Value leaks
Pilots work in one team and stall there. MIT found that 95% of organizations get no measurable return from generative AI. Without a shared map, no one knows which steps to move next, or what it saved.
The system of record for human + AI work in life sciences
Where humans + AI agents
make the miracles of science
happen.
Map every process, from R&D to commercial operations. Give each step to a human, an AI agent, or both. See how AI works across the whole organization, and keep every process optimal, auditable on-chain and running, even when a model goes down.
- Human
- Human + AI
- AI
- Breakthrough
The problem
Pharma is adding AI everywhere. Nobody can say who does what.
AI arrived one team and one tool at a time: a translation engine here, a literature screener there, a copilot at the agency. Nobody designed the split between people and AI, so nobody owns it.
Risk builds up
When QA or an inspector asks whether a person or a model checked something, the answer sits in someone's inbox.
People disengage
Teams hear "AI" and think "replacement". Without clear roles, adoption turns into resistance.
Source: MIT NANDA, "The GenAI Divide: State of AI in Business 2025."
Why now
AI is in the workflow.
Regulators want proof of who is in charge.
AI floods the review queue
Generative AI makes it cheap to create more content, more local variants and more documents. Every piece still needs medical, legal and regulatory review, so the bottleneck moves to the human steps.
Human oversight becomes a requirement
The EU's draft GMP Annex 22 on AI requires human oversight for high-impact uses. The EU AI Act's high-risk obligations apply from December 2027. The FDA published draft guidance on AI used for drug regulatory decisions in 2025. Proof of oversight is becoming a cost of doing business.
Same teams, more work
Patent expiries and pricing pressure push every function to do more with the people it has. Deciding what AI should take is now a leadership question.
Sources: EMA / European Commission draft GMP Annex 22 (July 2025); EU Digital Omnibus on AI (in force July 2026); FDA draft guidance on AI to support regulatory decision-making for drug and biological products (January 2025).
How it works
- MapOpenPetri drafts every process from your SOPs. Your team checks it.
- AssignEach step goes to a human, an AI agent, or both.
- RunPeople get their tasks in Teams, Slack or email. Agents do the routine.
- ProveEvery decision, switch and sign-off is logged for QA.
The product
See it on a real process
Pick a process and map it. Then optimize it with OpenPetri for time and cost, simulate what happens under pressure, see every AI tool, system and person it depends on, and see how each person gets just their own tasks in Teams, Slack or email.
Same team, better work. Nothing here removes people. Every step keeps a named owner, and every hour AI takes on goes back to the team for strategic work.
Sample data. Owners, systems and figures are fictional.
Book a demo
20 minutes. We walk through a process you choose.
For the people doing the work
No complex maps.
Just a simpler way to work.
Process maps get complex, and most people should never have to read them. With OpenPetri they don't. Each person gets a short message in Microsoft Teams, Slack or email with only the tasks that need them, and everything ready to finish them. Less searching and fewer tables, more time for the work that matters.
- Tasks come to youNo new tool to learn. OpenPetri messages each person in the app they already use, and only when a step needs them.
- AI does the prepDrafts, checks and references arrive ready, so people spend their time on judgment, not on copying and pasting.
- Done in one clickApprove, edit or sign off from the message. The decision is recorded with the person's name, ready for QA.
OpenPetri Morning, Laura. Two things need you today. Everything else in the campaign is handled or waiting on others.
AI draft readyDue today
Personalised email variants
12 variants built from approved modules. Check tone and fit, then approve.
Your turnDue tomorrow
Channel and sequence planning
Segments are ready. OpenPetri suggests a three-touch sequence for you to adjust.
AI agents, managed
Hire AI agents like you hire people.
Find them inside your company or out.
Every AI agent in OpenPetri gets what a new team member gets: a clear role, a manager, a budget and regular reviews. So leaders always know what each agent does, who answers for it, what it costs and whether it earns its place.
Job description
What the agent may do, and what it may never do.
