← Resources

Illustrative deliverable

AI Decision Brief and 90-Day Roadmap

Synthetic scenario: triaging research-funding opportunities for a university research office.

This is not client work.

The organisation, evidence and recommendation below are fictional. The example demonstrates the structure and level of specificity of a Discovery & Governance Sprint deliverable without implying a client, testimonial or completed result.

Executive decision

Proceed with a bounded triage pilot

Test whether a retrieval-based assistant can reduce the time spent finding, classifying and routing funding opportunities. Do not use the system to assess proposal quality, write reviewer feedback or make funding decisions. Continue only if the pilot improves routing speed without reducing factual accuracy or staff confidence.

1. Decision context

Current state

Research-office staff monitor multiple funder sites and mailing lists, manually interpret eligibility, and route opportunities to departments. Information arrives in inconsistent formats and relevance depends on local expertise.

Decision required

Whether a small AI-enabled workflow could improve discovery and routing without creating unreliable advice, displacing judgement, or exposing confidential research plans.

2. Use-case assessment

Synthetic assessment of four AI use cases by value, feasibility, risk and recommended decision
Use caseValueFeasibilityRiskDecision
Opportunity triage and routingHighHighLow–mediumPilot
Eligibility and deadline extractionMediumHighLowInclude
Drafting applicant feedbackMediumMediumHighDefer
Automated funding decisionsUnclearLowVery highDo not pursue

3. Governance boundaries

Permitted in the pilot

  • Public funder guidance and published opportunity data.
  • Extraction of deadlines, themes and stated eligibility.
  • Human-reviewed routing suggestions with source links.
  • Logging of corrections and uncertain classifications.

Outside the boundary

  • Confidential proposals or unpublished research plans.
  • Eligibility claims without a cited funder source.
  • Ranking researchers, proposals or departments.
  • Autonomous notifications or decisions.

4. Pilot specification

Users
Two research-office staff and representatives from two academic departments.
Test set
A fixed sample of recent public opportunities with an independently checked reference classification.
Success measures
Routing accuracy, time per opportunity, correction rate, source traceability and user confidence.
Stop conditions
Unsupported eligibility claims, systematic missed opportunities, or a workload greater than the current process.

5. 90-day roadmap

  1. Days 1–30, evidence: establish the reference set, confirm data boundaries, document the current baseline, and agree decision ownership.
  2. Days 31–60, pilot: build the smallest viable workflow, test with named users, and record every correction and exception.
  3. Days 61–90, decide: compare results with the baseline and choose to stop, revise, or move into controlled implementation.

Need this level of clarity for a real workflow?

Start with a free fit check. I will tell you whether a Discovery & Governance Sprint is proportionate before asking you to complete the full intake.