Which AI investments deserve backing?

Connect the customer opportunity to a commercial outcome.

Start with a customer task worth improving: choosing the right product, understanding an offer, completing a purchase or getting a problem resolved. Establish what a better experience would change for the business, such as more completed purchases, fewer cancellations or less repeat service work.

Compare the AI proposal with other ways of improving that task. A clearer offer or a simpler process may be the better investment. Follow the expected benefit through the journey so an improvement at one stage is not lost in confusing terms, an incomplete order or a difficult handoff.

Include the work that continues after launch.

Building or buying an AI capability introduces responsibilities for checking quality, handling exceptions and maintaining the experience as policies, products and systems change. The investment case needs an owner and capacity for that work.

A demonstration supported by intensive expert intervention can be a useful experiment. Expanding it requires accounting for the human work that made it possible. Treat an unfunded supervision or testing requirement as an unresolved part of the case.

From my writing: Before You Say “We Could Build That With AI”.

Give the pilot a decision to resolve.

Before approving a pilot, name the uncertainty that could change the investment decision. It might be whether customers can complete the task, whether employees can manage exceptions or whether the operating cost supports the commercial case.

Agree what finding would justify proceeding, changing direction or stopping. A successful demonstration is weak grounds for a larger commitment when the most important business assumption remains untested.