How we ship client products in 38 days (not 120) - the operating model behind it

Two years ago, Ailoitte delivered a production-ready SaaS product in 34 days on a fixed-price contract. The client had budgeted four months.

That wasn't a fluke. We've repeated it across 300+ products in 21 countries, with a median of 38 days. Here's the honest breakdown of how.

The problem with time-and-materials

In 2023, we ran T&M contracts. The structural problem: our revenue went up when projects took longer. When AI started compressing implementation time by 40–50%, we had a choice: quietly pocket the efficiency gain or restructure the model to reflect what work actually cost.

We restructured.

What we built: the AI Velocity Pod model

Small cross-functional units, 3–6 humans, governing a structured layer of specialised AI agents.

• Pod Lead — defines intent and specs precise enough for agents to execute without ambiguity
• Review Engineer — reviews diffs at milestones against acceptance criteria, not line by line
• QA Orchestrator — runs a continuous agentic QA pipeline at the commit level, not sprint end
• Domain Expert — carries business logic and edge case knowledge that AI can't infer

Agents handle code generation, refactoring, test suites, and first-pass review.

What fixed price is actually required internally

Scope discipline. Every engagement starts with a discovery sprint producing one document: exactly what ships, what doesn't, and what triggers a change order. Writing this precisely is harder than the development that follows. It's also what makes 38 days repeatable.

A different P&L. Under fixed-price, margin = fixed price minus cost to deliver. Every efficiency gain from AI goes directly to margin. We profit from shipping faster. That alignment is what keeps the model honest.

What failed before it worked

• Vague scope documents — ambiguities surfaced mid-sprint during deadline negotiations. Fixed with a mandatory second-pass scope audit checklist.
• QA deployed after the agents — We shipped faster but caught defects later. Fixed by making QA infrastructure week zero, always.
• No formal change order process — small mid-sprint additions eroded the model from within. Fixed with a written protocol: any scope addition is a new engagement with its own price and timeline.

The numbers

• Median delivery: 38 days across 300+ products
• Change order rate: under 15%
• Cost to clients: 70–85% lower than traditional agency engagements
• Countries shipped: 21

One open question

We've refined the Startup MVP Velocity model across 300+ engagements, but one challenge stays active: clients whose procurement systems are built for hourly billing.

How do you handle buyers whose internal systems ask for an hourly rate when you're operating a fundamentally different model?

Curious what others have found.