What is The Foundry Method?
The Foundry Method is an evidence-gated AI consulting framework for moving from broad AI ambition to a decision a sponsor can fund and an engineering team can act on. It is designed to run repeatedly across clients, portfolio companies, business units, and functional teams without turning every engagement into a new spreadsheet and deck.
The six-step framework
- Scan — start with a company URL and automatically gather dated public evidence about customer, competitive, regulatory, supplier, talent, and capital pressure.
- Source — add company pages, documents, spreadsheets, interviews, and optional consent-first workflow capture.
- Evidence — build a structured map of workflows, pains, data, skills, assets, offerings, constraints, and relationships. Machine-drafted facts remain proposals until reviewed.
- Review — test the evidence against 52 opportunity patterns, rank candidates on Impact × Feasibility × Moat, expose missing proof, and separate quick wins from strategic bets and prerequisites.
- Prove — turn selected opportunities into business cases, agent blueprints, eval suites, safety checks, and pilot plans with fixed readiness gates.
- Build — export a starter-code scaffold and production handoff package after the required gates clear. Foundry prepares the package; your technical owner decides what is deployed.
A framework that runs
Well-known consulting frameworks such as BCG's 10-20-70 rule, McKinsey's Rewired capabilities, and Deloitte's Trustworthy AI dimensions help leaders structure transformation and governance. The Foundry Method addresses a narrower execution question: which AI opportunities does this company have evidence to pursue, in what order, and what must be true before each one is built?
The methodology is implemented in a working assessment system. The same rules gather and review evidence, fire opportunity patterns, calculate inspectable scores, sequence prerequisites, and generate implementation artifacts. That makes results more consistent across engagements without making recommendations generic.
The controls behind the method
- Automatic discovery drafts evidence; a person controls what enters the company record.
- Deterministic triggers and prescores keep the LLM from silently changing opportunity order.
- Generic ideas are demoted when company-specific pain, data, or advantage is absent.
- Readiness verdicts come from evidence and fixed gates, not from an AI grading its own output.
- Public signals remain sourced hypotheses until validated against internal reality.
Where consultants use it
Advisors can begin before the first interview with a source-linked discovery brief, use workshops to close the highest-value evidence gaps, and leave the engagement with a roadmap and build package rather than a presentation alone. The method also supports repeatable portfolio reviews, transformation programs, and function-level discovery.