We help small teams run clinical supply like large ones.
Veridose is a decision layer for clinical trial supply. It connects your IRT, ERP, CDMOs, depots and couriers, recomputes the plan when the trial changes, sends agents to do the chasing, and leaves every decision with your supply manager.
- ~$500K
in lost sales for each day a trial is delayed, plus about $40K a day in direct costs.Tufts CSDD, 2024 · source
- ~50%
median waste of investigational product kits across ten benchmarked companies.McKinsey, 2021 · source
- 77%
of Phase II protocols had a substantial amendment, 2.2 on average.Tufts CSDD, 2016 · source
- 28%
chance a Phase II asset succeeds. Supply shouldn't be what fails.Citeline, 2014–2023 · source
Where clinical supply breaks down
Emerging biopharma companies started 68% of trials in 2025, up from 56% in 2019. Many run supply with a small team that keeps drug moving to dozens of sites through several CDMOs, depots, couriers and an outsourced IRT.IQVIA, 2026 · source
Forecasts built by hand
Spreadsheets cover uncertain enrollment with blanket overage, so drug expires on the shelf or runs out at a site.
Amendments that break the plan
A new cohort or country changes demand, batches, labels, IRT settings and import routes at once. The plan is rebuilt over days of email.
Status chased vendor by vendor
Courier portals, depot emails and CDMO calls take the week that should go to decisions.
Documents that drift apart
The same storage, shelf-life and label facts are retyped into the IMPD, pharmacy manual, IRT specification and label proofs.
What Veridose does
One record of the trial, with forecasting models, agents and language models working on it, and a qualified person approving every change.
-
Forecasts tuned to your trial
In-house simulation and optimization models are configured to your protocol, enrollment and visit schedule. They show the range of likely demand, and every assumption can be edited.
-
Amendment impact, fast
Veridose reads the change, recomputes the plan, traces the effects on manufacturing, labels, IRT and import, and presents costed options for the supply manager to choose from.
-
Agents that do the chasing
They query CDMOs, depots, couriers and sites, follow up when answers are late, check weather and holidays on each lane, and bring back only what needs a decision.
-
Answers for every role
Updates tailored to the study team, scientists and executives, plus an assistant anyone can ask. Answers cite their sources, respect permissions and never break the blind.
-
Documents and records
Supply sections of the IMPD, pharmacy manual, IRT specification and labels are drafted from one record, inconsistencies are flagged, and each draft goes to its owner. Every change is logged to 21 CFR Part 11 and EU Annex 11 audit-trail standards.
What it's worth
Better supply decisions help the patients in the trial, the people running it, and the budget behind it. Until our pilots report, the figures below are industry benchmarks, not Veridose results.
For the trial and its people
- PatientsTheir dose is at the site on the day of their visit, not held up by a stockout or a customs delay.
- SitesCoordinators confirm stock in a minute and get resupplied before they run low.
- Supply managersLess time chasing vendors and rebuilding spreadsheets, more time on decisions, with every call documented.
- Scientists and study leadsAnswers on cohort and country readiness when they ask, without waiting for a status meeting or breaking the blind.
- Executives and QAOne view of timeline risk and spend, and an audit trail that is ready for inspection.
For the budget
- 15–20%
lower clinical supply costs when supply planning reaches best-in-class, mostly from wasting less drug.McKinsey, 2021 · source
- ~$40K
in direct costs for each day a trial runs late. Catching a stockout or import hold early saves these days.Tufts CSDD, 2024 · source
- $141K
median cost to implement one Phase II substantial amendment, which also adds about three months to the study.Tufts CSDD, 2016 · source
Example: one Phase II trial with a $4M supply budget
| Supply costs 15% lower | $600K |
| Two weeks of delay avoided (14 days × $40K) | $560K |
| Half the implementation cost of one amendment | $70K |
| Potential saving | ~$1.2M |
Illustrative. Uses the low end of the benchmarks above; your numbers will differ. A pilot measures them against your own baseline.
How it works
Veridose sits above the systems you already use. It reads from them, works on one record of the trial, and writes approved changes back.
Your systems
One trial record
Continuously reconciledProtocol, enrollment, stock, lots, shipments and documents. Patient identifiers are removed or tokenized on entry.
Intelligence
Models set the numbersForecasting, simulation and optimization. Versioned and validated: same inputs, same answer.
Agents and language models do the legworkChase, check, explain, answer and draft. They never set the numbers.
Decision
A named person approvesSupply, QA or regulatory, depending on the action. Every approval is logged with its reason.
Outcomes
Each task goes to the model suited to it, open or frontier, hosted in region under a BAA. MCP became a shared standard under the Linux Foundation’s Agentic AI Foundation in December 2025 (source).
One trial, three views
Everyone works from the same record. Each role sees what it needs first.
How we're different
Works across vendors
IRT and platform vendors announced AI agents in 2025–26, each working on its own system. Veridose works across all of them, where the supply decisions actually sit.Suvoda · Veeva · IQVIA
Math sets the numbers
Forecasts, buffers and overage come from fixed, validated models. That follows draft EU GMP Annex 22, which limits critical GMP uses to static models and keeps generative AI out of critical decisions. Language models explain the numbers; they don't produce them.European Pharmaceutical Review
A person makes every call
Agents ask, check, draft and flag. Booking a shipment, ordering a batch, releasing stock and changing IRT settings all need a named approval.
Built to be adjusted
Supply managers asked for this: edit assumptions, override a recommendation with the reason logged, and run your own what-ifs.
How we partner
Start small on one trial, then choose where Veridose runs.
How we start
Design partner
Bring a live or recent trial, shape the product with us, and decide which integrations come first.
Pilot
One trial, with scope agreed up front. We measure drug waste, days of delay avoided and hours saved against your own baseline.
How it runs
Managed
We host and operate Veridose in region under a BAA, and ship validated releases.
In your environment
Deployed in your own cloud account, so trial data never leaves it.
Configured with your team
We tune models and scenarios to your trials together. Your team owns the assumptions; we own the validated models.
AI recommends. A qualified person approves.
Clinical supply affects patients, so it's regulated. In April 2026, FDA cited a drug manufacturer for the first time for relying on AI-generated cGMP documents without proper quality review. These commitments shape the design from the start.DLA Piper · source
What we don't claim yet: no validated deployments, no security certifications and no named integrations. We say "designed for Part 11", not "Part 11 compliant", because compliance is established with each customer.
- Human approval is explicit and loggedEvery committed action is attributed to a named person.
- Fixed, versioned models where GMP is involvedValidated before each release; nothing retrains silently.
- A stated use and risk tier per capabilityFollowing FDA's draft guidance on AI in drug development.
- Minimal patient data, kept in regionRemoved or tokenized on entry, and aligned with ICH E6(R3), HIPAA and GDPR.
Build it with us
We're raising a seed round to build amendment impact analysis, and looking for supply teams to shape it.
Or email [email protected], [email protected] or [email protected]. We reply within a week.