Is It Worth It (Make AI Work - Part 2)

Part 2 of the mini-series on making AI work for your business. Last time, we covered problem selection: pick the right problem, define what a solution looks like, and avoid vague mandates. Today: once you have a problem, how do you know it's worth solving?

The Cost of Doing Nothing

Here's a question that doesn't get asked enough: What happens if this problem remains unsolved for another year or two?

If the answer is "minor inconvenience, we'd adapt," that's a signal. Not necessarily to stop, but to be honest about priorities. Plenty of AI projects get greenlit because they sound impressive, not because the underlying problem is urgent. And urgency matters. The organization's willingness to push through the inevitable rough patches of an AI initiative is directly proportional to how much the problem actually hurts.

On the other end, if the problem is blocking a strategic priority, you already know that. The tricky cases are in the middle: ongoing operational friction that's become so normal, nobody questions it anymore. "That's just how things work here." These are the problems worth digging into. They've been quietly draining resources for so long that people have stopped noticing.

One signal to look for: Has anything been tried before to solve this? If not, that tells you a lot about urgency. If yes, great. What was tried, and why did it fail?

Measuring Success (Before You Build Anything)

I've written before about how even soft outcomes can be measured. If something bothers your organization enough to warrant action, it must create observable consequences. Higher turnover, more complaints, slower delivery, missed deadlines. Find the observable effect and measure that.

The worst time to figure out your success metrics is after you've built the thing. "Did it work?" shouldn't be a philosophical question. Agree on what the numbers need to look like before you start, and a lot of the ambiguity around AI ROI disappears.

No metrics, no baseline, no business case.

Where Does the Freed-Up Time Go?

This is the one that trips people up the most. Say your AI initiative succeeds beyond your wildest dreams. It saves your team ten hours a week. Fantastic. Now what?

If nobody has a good answer, you don't have a value story. You have a cost story. And cost stories are weak, because the math never looks as impressive as you'd like. I've shown this before: if someone earns $100k and you free up 25% of their time, the naive calculation says that's worth $25k. But if that person generates $200k in economic value at full capacity, the freed-up time is worth $50k. You're unlocking their ability to do the high-value work they were hired for.

That only works if there is high-value work waiting for them. "General productivity improvements" is code for "we haven't thought about it." Specific, identified, revenue-generating activities that are currently neglected because everyone's drowning in busywork? That's a real answer.

And don't forget the Theory of Constraints angle. Freeing up time in a part of the workflow that isn't the bottleneck doesn't speed up the overall system. It gives someone more idle time. Make sure the problem you're solving sits on or near the constraint.

Before you invest in making things faster, know where the time goes and whether the system can absorb it.