Links each claim to its approved reference. Never approves content.
Manager
A named person who answers for the agent’s work.
@a.ricci, Medical information
Token budget
A monthly token limit, with an alert before it runs out.
6M tokens a month · 4.1M used
Performance review
Quality, speed and cost against the human baseline. Keep, retrain or replace.
98% accurate on sampled checks · keep
Example: a claim-reference agent in MLR review. Sample data.
Try the agent roster: every AI agent in the company, who manages it, what it costs and how it performs.
Sample data. Agents, owners and figures are fictional.
Didn’t find an AI agent with the right skills?
Find one on the market. Post a job for an AI agent, just like you would for a person.
OpenPetri screens every candidate for compliance and your hiring criteria before it reaches you.
Vendors never get access to your data. OpenPetri deploys their AI agents inside your own environment, smoothly and in full compliance and security.
- Post the jobWhat the AI agent must do, what it must never do, and which data it may touch.
- Test the applicantsVendors and builders apply. Each AI agent runs on samples of your past work, inside your environment.
- Hire the best fitCompare quality, cost and validation. It deploys in your environment and joins your roster with a manager.
Failover for AI
When one AI goes down,
the process doesn't.
OpenPetri lets you build flexible processes. Every AI-assisted step has a ready alternative: another model, another provider, or a person. If a tool fails, becomes too expensive or is no longer available, OpenPetri switches the step and the work keeps moving.
- PredictOpenPetri shows which steps depend on a single provider or region, before it becomes a problem.
- PrepareEach AI-assisted step gets a pre-approved alternative, validated in advance.
- SwitchIf a tool is lost, the step moves to its alternative. Qualified-person sign-offs stay in place, and every switch is logged.
ScenarioA US AI service is no longer available in Europe
- Personalised email draftsUS-hosted language modelEU-hosted language modelRunning
- Local adaptation of claimsUS translation serviceEU translation engineRunning
- Literature screeningUS search modelIn-house modelRunning
- Medical reviewQualified personRunning
3 of 4 steps depend on US-hosted AI. OpenPetri has an alternative ready for each.
Illustrative scenario. Tool names are generic.
$500K
Why it pays off
The cost of one day of delay for an average drug.
A day of a phase II or III trial costs another $40,000. OpenPetri finds where work waits, between people, tools and sign-offs, and measures the days it takes out.
What a pilot proves. We measure one process before and after: cycle time, review rounds, routine hours per asset, and the share of AI-assisted steps with documented human oversight. Those four numbers make the business case for the next function.
Source: Smith, DiMasi & Getz, Tufts CSDD, "New Estimates on the Cost of a Delay Day in Drug Development," Therapeutic Innovation & Regulatory Science (2024).
Competition
Others see part of the picture.
We map the whole of it.
Today teams piece this together from tools built for something else. Ratings reflect what each category typically offers.
| How teams solve it today | Maps real work | Decides human vs AI | Audit-ready evidence | Stays current | Works across every AI provider |
|---|---|---|---|---|---|
| Pharma suites with built-in AI e.g. Veeva | Partly | Partly | Yes | Yes | No |
| Process mining e.g. Celonis | Yes | No | Partly | Yes | No |
| AI governance platforms e.g. Credo AI | No | Partly | Yes | Partly | Partly |
| Consulting projects | Yes | Yes | Partly | No | No |
| SOPs and spreadsheets | Partly | No | No | No | No |
| OpenPetri | Yes | Yes | Yes | Yes | Yes |
Neutral by design
We don't sell models or content tools. OpenPetri routes each step to the best option, including the tools clients already own, and switches when one fails.
Designed for GxP from day one
Guardrails, audit trail and validation evidence are the core of the product, not add-ons. That's what gets past QA in pharma.
Hard to replace once adopted
The map becomes where AI decisions are recorded and signed off, and every map teaches OpenPetri about the next one. Once QA relies on it, it stays.
Market and business model
Start with one process.
