Know where your clinical AI stands against the 44 recommendations, and what to fix first.
Independent assurance, built on the UK Government’s National Commission recommendations. Take the free 3-minute self-check, see your readiness radar instantly, then turn it into evidence you can hand to a commissioner.
Start the free self-check Skip to the Readiness CheckAnswer sixteen quick questions
Across the eight areas buyers and commissioners will ask about. Scored 0–100, shown as a traffic-light radar.
Illustrative examples, not real organisations. Complete the check below to see yours.
Your readiness at a glance
Your three priorities
All eight dimensions
What the colours mean
| Please improve | Under 40. A material gap a commissioner or buyer will flag. Close it before it is asked about. |
| Nearly there | 40 to 69. A foundation in place, with a specific gap to close. |
| You’re ready | 70 and above. Strong enough to evidence; keep it under review. |
Turn these gaps into a commissioner-ready report
Send your details using the short form below and I will come back to you personally with how the oDIX8 Commission Readiness Check would map all 44 recommendations for your specific tool and workflow.
Request the Readiness Check →Turn this into evidence you can hand to a commissioner
This self-check is indicative. The oDIX8 Commission Readiness Check maps all 44 recommendations for your specific tool and workflow, ranks your exposures and gives a concrete remediation route for each.
See the full Readiness Check →Your sixteen answers stay in your browser and are never sent. To improve the tool we record only your anonymous score, your overall number and the eight dimension scores, with no answers, no email and no identifier.
Two kinds of organisation need this now
Whatever your radar just showed, the stakes are the same for two groups, and the rest of this page is how to close the gaps it found before anyone else asks about them.
NHS providers
You are deploying or procuring AI and need to show a board or a commissioner that it is safe and that the evidence behind it holds.
Clinical-AI vendors
You sell into the NHS, and buyers increasingly want independent evidence that your tool does what the brochure says.
Roughly 40 to 60% of B2B decisions are lost not to a competitor, but to doing nothing
With clinical AI, doing nothing has a price that lands later, and larger. A recent Lancet study of a patient-facing conversational diagnostic AI in real primary care put the single most likely diagnosis right only 56% of the time, and its clean safety record held only under live clinician supervision. When several AI tools sit in one workflow, small errors compound down the chain.
| If you wait | Likely cost | With the Check |
|---|---|---|
| A tender asks for independent evidence you do not have | one lost contract, often £50k–£500k+ over its term | evidence ready, you stay in the running |
| An audit is demanded later, at full rate | focused audit £8k+, notified-body route £33k–£70k | a fraction of that, now |
| A coding or safety gap surfaces in production | remediation + reputational cost, hard to bound | caught on paper, before go-live |
| Spring-2027 guidance lands and you are not ready | months of delayed market entry | ten working days to clarity |
Audit and notified-body figures from public market sources (2026).
What being unready costs you, in three scenes
Independent assurance buyers and boards trust
oDIX8 is independent assurance for clinical AI. We assess clinical-AI tools and we do not sell them, so the verdict is yours to put in front of a buyer or a board. We exercise your declared workflow with synthetic probes and show exactly where the product holds and where it breaks, with grading separated from production so the result stands up to a second party and a third. No patient data, no data-sharing agreement, no source code, nothing to integrate.
We saw this coming. A body of peer-reviewed and open-access work, time-stamped before the Commission reported, identified several of the gaps it went on to recommend addressing, from regulating by function rather than by whole system, to correlated failure risk across AI pipelines. The full record is public on ORCID (0000-0002-6273-8014).
oDIX8 is founded by a senior researcher at Martin-Luther-Universität Halle-Wittenberg, in the group of Professor Thomas Frese, President of WONCA Europe.
The evidence behind the method
This is not an opinion assembled last week. The method rests on peer-reviewed and open-access research that, time-stamped before the Commission reported, named several of the failure modes it went on to recommend addressing:
- Regulate by function, not by whole system: our preprint Cascade Amplification in Administrative AI Pipelines, the gap the Commission flagged at report pages 29–30.
- AI that co-produces clinical and administrative reality: Governing the Loop (under review, PLOS Digital Health).
- Correlated failure risk across AI pipelines: Feedback-Loop Amplified Operational Risk (SSRN).
- Drift and manufactured prevalence: Artificial Epidemiology and The LOOSE-Rescue Paradox (697,021 synthetic encounters).
A Springer monograph, first-author studies in BMC journals, and governance work under review at PLOS Digital Health and the European Journal of General Practice, with open-access preprints on Cambridge Open Engage. Full, verifiable record: ORCID 0000-0002-6273-8014.
