PROJECT [CODENAME]

[One-line product descriptor]

[Scope line: what moves from where to where, or which system is replaced]

Executive Summary

[Author name] [Role, programme, business unit] [Country]

[Second author name] [Role, secondment or reporting line] [Country]

[Month Year] Classification: [Internal / Commercial in Confidence]

\newpage

Financial Summary

[Two to three sentences: what was delivered, in what elapsed time, with what team size, versus what benchmark. State the multiple.]

[Two to three sentences: blended rate per developer-month applied to both scenarios, total delivery cost, benchmark cost range, saving in absolute terms and as a percentage. Monthly tooling cost.]

Team, Duration and Cost Comparison

Approach	Team	Duration	Total Delivery Cost
Traditional (Benchmark)	[n developers]	[n months]	[range]
AI-Assisted (Actual)	[n developers]	[n weeks, with dates]	[actual, approximate]
Savings	[multiple]	[multiple]	[absolute and percentage]

Cost assumption. [Blended rate per developer-month, applied to both scenarios. What the AI-assisted total includes (personnel plus named tooling). Anything excluded as a customer-side operating cost.]

Provenance of the benchmark. [COCOMO II, internal benchmark against comparable projects, or the client’s own estimate. Say which.]

Time Investment Analysis

Phase	Traditional Estimate	AI-Assisted Actual
Requirements and Analysis	[weeks]	[actual]
Design, PRD and Implementation Plan	[weeks]	[actual]
Foundation (repo, API, frontend, database, CI)	[weeks]	[actual]
[Core feature block]	[weeks]	[actual]
[Second feature block]	[weeks]	[actual]
Quality Polish and Hardening	[weeks]	[actual]
Testing and QA	Separate QA phase	Continuous (included)
Documentation	Separate phase	Generated alongside code
TOTAL	[total]	[total]

A note on the elapsed time. [Holidays, late secondments, environment access delays. Explaining lost time strengthens the number.]

\newpage

Part 1: Project Overview

What is [Codename]?

[Three to five sentences: what it does, who operates it, what it replaces, how it is deployed and why.]

The Problem We Solved

• [Pain point one, quantified]
• [Pain point two, quantified]
• [Structural or data-model problem that makes the work hard, not just tedious]
• [Typical cost or effort per customer or per project]
• [Commercial consequence]

Our Solution

• [The central design idea in one line]
• [Deployment and platform fact]
• [Frontend fact with concrete counts]
• [The pipeline or workflow that carries the main use case]
• [How AI is governed inside the product, or “AI was the build engine only; the product calls no model at runtime”]
• [The data-integrity or auditability property a reviewer would ask about]

Part 2: Scale of Work Completed

Figures captured from the repository on [date]. The product is approximately [n] percent complete. The remaining [n] percent covers [remaining scope], itemised in Part 10.

Metric	Value	Industry Context
Total Lines of Production Code	[n]	[size comparison a non-engineer can picture]
Backend [language]	[n]	[across n production files]
Frontend [language]	[n]	[framework and build tool]
Total Commits	[n]	[start date, average per day]
Pull Requests	[n]	[percent with independent human review before merge]
Issues Tracked	[n]	[structure of the backlog]
Database Migrations	[n]	[versioned schema evolution]
API Endpoints	[n]	[API style]
Frontend Components	[n]	[reusable building blocks]
Frontend Pages	[n]	[full application screens]
Backend Test Cases	[n]	[unit and integration approach]
Frontend Test Assertions	[n]	[across n test files, tooling]
Documentation Lines	[n]	[what the documentation covers]

Part 3: Technical Architecture

Backend Architecture

Component	Technology	Why This Choice
Runtime	[platform and version]	[operational justification]
API Style	[style]	[justification]
Data Access	[technology]	[justification]
Database	[engine and version]	[justification]
Logging	[library and format]	[justification]
Security	[library]	[hashing or token strategy]

Frontend Architecture

Component	Technology	Why This Choice
Framework	[framework and version]	[justification]
Language	[language and version]	[justification]
Build Tool	[tool]	[justification]
Component Library	[library]	[justification]
Styling	[approach]	[justification]
Routing	[library]	[number of routes]

AI Runtime Architecture

[Delete this subsection and its table if the product calls no model at runtime, and say so in one sentence in Part 4.]

Component	Technology	Why This Choice
Gateway	[gateway]	[key management and policy centralisation]
Reasoning Model	[model]	[tasks it is used for]
Bulk Model	[model]	[cost multiple versus the reasoning model]
Concurrency	[worker range and retry policy]	[rate limits respected]

Infrastructure

Component	Technology
Containerisation	[technology and composed services]
Staging	[environment and its purpose]
CI/CD	[pipeline and what runs on each push]
Code Quality	[linters, analysers]
Integration Testing	[real dependencies or mocks]
End-to-End Testing	[tool and coverage]

Part 4: The AI Approach

AI for Building the Product

Engine: [tool and licence type]

[What the tool did across the lifecycle, how work was structured, the independent human review rate, merged PR count and elapsed period.]

