DUN-AI private beta
Before you invest in AI, prove that you need it.
Bring one real AI investment question your organisation is considering. DUN-AI evaluates it and returns a decision: invest, defer, remediate first, use a non-AI alternative, keep the process human-led, or run a restricted pilot — with the reasoning, the economics and the conditions attached.
5 of 5 places remaining. Counted from applications actually accepted, not from a figure typed into this page.
Recommended outcome
Use conventional automation
No new AI
The task is deterministic and fully specified. Rules are cheaper to build, cheaper to run, and can be audited line by line. An AI system here would add cost and opacity without adding capability.
- Alternatives compared
- 13
- Blocking gates
- None triggered
- Determined by a model
- No
Most organisations are asking the wrong question
“Where can we use AI?”
It assumes the answer. A business case written after the technology has been chosen is not a comparison, it is a justification — and it will usually pass, because nothing in it was built to fail.
“Should AI be used here at all?”
DUN-AI starts one step earlier. If the answer is yes, it identifies the smallest intervention that can justify itself. If the answer is no, it says what to do instead — and that is a first-class result, not a failure to conclude.
Badly selected AI investments cost more than the money. They add operating complexity, integration burden and governance overhead to a process that was not the constraint in the first place.
What you receive
One executive decision record for the question you bring. It is the same document the specimen above is a sample of.
- A decision from a closed set of fifteen outcomes
- The reasoning that produced it, dimension by dimension
- The strongest non-AI alternative, scored on the same basis
- The smallest justified intervention
- Cost and value as ranges, where the evidence supports them
- Break-even conditions, or a statement that they cannot be reached
- Risks, and any blocking gate that closed the decision
- What evidence is missing and would change the answer
- A concrete next action for management
- The conditions under which it should be reassessed
DUN-AI is willing to tell you not to buy AI
An assessment resolves to exactly one of fifteen outcomes. Nine of them authorise no AI spending at all. That is the point of the product, not a caveat on it.
- Reject the investment
- Use conventional automation instead
- Use platform capability you already license
- Keep the process human-led
- Fix the data first
- Redesign the process first
- Defer until conditions change
- Refer for specialist review
- Run a restricted pilot
- Proceed, with conditions attached
The engines that produce these cannot read commercial, billing or vendor data. That separation is enforced by a test, not by a policy — which matters, because DUN-AI is built by a company that also does implementation work.
Who the beta is for
Roles
CEO, COO, CFO, CIO, CTO, CDO, Head of AI, Head of Transformation, enterprise architecture and digital leaders, business-unit leaders and AI programme owners. You need to be close enough to the decision that the answer changes something.
Sectors
Telecommunications, banking and financial services, energy, government and public sector, and healthcare each have their own content pack. Management consultancies, AI consultancies and systems integrators are welcome as secondary candidates. Outside these five, an assessment runs on core content only and we will tell you that before you start.
What this is, and what it is not
- This is an invitation-only private beta. It is supervised, limited, and not a public product.
- No organisation has used DUN-AI to make a real investment decision yet. There are no customers, no case studies and no results to quote, and this page carries none.
- DUN-AI is decision support. It does not give legal, regulatory, financial, medical or clinical advice, and it does not replace an accountable executive, a lawyer or a clinical safety officer.
- It makes no promise about the outcome of any investment, and no promise of savings, return, compliance or delivery success.
- Content packs carry a validation status that is shown inside every report. You will be able to see exactly how much assurance the content behind your assessment carries.
The full list of what is not built and what is weak is published in advance, because a buyer should know it before a procurement conversation rather than after one.
Apply
Every application is read by a person and you will hear back either way. Applying takes about five minutes and asks for nothing confidential.