AI Products
How to Validate an AI Product Idea Before You Write Code
A founder-friendly validation playbook for AI products: problem interviews, fake-door tests, wizard-of-oz MVPs, willingness to pay, and kill criteria.

AI makes demos easy and businesses hard. The failure mode is building a flashy prototype nobody budgets for. Validation means proving a specific user will change a workflow and pay—before you invest in agents, RAG pipelines, and multi-tenant SaaS.
Use idea lists from 25 AI product ideas only as starting points.
Step 1 — Name the job, not the model
Write one sentence: “Helps [role] complete [job] so that [measurable outcome], today done with [current tool/pain].” If you cannot fill this, you have a technology hobby, not a product.
Step 2 — Ten problem interviews (not pitch meetings)
- Ask for the last time the pain happened
- What they tried; what it cost in time/money
- Who else is involved in buying
- What “good enough” looks like
Stop talking about transformers. Listen for budget and workarounds.
Step 3 — Wizard-of-Oz MVP
Deliver the outcome with humans + ChatGPT + Notion behind a simple form. If users will not use a concierge version, they will not use your agent either.
Step 4 — Fake door and pricing signals
- Landing page with waitlist and price anchor
- Letter of intent for B2B pilots
- Paid pilot (even small) beats vanity signups
Step 5 — Data and risk feasibility
Can you legally access the documents/APIs? Is accuracy threshold achievable with retrieval + review? What is the cost of a wrong answer (refund, lawsuit, safety)?
Kill criteria (write them down)
- Fewer than 3 design partners who will do weekly calls
- No one names a budget owner after 15 interviews
- Error cost requires 100% accuracy you cannot guarantee
- Incumbent can ship the same feature as a checkbox next quarter and you have no wedge
What “validated” looks like
- Repeated pain stories with quantified cost
- Users completing the workflow weekly in your pilot
- Money or irrevocable procurement step
- A clear next integration that deepens the wedge
Frequently asked questions
How long should validation take?
Often 2–6 weeks of focused conversations and pilots—not 9 months of silent building.
Build AI products with Auroviq
Auroviq (AuroviQ) helps founders and product companies design, build, and scale AI-powered software—from MVP to production agents—across the UK, Netherlands, Singapore, and India.
Tags
Next step
Building AI products that ship?
AuroviQ helps teams design, build, and scale reliable software and AI systems — from mobile apps to enterprise platforms.