You want AI ROI? Know what drives your business

By Michael Domanic, Section Head of AI

I hear some version of this almost every week: “We need to start seeing ROI from AI.” My first question back is always: What metric are you expecting to move?

The answer is often employee productivity. And while that’s a real outcome, it’s a tough metric to use to justify your AI investment in years 1 and 2. It’s slow to emerge, hard to measure, and mostly visible when you can freeze or cut headcount – which not all teams want to do.

There’s a faster path to visible AI ROI, and it requires knowing four (relatively simple) things: What your business goals are, which metrics ladder up to those goals, what work actually drives those metrics, and where AI can best accelerate the work.

The reverse engineering exercise

The approach I've been advocating for is pretty straightforward, even if most companies haven't done it yet.

1. Start with your business goals. Not AI goals - business goals. Your revenue target for next year, your margin goal, your customer retention number, whatever the leadership team has already agreed matters most. These should exist with or without AI.

2. Break those down into the contributing metrics. If your goal is revenue growth, what feeds that? In a sales organization, it's probably some combination of meetings booked, conversion rates at each stage, average contract value, and time to close. In a product organization, it might be feature velocity, customer adoption rates, and expansion revenue. Every business has these building blocks, and most leaders can identify them if they sit down and think about it for an hour.

3. Ask: Which of those levers can AI accelerate? Not in a theoretical “AI could help with anything” way, but specifically. Could you book more meetings if you had AI-powered SDR outreach? Could you accelerate time to close with AI-assisted proposal creation?

When you go through this exercise, you end up with a much clearer picture of where to point your AI investment. Instead of “everyone should use AI more,” you get something like, “Our biggest lever for revenue growth is conversion from first meeting to opportunity, and we think AI can improve that by giving AEs better preparation and faster follow-up.” That's an investment thesis you can actually measure against.

4. Measure on the KPI level. “Is AI making us more productive?” is almost impossible to answer at the org level in any satisfying way. "Did AI-assisted discovery calls convert at a higher rate than non-assisted ones?" is a question you can actually answer with data.

This is what makes the building-block approach so much more practical. You get quick wins that are genuinely meaningful, even if they're small. Increasing qualified meetings booked by 15% might not change the business overnight, but it can be extrapolated to revenue growth. If you know your conversion rates and your average deal size, you can model what that 15% lift is worth over a year. That’s a real number that your CFO can work with - and it's infinitely more useful than “people report saving about two hours a week.”

In reporting AI-influenced KPIs, you’ll probably get a question like: “How do you know it’s AI moving this and not some other variable?” The honest answer is you won’t, entirely. Perfect attribution is hard - not just with AI, but with pricing changes, positioning updates, new hires, training programs, etc. But you can still build a credible case. The simplest is before-and-after - baseline the metric before you deploy AI, then track the change. If your discovery call conversion rate was 22% for six months and it moved to 28% in the two months after you deployed AI (with no other huge compounding variables), the correlation is strong enough to be meaningful.

Another way to do this: Compare AI-assisted work to non-assisted work happening in parallel. If half your AEs are using AI-powered meeting prep and half aren't, the performance difference between the two groups is about as close to a controlled experiment as you're going to get in a business setting.

What this means for your AI program

If you take this approach seriously, it changes how you deploy AI, how you do change management, and how you measure success.

On training: Instead of rolling out general-purpose AI coaching to everyone, you can focus on building use cases and automations that directly target the levers you've identified. Your marketing team’s AI program looks different from your sales team’s program because they’re targeting different building blocks.

On champions: Your AI champions have a clearer mandate. Instead of “help people get better at using AI generally,” it's “help your team build automations and redesign workflows that improve these specific metrics.” That focus makes their work more impactful and easier to evaluate.

On measurement: You have a before-and-after that's tied to something the business already cares about. You're showing movement on metrics that leadership was already tracking, and attributing some of that movement to the AI investments you made.