the kind of help: technical solution architecture, deployment or stakeholder communication.
      Step 2 · AlertBoth sides are told in TeamsIntake flowThe engineer gets a confirmation. The governance owner gets an alert that includes the support answer, and the idea counts toward Needs support on the dashboard.
      Step 3 · TriageA support contact is assignedGovernance ownerReviews the idea and assigns a support contact, who can edit the ideas they are helping with. The assignment appears in the activity feed.
      Step 4 · Approved to DeployedThe engineer hears every moveStatus change flowAt Approved, the Teams message names the support contact, who will reach out to plan the first build steps. In Build and Deployed arrive the same way.
      05 · Step 6 · Sensitive data review

      Flag the few. Don't stop the many.

      Answer the sensitivity question as an engineer, then act as the governance owner. Watch who can see the record.

      The engineer answers Would it touch confidential, R&D, customer or export-controlled material?
      Who can see "Supplier Quality Insights"
      Audit trail
        How a sensitive idea is handled, in order
        • 1 · The engineer's answer sets the flag: any answer other than No, including Not sure, sets it. Over-flagging is the safer error.
        • 2 · The intake flow raises it: the governance alert starts with [SENSITIVE DATA REVIEW] and the idea lands in a dedicated Sensitive Review view for the governance owner.
        • 3 · The governance owner restricts it when needed: item-level permissions limit the record to governance owners and the assigned support contact. The wider engineering community can no longer see it.
        • 4 · Same register: sensitive ideas stay in the one list, so reporting and the lifecycle work the same way.
        06 · Step 7 · The leadership dashboard

        The question the sponsor started with, answered

        What AI work exists, on which platform, built by whom, and where it stands. This dashboard reads the same register as the board you just moved cards on.

        Ideas by status
        9ideas
          Ideas by business area
          Sanitised data Counts come from measures over the register
          What leadership gets
          • At-a-glance counts: total ideas, and those needing triage, needing support, in build, deployed and under sensitive review.
          • A built-in check: with no status filter, the seven stage counts add up to the total; that is how the team checks the numbers.
          • A live activity feed: status changes, Teams messages, new submissions, flags and support assignments, with who or what did each.
          • An inventory of shipped AI: deployed ideas form a running list of tools the hub can show to the wider business.
          Register and dashboard: who uses which
          • The register is the record: every idea, owner, stage, support need and triage note lives in the AI Idea Registry SharePoint list, with version history.
          • The dashboard is a read-out: Power BI counts, charts and the board are built over the register. A card move is written back to the register.
          • Engineers: submit in Microsoft Forms and hear back in Teams. They can view the dashboard, views and board, but cannot edit items or move cards.
          • Governance owners: manage the portfolio. They triage, set priorities, assign support and set permissions on register items, and move cards on the board.
          07 · Under the hood

          A closed loop around one SharePoint list

          No separate database, no external API, no third-party SaaS. Pick a direction to trace how data moves.

          1 · Capture
          2 · Process
          3 · Store
          4 · Present and notify
          Built to be owned by the client
          • Three flows and one function: intake, status-change notification and board write-back, plus the UpdateIdeaStatus function in the client's Azure subscription.
          • Standard licensing: intake, register and notifications run on E3/E5, with no external SaaS, no Dataverse and no Copilot Studio dependency.
          • Details that matter: column names and choice values match the form character for character, or SharePoint rejects the write.
          • Handed over in full: a technical design document and a rebuild guide with a check at every step, so the hub runs it without Paramount.
          Proof

          Inside the platform

          Product screens with sanitised data. Tap any screen to see it full size.

          08 · Expected results

          One register. No bureaucracy.

          Work in progress

          An ongoing engagement. These are the outcomes the platform is designed to deliver as rollout continues across the hub.

          ~80engineers with one place to register AI ideasExpected
          < 2 minto submit an idea, from tools engineers already useDesign target
          0approval gates before an engineer starts buildingBy design
          6purpose-built dashboard views, from Needs Triage to Sensitive ReviewIn scope
          Designed to deliver
          Engagement scope · build and advisory
          The expected resultLeadership sees grassroots AI activity as it happens, engineers who need help get a named contact, and nobody waits on an approval step.
          Microsoft FormsSharePoint Online ListsPower AutomateMicrosoft TeamsOffice 365 UsersPower BIAzure FunctionsMicrosoft 365 E3/E5
          What comes next
          • Automated last-reviewed dates: under a dedicated service account with a loop guard.
          • Smarter routing: reviews sent to different governors by business area or sensitivity, or through a shared governance channel.
          • Richer Teams messages: Adaptive Cards with headers, fact tables and action buttons, and KPI tiles refreshed by a scheduled flow.
          • AI-assisted triage: suggested titles and problem summaries through an approved model endpoint.
          AI governance, without the bureaucracy

          See every AI idea. Slow none of them down.

          An intake and governance layer built on the Microsoft 365 tools your teams already use.

          What a first conversation covers
          1. 01Your business objective
          2. 02The technical context
          3. 03The right next step
          Discuss a project View all case studies →