Find the AI worth building, and the controls that let you ship it
Centricone Technologies runs the assessment before the build: which use cases repay the effort, whether your data can support them, what the regulation in your sector actually requires, and the governance and audit trail that get a project past a risk committee.
Most AI assessments produce a slide deck. This one produces a shortlist you can start on, the list of use cases we told you not to build, and the controls that make the first one shippable.
support coverage across US and Canada time zones
of code reviewed, tested, and documented before release
The pressure to do AI arrived before the plan
Everything is a candidate and nothing is scoped
A long list of ideas from across the business, none costed, none tested against the data, and no agreed way to rank one against another.
Nobody owns the risk
Tools are already in use, procurement has questions, and there is no policy, no inventory, and nobody who can say whether a given use is allowed.
The regulation is a moving target
The EU AI Act, sector rules, and client questionnaires each ask a different question, and the answers do not exist in writing anywhere in the organisation.
for teams deciding what to build and how to govern it
Centricone covers both halves: the readiness work that ranks use cases against your data and your economics, and the governance work that gives you a policy, an inventory, and the evidence a regulator or a client will ask for.
Deliverables
What you get
- A ranked, costed shortlist of use cases
- A written read on your data readiness
- An AI policy and a system inventory
- Risk classification against the regime that applies to you
- A control set with named owners
- A scope and estimate for the first build
How an assessment runs
Gather the candidates and the constraints
Interviews across the business, the ideas already in flight, the tools already in use, and the rules your sector and your customers apply.
Test the shortlist against the data
The strongest two or three candidates checked against what your data can actually support, before anyone commits to a build.
Write the policy and classify the risk
Policy, inventory, ownership, and classification, sized to your organisation rather than copied out of a template.
Hand over a plan you can start on
A costed first build, the controls it needs, and the list of what we advised against and why.
When you should skip the assessment
This is a decision-making engagement. If the decision is already made, buy the build instead and keep the money.
- You already know what you are building and it is not contentious — go straight to the engagement that fits it.
- You want a document to satisfy a board rather than a shortlist you intend to act on. We are not the right firm for that.
- The organisation is small enough that the policy is one page and the inventory is three tools. Write it yourself; we will review it in an hour.
- The real blocker is that the data does not exist yet. That is a data engineering project, and it has to come first.
Why teams start with an assessment
A shortlist you can act on, not a slide deck
The use cases we advised against, in writing
Data readiness tested before anything is scoped
Controls mapped to the regime that actually applies
The capability behind this engagement
Once the shortlist exists, the AI development services page covers what building the first one involves: retrieval, model selection and evaluation, guardrails, logging, and cost control.
Frequently asked questions
Not sure this is the right shape? Tell us the situation and we’ll say which engagement fits.