Expand to the whole company.
- Land
Commercial content, MLR & omnichannel
High volume, highly repetitive and fully regulated. Every global pharma company runs it, AI is already arriving there, and our team has run it for years. The fastest place to show the time AI gives back to teams.
- Expand
Every function in pharma
R&D, clinical, regulatory, safety, medical, market access, manufacturing and sales. The same map, markers and guardrails, one process at a time.
- Extend
All of life sciences
Scaling biotechs, CROs, CDMOs, medtech and diagnostics. Later, any industry where AI needs human oversight.
Function heads and transformation leaders at global pharma, and operations leaders at scaling biotechs building AI-native from day one.
QA and compliance, who need evidence of human oversight for every AI-assisted step.
Annual subscription per function. It grows with the processes mapped and the AI steps OpenPetri runs.
A 12-week paid pilot on one process, with the baseline measured in week one. Then expansion across functions, markets and affiliates.
$3.2B
Veeva's revenue in fiscal 2026, built as a system of record for life sciences. Pharma pays for systems of record. Human + AI work doesn't have one yet.
Source: Veeva Systems, fourth quarter and fiscal year 2026 results (March 2026).
The plan
What the seed round buys
- Interactive prototype across three pharma processes
- Wedge chosen: commercial content and MLR review
- A team with 10+ years inside seven global pharma companies
- Q1 2027First paid pilot live in commercial content
- Q2 2027First process in production, with a full audit trail
- Q3 2027Three paying customers
- 2028Every function running on OpenPetri at one pharma company
The team
We lived this problem inside global pharma
10+years in pharma
8global pharma companies
Founders
Advisors
- ConfidentialR&D AdvisorLinkedIn on request
- ConfidentialRegulatory AdvisorLinkedIn on request
- ConfidentialMedical AdvisorLinkedIn on request
- ConfidentialP&C AdvisorLinkedIn on request
- ConfidentialCommercial Strategy AdvisorLinkedIn on request
- ConfidentialIT AdvisorLinkedIn on request
Team coming from
- Novartis
- Takeda
- Almirall
- Boehringer Ingelheim
- Sanofi
- Roche
- AstraZeneca
- Galderma
- Microsoft
Company names describe team members' professional experience. They don't imply any partnership or endorsement.
The mission
Halve the time from molecule to patient.
Every day a medicine waits costs about $500,000. We start by taking the waiting out of every regulated process, with humans at the center and AI agents doing the routine.
Every process mapped the way it really runs, not the way the SOP says.
Every step given to a human, an AI agent or both, with a named person accountable.
AI takes the routine. People get time back for the science, the care and the ideas only they can bring.
Q&A
Questions we hear
Does OpenPetri reduce headcount?
No, and that isn't the goal. Every step keeps a named owner, even when AI does the work.
The hours AI takes on go back to the same team for strategic work: better science, closer customer relationships, sharper decisions. OpenPetri measures that time so leaders can see where it went.
Isn't this a consulting project?
No. A consulting project delivers a snapshot that is out of date within months. OpenPetri is software: the map stays current as tools, teams and rules change, and every decision is recorded. It runs at software margins, not consulting rates.
What stops Veeva or an AI lab from building this?
Suite vendors build AI into their own apps. AI labs want work to run on their own models. Pharma companies run dozens of tools and several AI providers, and they don't want one vendor deciding what goes where.
OpenPetri is neutral. It maps work across every tool, including Veeva, and can switch providers when one fails or gets too expensive. A suite vendor or a lab can't credibly offer that.
How big can this get?
Pharma already pays for systems of record: Veeva reached $3.2B in revenue in fiscal 2026 selling them to life sciences. Human + AI work has no system of record yet.
We land in commercial content and MLR, expand to every function in pharma, then to the rest of life sciences and other industries where AI needs human oversight.
Let's talk
We're raising our seed round and looking for design partners in pharma. Tell us who you are and we'll set up a demo on a real process.