Our AI consulting methodology is one loop: diagnose before prescribing, build with a human in the loop, transfer everything, exit clean.
The Method, Stated Plainly
Blankpage's AI consulting methodology runs in four phases: Listen & Map, Design Sprint, Build & Guide, and Transfer & Exit. Every phase has a fixed scope and a dated deliverable, and every artifact — code, models, documentation — belongs to the client at exit.
Most methodologies are decks. Ours is a sequence of deliverables you can hold. Working software over slideware is not a preference; it is the acceptance criterion.
The Four Phases
1. Listen & Map — one to two weeks. We read the code, query the data, and interview the people who run the system, because every system is socio-technical: half software, half habit. Output: a systems map and a problem statement you sign before anything gets built.
2. Design Sprint — runs inside the two-week Diagnostic Sprint. We prototype the riskiest assumption first, against your real data. Output: a working prototype or test harness, a findings memo, and a build/no-build recommendation with a fixed price.
3. Build & Guide — six to twelve weeks. Weekly demos of running software, decisions logged with thinking traces, scope changes only through a new signed scope. Output: a system in production that meets the acceptance criteria from phase two.
4. Transfer & Exit — the final week. Repositories, model weights, prompts, pipelines, runbooks, and a training handoff move to your team. Output: your engineers operating the system without us.
Human-in-the-Loop, as Policy
Autonomy is a dial, not a religion. We set it per decision, not per system: draft-only where stakes are high, full autonomy where errors are cheap and reversible.
Every consequential action an automated system takes has a named human owner and an audit trail. Human-in-the-loop AI is not a caveat for lawyers; it is how systems stay correctable.
What You Own at Exit
Everything. Source code, model weights, fine-tuning data, prompt libraries, infrastructure definitions, runbooks, and the decision log. Full IP transfer is the default contract, not an upgrade.
If you need us forever, we built it wrong. The three engagement models are designed around that sentence.
How the Methodology Maps to Engagements
The two-week Diagnostic Sprint covers Listen & Map and the Design Sprint. Design & Build covers phases three and four. Strategic Advisor keeps the loop honest after exit — ongoing, cancelable, no lock-in.
The same AI consulting methodology runs across all eight disciplines, from private LLMs to incident response. Different domains, same loop.
The commitments behind the method live in the manifesto. Common objections are answered in the FAQ.