Insights · AI Adoption
Start with the decision, not the model
The first question about AI should not be which tool to buy. It should be which decision, made faster or more accurately, would change the business.
Most conversations about AI in a business begin in the same place: a tool. Someone has seen a demonstration, a vendor has made an offer, or a competitor has announced something. The organization then goes looking for a problem the tool can solve.
That order is backwards, and it explains why so many AI pilots work in a technical sense and still change nothing. The tool performs as promised. Nobody's work is different afterwards.
Begin with the decisions you make every week
Every organization runs on a small number of decisions it makes again and again. They are rarely dramatic, and they are rarely written down as decisions at all. They look like this:
- Which enquiries to answer first, and which to decline.
- What to quote for a piece of work.
- Which invoices to query before they are paid.
- Which clients are likely to leave, and what to do about it.
- What a document says, and whether it matters.
These are where the value is. A decision made a little better, many times a year, is worth more than an impressive capability used once.
Three questions for every candidate
When we help a leadership team choose where to start, we put each candidate decision through three questions.
- How often is it made? A decision made daily by several people is a better place to start than one made once a quarter by the board.
- What does a wrong or late decision cost? Not in theory, but in this business: lost work, rework, write-offs, staff time, client goodwill.
- Is the information needed to make it already recorded somewhere? If the answer depends on knowledge that lives only in one person's head, the first project is writing it down, not automating it.
A decision that scores well on all three is a strong first use case. One that scores well on the first two but not the third is a strong second project, after the groundwork is done.
Then decide what the machine should do
Only now does the question of tools become useful, and it becomes much narrower. In most businesses, AI is best placed to do the preparatory work around a decision: gather the relevant information, summarize it, draft a response, sort a queue, flag what looks unusual.
The decision itself, and the accountability for it, stays with a person. That is not caution for its own sake. It is what lets a team trust the new way of working enough to use it, and it is what lets the business explain its decisions to a client, a regulator or a court.
The useful question is not what AI can do. It is which decision you would like to make better, and what a person needs in front of them to make it.
What to bring to a first conversation
If you are considering where AI could help your organization, you do not need a technical brief. Bring a list of the decisions your people make most often, a rough sense of which ones go wrong and what that costs, and an honest view of where the information behind them lives. That is enough to tell whether there is a project worth doing, and where it should begin.
NLM Cognitive Solutions
To discuss this article or your own situation, write to hello@nlmcognitive.com.