Service
AI opportunity discovery
AI opportunity discovery is a short, structured engagement that ends in a decision. We map the repeated decisions in your business, size what each one costs today, test whether a correct answer can be defined, and recommend which single problem to build for first, including the recommendation not to build.
Who this is for
Teams that have been asked to "do something with AI" and want to arrive at a defensible answer rather than a list of pilots. It suits companies with more candidate use cases than capacity, which is almost all of them.
The business problem it addresses
Choosing wrongly at the start is the most expensive mistake in an AI programme, and it is invisible for months. Discovery front-loads that decision while it is still cheap to change.
The output is not a slide deck of possibilities. It is one recommendation, with the reasoning and the evidence that would confirm or kill it.
Signs you may need this
Several pilots running with no shared measure of success. A use case chosen because a vendor demonstrated it. Disagreement inside the business about what the system should even produce. An AI budget with no way to tell whether it worked.
What the engagement includes
Interviews with the people who do the work today. A map of the decision, the data behind it, and the cost of getting it wrong. A draft rubric defining a correct answer. A feasibility read on the data that would ground the system. A written recommendation with the risks stated plainly.
What is delivered
A written recommendation, a first evaluation rubric, an assessment of data readiness, an implementation outline for the recommended option, and an explicit list of what we could not establish.
When this is not appropriate
When the decision is already made and funded and you want validation rather than assessment. When the process is being redesigned anyway. When no one in the business can commit the few hours of expert time discovery requires.
Questions buyers ask
How long does it take?
Usually a few weeks, depending on how many people need to be interviewed and how accessible the data is.
What if the answer is not to build?
That is a valid and common outcome. It is cheaper to reach it in discovery than in month four of an implementation.
Do we need clean data first?
No. Part of the work is establishing what state the data is actually in, which is usually different from what people believe.
Who needs to be involved?
The people who perform the decision today, whoever owns the outcome, and someone who can speak for security and data access.