AI Readiness Assessment
Six questions that predict whether an AI project will reach production. Run them yourself before anyone quotes you for anything.
- 01
Data accessibility
Can you get at the data an AI system would need, through an API or an export, without a project to build one first?
This is the strongest single predictor of how a project goes. Inaccessible data does not stop AI work, it just moves the cost into integration before anything intelligent happens.
- 02
Data quality
Is the data consistent enough that a person could work from it without correcting it first?
Models do not repair bad data, they propagate it faster and with more confidence than the process they replaced.
- 03
Process definition
Could you write down the steps of the process you want to automate, including the exceptions?
If nobody can describe the current process, automating it encodes whatever it currently does, including the parts nobody intended.
- 04
Regulatory position
Do you know which rules govern this data and whether they permit third-party processing?
This determines the architecture before anyone picks a model. Discovering it late is what kills projects at the point they were about to go live.
- 05
Success definition
Can you state the number that would have to move for this to have been worth doing?
Projects without one do not fail, which sounds good and means nobody can tell whether to continue funding them.
- 06
Ownership
Is there a named person accountable for this working, with authority to change the process around it?
The most common cause of a technically successful pilot never reaching production is that nobody owned the change it required.
How to read your answers
A clear yes to all six means you are ready to build, and the next step is picking the right first use case rather than another assessment.
A no on data accessibility or data quality means your first project is a data project. That is unwelcome news and far cheaper to receive now than after a pilot has spent four months discovering it.
A no on regulatory position is the one to resolve before anything else, because it decides the architecture. It is also the one most often deferred, on the assumption it can be sorted out later. It usually cannot.
A no on ownership is the quietest killer on this list. Pilots that work technically and never reach production almost always failed here, and it is the only item nobody can fix with budget.
Or have someone walk it with you
The six questions above are yours to run. If you would rather have the answer than work it out, the free AI assessment is forty-five minutes with a developer who maps your processes live and tells you which are worth automating. No pitch, and you keep the shortlist either way.
The formal version
Our AI strategy and roadmap engagement does this properly: a structured audit of data infrastructure, technical capability and workflows, every viable use case scored against impact and feasibility, a vendor-neutral technology recommendation and a phased roadmap with go/no-go gates. About six weeks, fixed fee, and we publish the range.
Common questions
Is this a paid engagement?
The self-assessment on this page is free and yours to run. Our formal assessment, which produces a scored baseline, a prioritized use-case list and a costed roadmap, is a paid engagement lasting about six weeks.
What if we score badly?
That is a useful result, not a bad one. Poor data accessibility usually means the first project should be a data project. That is a cheaper thing to discover here than four months into a build.
How long does the formal assessment take?
About six weeks, including a discovery workshop, system review, use-case scoring against an ROI framework, and a presentation to leadership with a business case ready for board discussion.
Walk through your answers with us
Bring your six answers to a 30-minute call and we will tell you what we would do first, at no charge.
Book a callThe firm behind the firm