Build it. Test it.
Defend the next decision.

Set a safe, evidence-led way of working with AI before making venture decisions.

• Distinguish assistance, automation, and accountable human judgment.
• Set privacy, verification, disclosure, and escalation rules.
• Establish a baseline for AI readiness and decision quality.

Responsible working protocolA signed AI-use charter, readiness baseline, and named decision owner.

Turn an interesting idea into a falsifiable opportunity thesis.

• Frame a problem around a specific actor, context, and costly struggle.
• Separate facts, assumptions, and AI-generated possibilities.
• Prioritize opportunities using impact, access, evidence, and fit.

Falsifiable venture thesisA problem statement, target actor, assumptions map, and explicit stop condition.

Replace synthetic validation with evidence from real people and observable behavior.

• Run non-leading customer discovery conversations.
• Triangulate interviews, secondary research, behavior, and commitments.
• Make a go, pivot, pause, or stop decision with calibrated confidence.

Evidence reviewInterview notes, secondary sources, contradictory evidence, and an evidence-strength rating.

Design trust controls alongside the venture instead of adding compliance at the end.

• Map data, models, people, providers, and affected stakeholders.
• Identify misuse, bias, privacy, security, and human-oversight risks.
• Select proportionate controls and accountable owners.

AI assurance passportAI/data inventory, system card, threat model, risk register, and human-oversight plan.

Connect customer evidence to value creation, capture, and the next cheapest test.

• Build an evidence-linked business model and value proposition.
• Expose the riskiest desirability, viability, feasibility, and integrity assumptions.
• Design a measurable experiment with a decision threshold.

Experiment-ready modelBusiness model, value proposition, riskiest assumption, experiment card, and threshold.

Define the smallest trustworthy product that can test the venture thesis.

• Map the user journey and select one high-value core flow.
• Separate minimum learning scope from feature accumulation.
• Write testable acceptance criteria and a product experiment plan.

PRD-lite and test planUser journey, three core capabilities, exclusions, acceptance criteria, and observed-test plan.

Choose an architecture by measured quality, safety, latency, and cost—not novelty.

• Choose build, buy, model, retrieval, workflow, and human-review boundaries.
• Create a small golden evaluation set and measurable release thresholds.
• Estimate cost per trusted outcome and define fallback behavior.

Architecture decision recordSystem diagram, model/provider comparison, eval set, thresholds, unit cost, and fallback plan.

Build one working core flow, observe people using it, and revise it.

• Build through a no-code, code-assisted, or Wizard-of-Oz track.
• Run observed usability and failure-path tests.
• Prioritize issues and ship one evidence-led revision.

Working prototype reviewDemo link or recording, three observed tests, issue log, before/after evidence, and revision note.

Know what a trusted outcome costs, what customers may pay, and how long the venture can learn.

• Model pricing, gross margin, acquisition, retention, and support assumptions.
• Include inference, retries, tools, review, failures, and data work in AI cost.
• Run base, upside, downside, and model-cost shock scenarios.

Economic decision packPricing rationale, unit economics, scenario model, cost per trusted outcome, and runway or investment need.

Turn positioning into a measurable route to the first repeatable customer or adopter.

• Choose a focused segment, positioning, channel, and sales motion.
• Create proof-based messaging with appropriate AI disclosure.
• Run one channel or sales experiment with a decision threshold.

Market-entry experimentPositioning, outreach or adoption asset, experiment results, funnel measures, and next decision.

Create the decision rights, roles, measures, and cadence needed to operate responsibly.

• Define decision rights across product, model, data, safety, and customer operations.
• Choose metrics that connect learning, value, quality, safety, and economics.
• Design the first-hire or cross-functional capability plan.

Operating reviewRole charter, decision-rights map, metric tree, meeting cadence, and practiced stakeholder narrative.

Choose the right financing or approval path and enter diligence with organized evidence.

• Compare bootstrapping, revenue, grants, equity, debt, and internal investment.
• Identify material entity, contract, IP, data-rights, employment, and AI obligations.
• Read term-sheet economics and build a diligence index.

Diligence readiness reviewFunding or approval memo, ownership scenario, legal/data checklist, red flags, and indexed evidence room.

Demonstrate what works, acknowledge what does not, and defend the next 90 days.

• Synthesize venture, product, economics, trust, and operating evidence.
• Demonstrate the prototype and respond to a live incident scenario.
• Defend a go, pivot, pause, or stop recommendation and a 90-day plan.

Final evidence defenseVersioned venture dossier, live prototype, decision ledger, incident response, pitch, and 90-day control plan.