In development. Design partners welcome.

AIV Engine is AI validation software for GxP quality teams

The AI drafts. A person signs.

It classifies an AI system by GxP risk, sizes the validation to that risk and drafts the evidence set. A named person reviews and signs every record.

Illustration with sample data

The problem AIV is built for

  1. You have an AI system in quality, or one on the way.
  2. Most teams have no rule for how much validation an AI system needs.
  3. Every team ends up improvising the answer, and an inspector will ask how you reached it.

AIV runs the 4-tier method from my book, Validating AI in GxP Environments: A Practitioner's Handbook, the same way every time.

One AI system goes through 4 steps

A classifier, a vision model, an LLM application, a RAG assistant or an agent, built in-house or bought from a supplier. Each step leaves a record.

  1. 01

    Classify the use case

    4 questions set the GxP risk tier by a published rule, from what happens when the output is wrong.

  2. 02

    Tier the risk

    The tier decides how much validation the system gets. A dashboard nobody acts on gets a light touch. A system that routes a CAPA gets the full treatment.

  3. 03

    Draft the validation package

    Intended use, acceptance criteria, protocols and records come out ready for review. The AI layer drafts them.

  4. 04

    A qualified person reviews and signs

    Every signature carries a name, a role, its meaning and a timestamp, the shape 21 CFR Part 11 asks for. Nobody can approve their own work.

After release, monitoring watches for drift and sends the system back for reclassification when its scope, its data or its performance changes.

What AIV never does

The AI layer drafts documents, proposes acceptance criteria and extracts facts from the material you give it. It never classifies a system, never signs a record and never lowers a criterion.

Your data trains nothing. The AI layer runs on a private endpoint, and what you put into AIV stays in your evidence record.

It also stays out of 4 things on purpose

  • Non-AI software. Your ERP, LIMS and eQMS are validated the way they are today.
  • Equipment, process or cleaning validation. These stay in your existing validation programme.
  • Test scripts for your other systems. AIV writes the evidence set for the AI system only.
  • Measuring the model. AIV takes accuracy and drift numbers from your own tools and turns them into signed evidence.

Who the pilot is for

Design-partner conversations are open with 2 kinds of organisation. Both get the same thing and are asked for the same thing.

Quality teams

You run quality in pharma, biotech or a CDMO and have 1 or 2 AI systems in use or planned

You get early access to the working prototype, my own hands on the first validation and a say in what the product becomes.

AI vendors

You sell AI into pharma and your customers keep asking how to validate it

You get one validated method your customers can inherit, my own hands on the first validation and a say in what the product becomes.

What I ask in return: one real AI system, honest feedback and permission to describe the outcome without naming your company.

Try the classification step now

Answer 4 questions about one AI system and see its GxP risk tier and the validation shape it calls for. A simple model can sit higher than a complex one.

How AIV Engine classifies an AI system into 1 of 4 GxP risk tiersDecision tree. Q1: does the output influence a GxP decision? No gives Tier 1, Informational. Yes goes to Q2: if wrong on 10% of records for 6 months unnoticed, is there material risk? No gives Tier 1. Yes goes to Q3: does a qualified human approve every output first? Yes gives Tier 2, Assisted decision. No goes to Q4: is there a working human override exception path? Yes gives Tier 3, Guided workflow; No gives Tier 4, Autonomous action.Q1Does the outputinfluence a GxPdecision?Q2Wrong on 10% for6 months unnoticed:material risk?Q3Does a qualifiedhuman approve everyoutput first?Q4Is there a workinghuman overrideexception path?YesYesNoNoNoNoYesYesTIER 1InformationalIntended use, data sourcesand a periodic review.TIER 2Assisted decisionScripted OQ, PQ withreal users, everyoverride logged.TIER 3Guided workflowFull IQ, OQ and PQ,monitoring andexplainability.TIER 4Autonomous actionTier 3 controls plusa written case for nohuman in the path. How AIV Engine classifies an AI system into 1 of 4 GxP risk tiersDecision tree. Q1: does the output influence a GxP decision? No gives Tier 1, Informational. Yes goes to Q2: if wrong on 10% of records for 6 months unnoticed, is there material risk? No gives Tier 1. Yes goes to Q3: does a qualified human approve every output first? Yes gives Tier 2, Assisted decision. No goes to Q4: is there a working human override exception path? Yes gives Tier 3, Guided workflow; No gives Tier 4, Autonomous action.Q1Does the output influence a GxPdecision at all?No → Tier 1YesQ2Wrong on 10% of records for 6 months,unnoticed: is there material risk?No → Tier 1YesQ3Does a qualified human approve everyoutput before GxP action is taken?Yes → Tier 2NoQ4Can a human override the decisionthrough a real exception path?Yes → Tier 3 · No → Tier 4TIER 1InformationalIntended use, data sources and aperiodic review.TIER 2Assisted decisionScripted OQ, PQ with real users,every override logged.TIER 3Guided workflowFull IQ, OQ and PQ, monitoringand explainability.TIER 4Autonomous actionTier 3 controls plus a written casefor no human in the path.
The whole classification rule. 4 questions give 1 tier. The tier comes from what happens when the output is wrong, so a simple model can sit higher than a complex one.

1. Does the output influence a GxP decision at all?

2. If it were wrong on 10% of records for 6 months without anyone noticing, would patient safety, product quality or data integrity be at material risk?

3. Does a qualified human review and approve every output before a GxP action is taken?

4. Can a human override the decision through an exception path that exists in the live workflow?

Start with one conversation

Pilots begin with a conversation about one real AI system.

If you'd rather talk, book a conversation with Sachin and bring the system with you.

Ask about a pilot

Tell me about the system you have in mind and I'll write back to you myself.

I use these details only to reply to you. Enquiries are deleted after 24 months. Privacy notice.

No mailing list. One reply, from me.

Questions people ask about AIV

Is AIV a replacement for my validation platform?

No. AIV sits beside your existing validation platform and handles the AI-specific part: the risk tier, the evidence set an AI system needs and the controls that keep it in check after release. Your ERP, LIMS and equipment validation stay where they are today.

Does AIV compute model metrics?

No. AIV takes the numbers your own tools produce, such as accuracy on a test set or drift readings in production, and turns them into evidence with acceptance criteria, a tester, a timestamp and a signature.

Is my data used to train anything?

No. Nothing you put into AIV trains a model, ours or anyone else's. The AI layer runs on a private endpoint, and what you enter stays inside your evidence record.

When can I buy it?

AIV is in development, and design-partner conversations are open now. Dates will be shared with design partners. If you have an AI system you need to validate, ask about a pilot and we'll talk about whether it fits.

Security, data handling and the company behind AIV are on the AIV product site.