AI Inside the Product

Engine: [gateway or provider, and what it fronts]

[One paragraph.]

• [Where keys live]
• [How a provider or model swap is performed]
• [How cost is routed and measured]

Key AI-Design Decisions

Criterion	Decision	Rationale
Security	[decision]	[what it protects]
Cost	[decision]	[how it is measured]
Speed	[decision]	[quantified improvement]
Reliability	[decision]	[failure mode it handles]
Flexibility	[decision]	[lock-in avoided]
Quality	[decision]	[how the model was selected]

Part 5: Security and Integration

Deployment Model

[Where it runs, what data crosses which boundary, which regulatory expectation that satisfies.]

Security Layers

Layer	Implementation
Authentication	[mechanism]
Authorisation	[permission model and where enforced]
AI Traffic	[routing, any direct external calls]
Rate Limiting	[where applied]
Upload Safety	[caps]
Correlation	[correlation ID propagation]
Audit Trail	[what is recorded on every change]
Provenance	[field-level source labels]

[The Human-in-the-Loop Rule, or equivalent control]

[The rule, where it is enforced, how AI suggestions are surfaced.]

Part 6: Core Features

1. [Primary workflow or pipeline]

[One paragraph plus the stages in order.]

2. [Domain model or dependency structure]

[What it does, how it is enforced, why it matters.]

3. [AI-powered capability, or the next major feature]

[What it does, how it is enforced, why it matters.]

4. [Review or approval workflow]

[What it does, how it is enforced, why it matters.]

5. [Provenance or data-quality feature]

[What it does, how it is enforced, why it matters.]

6. [Audit trail]

[What it does, how it is enforced, why it matters.]

7. [Export, integration or reporting]

[What it does, how it is enforced, why it matters.]

Part 7: The Invisible Excellence

Testing

[n] automated test cases across [n] test files, with [integration testing approach].

Layer	Location	Assertions / Cases
Backend unit	[path]	[n]
Backend integration	[path]	[real dependency approach]
Frontend	[tooling]	[n across n files]
End-to-end	[tool and path]	[coverage]

Code Quality

• [Type-safety settings]
• [Linting and static analysis]
• [Review rule and independent human review rate]

Observability

• [Logging format and correlation]
• [Cost or usage dashboard, if AI runs at runtime]
• [Tracing]

Documentation

Approximately [n] lines of developer and product documentation, generated alongside code rather than as a separate phase. Key artefacts include [three or four with line counts]. See the Appendix.

Part 8: Challenges and Mitigations

Technical

[Challenge one: the quality or accuracy failure]

[The problem in numbers. Root cause. The fix, and whether it was a prompt change or a new boundary between deterministic code and the model. Result against target.]

[Challenge two: a reliability or integration failure]

[Symptom, silent failure mode, mitigation and new instrumentation.]

[Challenge three: performance at scale]

[Workload size, observed time, cause, quantified improvement.]

[Platform or vendor gaps]

[What works, which specific capability would improve fit.]

Organisational

[Distribution, time zones or working conditions]

[Where each person worked, synchronous time, disruptions, how the workflow absorbed them.]

[Knowledge gaps at kickoff]

[What each person did not know, how long it took to close, what closed it.]

[Scope or requirements uncertainty]

[What was undefined, the decision taken, whether it paid off.]

Part 9: Value Delivered

Internal Delivery Savings

Item	Traditional	With [Codename] (Actual)
Team size	[n]	[n]
Elapsed duration	[months]	[weeks]
Personnel cost	[range]	[actual]
Tooling cost	[baseline]	[actual]
Runtime AI cost	Not applicable	[per unit, customer-side, or none]
External partners	[typical]	[used]
Total delivery cost	[range]	[actual, percentage saving]

Customer-Side Value

Process	Before	After	Improvement
[Specialist effort per project]	[baseline]	[new]	[saving]
[Data quality]	[baseline]	[new]	[what it enables]
[Safety or correctness property]	[baseline]	[new]	[why safer]
[Onboarding a new source or customer]	[baseline]	[new]	[time saved]

Commercial Value Case

The value case presented at [forum and date] estimates approximately [figure] per year, driven by:

• [Monetised customer effort reduction, figure and horizon]
• [Additional projects per year, figure and horizon]
• [Competitive differentiation in tenders]

Risk Reduction

• [Safety]
• [Regulatory]
• [Security]
• [Vendor lock-in]
• [Knowledge]

Part 10: Next Steps

Remaining [n] Percent of Scope

• [Deliverable, owner, target period]
• [Deliverable, owner, target period]
• [Customer pilots or onboardings]
• [Remaining test coverage]

First User Feedback

[Two or three sentences, the environment it came from, whether it agrees with the measured results.]

Appendix: Documentation Inventory

Key documents in the repository, by size. Figures are line counts.

Document	Lines	Purpose
[document]	[n]	[purpose]

Additional Resources (within the repository)

• [Agent context files]
• [Presentation decks with dates]
• [Kickoff material]

Document Version: 1.0 Last Updated: [Month Year]