01 / SpecifySix editable Word files
AI Feature Spec Pack
Document the use case, controls, edge cases, stakeholder explanation and gate decisions. Includes a completed fictional example.
£99Buy now
SpecRoom · Enterprise Edition
Build a defensible process for the AI work on your desk.
Practical tools for financial-services teams: specify a feature, assess an LLM, and make a clear governance decision.
Spec Pack · AI Product Manager Playbook · LLM Evaluation Scorecard · Complete Bundle
Actual pages from the AI Feature Spec Pack · Click “See Spec Pack pages” for more
The work behind the decision
Give every review a clear record.
Define the problem, document the model boundary, test what can fail and record who can approve the next stage. The pack gives teams a consistent structure for that work.
The launch collection
Three products.
One practical workflow.
Built for data, risk and analytics professionals making AI product decisions inside banks and asset managers today.
02 / LeadPractical playbook
AI Product Manager Playbook
A working guide to product decisions, data and vendor oversight, testing, stage gates and regulatory applicability, with a fictional case.
£79Buy now
03 / EvaluateGuide + Excel workbook
LLM Evaluation Scorecard
Set pass conditions, log test results, compare options and record the governance decision using a scorecard and editable workbook.
£69Buy now
The complete collection
Complete Spec Room Pack
Get the Spec Pack, AI Product Manager Playbook and LLM Evaluation Scorecard with its workbook together.
Inside product 01
The Spec Pack.
Six working documents.
Built to be completed by the delivery team and reviewed against the firm’s own controls. The worked example is fictional and clearly labelled.
AI Feature Spec Pack · 01–06
A working system for AI feature governance.
From a measurable use case through edge-case testing, stakeholder explanation and production sign-off.
Enterprise Edition£99
Six editable Word documents, including a fictional worked example. Adapt the evidence, controls and assessments to your firm.
- 01AI Feature Spec Template
- 02Worked Example
- 03Edge Case & Bias Register
- 04AI Prompt Design Brief
- 05Stakeholder Explainability Script
- 06Rollout & Sign-off Checklist
Document previews
See the actual pages.
Three pages from each of the six supplied Word files. Select a page to enlarge it, or read six interior pages in the accessible PDF excerpt.
Open the six-page Spec Pack sample PDF
01 / editable Word file
AI Feature Spec Template
A reusable specification with problem, suitability, risks, metrics and gate decisions.
02 / editable Word file
Worked Example
A completed fictional regulatory-document classification case.
03 / editable Word file
Edge Case & Bias Risk Register
Record edge cases, test outcomes, bias risks and affected populations.
04 / editable Word file
AI Prompt Design Brief
Version prompts, describe their constraints, fallback and test cases.
05 / editable Word file
Stakeholder Explainability Script
Prepare a risk-committee explanation with evidence and a clear decision ask.
06 / editable Word file
Rollout & Sign-off Checklist
Track four decisions from discovery through production, with owners and evidence.
Inside product 02 · Lead
A practical playbook for AI product decisions.
Eight chapters follow the work from product framing to launch. Six further sections provide records, checklists and a fictional pilot example.
Contents · Practitioner Playbook
- The product decision before the model
- Map the operating model
- Data quality and lineage
- Choose and govern the model or vendor
- Test for decisions rather than demonstrations
- Use stage gates as decisions
- Regulatory applicability triage
- Launch and operate
Working tools: one-page decision record, example decision narrative, stakeholder map, vendor governance checklist, independent validation brief and fictional pilot scorecard.
Includes a fictional reconciliation-assistant case. Example figures illustrate the method and are not real performance benchmarks.
Inside product 03 · Evaluate
Make an LLM choice you can explain.
A seven-part Word decision record and a four-tab Excel workbook. The supplied workbook is set up to compare two candidates.
Contents · Evaluation Scorecard
- Define the decision and evidence
- Set hard gates before scoring
- Design the test set
- Score observed performance
- Capture each test result
- Review supplier and operating model
- Record decision and monitoring
Excel tabs: Comparison, Criteria, Test log and Governance. Compare two named candidates against six weighted criteria and six evidence gates. The test log has 36 prefilled model-case rows: 18 paired scenarios, one row for each candidate per scenario. Additional blank rows can be filled in.
Inside the Excel workbook
One decision, two candidates, six criteria.
The weights shown are illustrative and should be agreed before testing. A high weighted score does not override a failed governance gate.
| Criterion | Weight |
|---|---|
| Task accuracy | 25% |
| Safety and misuse | 20% |
| Tone and instruction fit | 10% |
| Latency and reliability | 10% |
| Total operating cost | 10% |
| Governance suitability | 25% |
A sequence you can audit
The right evidence at each gate.
01 / Frame
Discovery
Define the problem, baseline and a credible non-AI option.
02 / Specify
Design
Set use boundaries, data responsibilities and proposed controls.
03 / Validate
Build & pilot
Record tests, edge cases, monitoring and human oversight.
04 / Decide
Production
Review pilot evidence and record the authorised decision.
About the founder
Built from financial-services delivery experience.
Spec Room is developed by Matisse Ghaddaf, a risk and analytics delivery practitioner. The pack gives teams a structure for recording design choices, controls, evidence and governance decisions.
Career experience is provided as background. Named employers and clients do not endorse Spec Room.
State StreetRisk and analytics practice leadership, including data controls and delivery across investment analytics.
RBC BlueBayRisk business analysis and data product ownership during an investment risk platform transition.
Insight InvestmentData product ownership focused on access controls for investment data.
Janus HendersonRisk business analysis and delivery across regulatory reporting and investment risk.
A note from the founder
“I built Spec Room around a familiar delivery problem: a promising feature reaches governance review, but the evidence, controls and decision record are scattered. These documents give teams a clearer way to prepare that discussion.”
Matisse Ghaddaf · Founder, Spec Room
Practitioner led
Informed by experience delivering data and analytics work in financial services.
Designed for review
Prompts for evidence, risk ownership, testing and gate decisions.
Firm-specific assessment
Helps users assess and document applicable obligations; it does not establish compliance.
Designed for scrutiny
What the pack helps you record.
Use it as a starting framework, then attach the firm’s actual evidence, risk decisions and approvals. Templates do not replace independent validation or legal advice.
- ScopeIntended use, prohibited use and responsible owners.
- EvidenceBaselines, metrics, test references and limitations.
- ControlsHuman review, fallback, monitoring and rollback.
- DecisionsGate conditions, authority and record location.
See the documents.
Judge the standard.
Inspect the Spec Pack pages above. The Playbook and Scorecard extend that work into product decisions and model evaluation; the bundle brings all three together.
About Spec RoomThe six documents are editable starting points. Your firm supplies its own evidence and authorises its own decisions.
Buy the complete pack — £179 →
Six editable Word documents · Instant download · All three products