Five Questions to Ask Before Funding an AI Initiative

AI initiatives are often asked to become certain too early.

A sponsor wants a business case. A team produces a roadmap, an architecture, and a benefit estimate. The proposal looks responsible because it contains familiar planning artifacts.

But the numbers may be resting on assumptions nobody has tested: that the problem is common, the data is usable, people will trust the behavior, and the new workflow will save more effort than it creates.

Funding should not require pretending those questions are settled. It should require a disciplined way to settle them.

1. What meaningful outcome could change?

Start with the work and its consequence.

Who experiences the current problem? How often? What is delayed, missed, repeated, or made unnecessarily expensive? If the initiative succeeds, what will become observably better?

Avoid outcomes such as “increase AI adoption” or “improve efficiency” until they are connected to a real workflow. A useful outcome might be reducing the time a specialist spends gathering case history, helping a manager identify a risk earlier, or making a service available in situations where it is currently too costly.

If the outcome cannot be described without naming the technology, the opportunity may still be too vague.

2. Why might AI be appropriate here?

AI is useful when the work involves language, images, patterns, ambiguity, or a large space of possible inputs. It can help interpret information, generate a draft, recommend a next step, or coordinate a bounded workflow.

It is not automatically the best answer to every slow process.

Ask what AI contributes that simpler software, search, rules, or process repair cannot. Then ask what new uncertainty it introduces: variable outputs, review needs, data exposure, monitoring, or unclear responsibility.

The purpose is not to disqualify AI. It is to understand the trade being funded.

3. Which assumption carries the most risk?

Every proposal has a belief that deserves attention before the rest.

Perhaps representative data cannot be accessed. Perhaps users will not act on a recommendation they cannot verify. Perhaps exceptions are too common for the proposed automation. Perhaps the workflow saves two minutes but creates five minutes of review.

Name the assumption that would change the investment decision if it were false.

That assumption should shape the first prototype and the evidence requested from the team.

4. What will the organization need to operate it?

A model response is only one part of an AI product.

The operating system around it may include data owners, permissions, integrations, human review, escalation, evaluation, monitoring, support, and a process for changing behavior over time.

Ask who will own those responsibilities. If the answer is “the project team,” ask who owns them after launch.

This question often reveals whether the organization is funding a demonstration or a durable capability.

5. What decision will the next phase enable?

Do not approve “build the product” when the opportunity is still uncertain.

Approve a bounded phase with a clear decision at the end. The team may need to determine whether users can complete the workflow, whether the data supports acceptable output, whether review effort is manageable, or whether one use case is valuable enough to pursue.

Define what evidence will support proceeding, changing direction, or stopping.

A responsible funding decision does not eliminate uncertainty. It pays to reduce the uncertainty that matters next.

Smaller commitments can create stronger momentum

Leaders sometimes worry that a narrow first phase signals a lack of ambition.

I see the opposite. A focused investment lets the team move quickly without turning enthusiasm into a large, fragile commitment. It creates an opportunity to learn from real behavior before architecture and organizational expectations harden around the wrong idea.

If the evidence is strong, the next investment becomes easier to explain. If the direction changes, the organization has learned while the cost of change is still low.

Before funding an AI initiative, ask for clarity about the outcome, the role of AI, the riskiest assumption, the operating reality, and the next decision.

The quality of those answers matters more than the length of the roadmap.