How we work, in six straight answers.

Red flag #1

Who actually delivers the work?

The engineers named on the proposal are the engineers who ship the work. No bait-and-switch to a junior bench after signature — the team is small and senior by design.

Every proposal names the people, their roles, and what each is accountable for. You can meet them before you sign.

Red flag #2

How are we billed, and what happens if scope changes?

Milestone billing, 30 / 40 / 30: 30% at kickoff, 40% at the midpoint demo, 30% at delivery — which includes 30 days of post-launch support. Your money tracks delivered work, never hours claimed.

Scope changes are re-scoped in the open at the midpoint demo. You will never discover a surprise invoice.

Red flag #3

Where does our data go?

Your infra, your keys, no public-API leakage by default. AI features are architected so your data stays on infrastructure you control; a written DPA is available on request.

For AI engagements we state in the proposal exactly which models see which data, and what is logged where.

Red flag #4

What happens after launch?

30 days of post-launch support is in the price of every tier — not a paid add-on. Bugs, tuning, and handover questions are covered while your team takes the wheel.

After that: a support agreement if you want one, or a clean handover if you do not. The codebase is documented for your in-house engineers either way — 100% code ownership, always.

Red flag #5

Have you actually shipped real AI systems?

Three-plus production AI systems in the last 18 months — including two we run ourselves: TalkDrill (50,000+ users, 4.9★, real-time voice AI) and PenLeap (10,000+ UK students, 95% pass rate, AI rubric scoring).

Running our own AI products means we live with the consequences of architecture decisions — latency budgets, model costs, moderation, uptime — not just the demo day.

Red flag #6

Is "AI-fast" real, or marketing?

Real, and here is the mechanism: we build with Claude-class coding agents, AI-assisted scaffolding, and generated test suites — the same class of AI we ship for clients. That is how an AI MVP takes 8 weeks against the 6-month traditional quote.

What stays human: architecture, security review, data-handling decisions, and every merge — senior engineers review all AI-drafted code before it lands. Fixed pricing keeps the incentive honest: compression shows up in your quote and calendar, not our margin.

Own-history proof: TalkDrill voice pipeline, 11 weeks to production. 365 automated publishing workflows last year. Client portal, live in a month.