Will This AI Do What I Need?

Most proposed AI solutions describe what the system will do. Fewer describe the business activities that will interact with the AI solution being suggested. That disconnect is usually the difference between a proposal that works and one that simply doesn’t meet expectations.

Four questions will tell you most of what you need to know.

What happens when the AI is wrong, and who finds out? Every AI system produces a wrong answer sometimes. A proposal that has thought this through will tell you how errors surface, who reviews them, and at what point a person has to be involved. One that does not mention it has not yet been designed for your business.

What does it cost to run the AI? Build cost is a number in the proposal. Running costs heavily depend on volume, and that volume will be enabled or constricted by business processes. Does the demonstration represent practical throughput? Do the math and figure out what a busy month costs.

What has to change about how our people work? If the answer is “nothing,” take a pause to consider if this is true. AI usually changes the steps people follow, the decisions they make, or the judgment they need to apply. A strong proposal should identify the roles affected, explain how their work will change, and show what support they will need to make the change stick.

Who owns the information going in and the output coming out, and where does it live? This is often where a proposal turns vague. If it’s not determined in advance, then this clarity will need to be part of the scope of work.

None of these are technical questions. You do not need to understand how the technology works to ask them, and you will learn a great deal from which ones get a clear answer.

The key point here is that there are many excellent technical people working on deploying AI. But for AI adoption to be successful, there are several business considerations that must be top of mind when it’s time to decide whether to proceed.