Independent scholarship confirms the gap: a 2026 study of six health systems in npj Digital Medicine finds existing evaluation frameworks lack guidance for exactly the stochastic, prompt-sensitive, adaptive properties of generative AI, the methodological gap oDIX8 operationalises.
Independently confirmed, three times
This is not the first time an independent body has reached the conclusions of this work. Each time, our dated, open-access research came first.
Austria’s national health institute (GÖG) independently reached the conclusions of our published video-consultation research (BMC Primary Care 2023; BMC Digital Health 2025).
Healthwatch England reported incorrect medication and diagnosis entries from AI scribes. Our preprint The Pipeline Gap had set out the mechanism fourteen days earlier (17 August 2026), alongside Artificial Epidemiology and Cascade Amplification.
Six of our pre-dated positions reappeared among its 44 recommendations: from regulation by function rather than whole system, to post-market drift and correlated failure risk.
Already proven in the real world
An independent assessment of an oncology AI tool ran to a 40-page report, our findings were independently replicated by the buyer’s consortium, and the tool was found not fit for clinical use.
A full independent assurance review of a UK clinical-AI product, with a findings register and a framework crosswalk (DCB0129, ISO 14971, EU AI Act).
A paid clinical-AI tool, built and delivered to a leading German university chair.
Anonymised. Full detail available on request under NDA.
The oDIX8 Commission Readiness Check
A fixed-fee, independent 20+ page report that maps your position against all 44 recommendations, in plain language, in ten working days, with no patient data and no software access required.
- 1. Scope call (45 min).You describe the tool and the clinical workflow it touches. Nothing confidential, no data.
- 2. We probe and map.We exercise the declared workflow with synthetic probes and score every one of the 44 recommendations for your case.
- 3. Report + readout.You receive the 20+ page written report and a 45-minute walkthrough. Ten working days, start to finish.
What the report reveals about your tool
- Which of the 44 you can already evidence in a tender, and the exact wording to claim it.
- The chain-level gap that appears when your tool meets the others in the workflow, the one most assessments never test.
- Where staff may be over-trusting the model, and the single change that cuts automation-bias risk fastest.
- The patient subgroup your tool may quietly underperform for, before an equity question finds it.
- Which gaps a commissioner will raise first, ranked, so you fix in the right order.
Everything in the Readiness Check
| Deliverable | What it does | Indicative value |
|---|---|---|
| Gap map against all 44 recommendations | Covered, exposed, or not applicable, scored for your tool | £2,500 |
| Your five highest-priority exposures, ranked | By severity and how soon they bite | £1,500 |
| Coverage view with tender wording | Language you can paste into a bid | £800 |
| Remediation routes | A concrete next step for every material gap | £1,800 |
| 45-minute readout call | We walk you and your team through it | £400 |
| Bonus: procurement evidence summary | One page to attach to a tender | £600 |
| Bonus: 12-month re-check at founding price | Lock the founding rate for a repeat | £950 |
| Total indicative value | £8,550 | |
What each component is worth bought separately, at market rates.
An illustrative sample (a fictitious tool, orange “Partly ready” result), so you can see the format.
Three ways to answer the question, and what each costs
| Do nothing | oDIX8 Readiness Check | Full external audit | |
|---|---|---|---|
| Cost | £0 now, unbounded later | £3,500 (founder £1,950) | £8,000–£70,000+ |
| Time | n/a | 10 working days | weeks to months |
| Patient data | n/a | none required | often required |
| You get | exposure | gap map, priorities, remediation | conformity / certification |
The fast, independent way to know where you stand and prove it.
The first ten organisations get the full Commission Readiness Check at the founding price, in exchange for a short case study we can anonymise. When the ten places are gone, it is £3,500.
Claim a founding placeImplementation lands by spring 2027 and procurement cycles are turning now. Founding-client places are limited only by how many Checks we can deliver properly each month. The organisations that can show where they stand will be the ones still in the room when the evidence is asked for.
Questions
Is the National Commission law?
The recommendations are not themselves law. The UK Government has accepted all 44 and is implementing them, with technology-specific guidance to follow.
What do you need from us?
A description of your tool and the clinical workflow it touches. No patient data, no data-sharing agreement, no source code.
How long does it take?
Ten working days from the scope call to the finished report and readout.
P.P.S. Only ten founding places exist at £1,950. When they are gone, the price is £3,500. If you are going to do this, do it as a founding client.