Building for ROI: stop measuring time saved and start tracking the real AI impact

As generative tools lower the technical barrier to building, measuring true AI ROI has become the ultimate differentiator between products worth creating and those that are simply easy to generate. Sara Maldon (Make, Inflect) breaks down the trap of internal duplication, why “hours saved” is a phantom metric and how builders can measure the true ROI of AI initiatives, ahead of How to Web Conference 2026.

“When building becomes effortless, everyone gets nudged to build. In an enterprise, you look up and realize 13 different people have built 13 different solutions for the exact same problem. You just paid to duplicate the same work 13 times over.”

Sara Maldon didn’t take the standard route into tech. She started out as a lawyer, realized traditional legal tools weren’t going to fix broken legacy systems and spent four years rewiring her brain to become a back-end engineer.

This chameleonic entry into the tech world gave Sara a rare dual perspective. She knows like the back of her hand how clumsy software adoption feels to someone outside the engineering team, but she also understands systems from the inside out.

We sat down with Sara ahead of her talk at How to Web Conference 2026, and we can confirm: she wears two distinct hats. At Make, she leads internal AI transformation for a company built entirely for builders. At Inflect, she works on the executive side of the equation, helping leadership teams measure whether their AI initiatives actually deliver full value in the market and showing them exactly how to get there.

As October draws near, Sara is gearing up to take the stage and tackle one of the messiest side effects of living in an era where shipping a feature takes a single weekend: why “hours saved” is a ghost metric for your productivity, and how to bridge the gap between pure velocity and actual business value.

Where a legal mindset meets back-end code

Transitioning from corporate law to software engineering is a journey where many might get lost in translation. Sara used it to build the ultimate foundation for driving enterprise change management.

“Moving from law to engineering took years of reprogramming,” Sara reflected. “They are entirely different mindsets. But going through that transition gave me a unique vantage point. I know how painful technology adoption is for a non-technical person, but I also know how the machine operates under the hood.”

While leading AI transformation at Make, Sara quickly realized that the hardest challenges had nothing to do with API keys or model selection – the friction belonged entirely to humans.

“Allowing the technology is just a prerequisite. Putting tools in people’s hands doesn’t mean a transformation has happened. Executive sign-in and leadership leading by example – that’s the ground-level enablement where the work actually sticks.”

When “anyone can build” backfires

The era of building anything at the push of a button has been upon us for a while now. What used to be a gatekept technical bridge between an idea and a working product has melted away. But with that melt comes a fresh operational challenge.

When building becomes cheap and accessible, every employee is encouraged to create their own custom workflows. In an enterprise setting, that enthusiasm often leads straight into the duplication trap.

For founders, product managers and builders creating tools for external users, this means that functionality alone won’t keep people around for long. The pendulum has swung toward craft, restraint, focus and user delight.

“Hours saved” or the ‘boo’ metric

When companies attempt to justify their AI spend, they almost always fall back on the same soft metric: estimated time saved. Sara, however, refused to track time saved early on in Make’s AI journey.

Instead of inventing abstract productivity metrics specifically for AI, builders should look at the hard operational numbers the business already tracks. If an AI project doesn’t move a core performance metric you were already monitoring, it’s just an expensive experiment.

The ultimate do’s and don’ts comparison? Stop counting “prompts generated per week.” Track whether your pipeline velocity or win rate actually improved.

The unused toolkit: solving problems users actually feel

Another trap in enterprise AI adoption is solving problems that sound great in theory but lack demand on the ground. Sara brings an example from her own experience.

Early on, Sara’s team built an AI-powered content generation toolkit designed to help the marketing team draft brand-aligned materials faster. Seven months after rollout, adoption was nearly zero.

“We were trying to solve a problem the team never wanted solved in the first place. An article generated by AI might hit the right topic and brand guidelines, but it misses human story craft.”

Before launching an initiative, builders must validate whether the end user actively wants that problem solved – or if leadership is simply forcing AI onto a process that was already working.

In a market obsessed with speed, moderation is your edge

If there is one core principle Sara wants builders, founders and investors to hold onto as AI capabilities accelerate, it is counterintuitive: slow down enough to ensure you are running in the right direction.

It’s the exact opposite of the relentless sprinting the tech industry has been through. But speed without intentionality simply gets you to the wrong destination faster.

Meet Sara Maldon at How to Web Conference 2026

The takeaway? Pair your ideas with intentionality, moderation and patience, and you’re already one step ahead.

Gear up to meet Sara in Bucharest this October – while she’s taking the stage to break down how to actually build for ROI and dissect the true impact of your AI initiatives. Sara will draw on two and a half years leading internal AI transformation to deliver a use-case-packed guide on what to track, where value gets diluted when giving every employee AI access and how to turn AI spend into numbers leadership can actually understand.

Get your tickets for How to Web Conference 2026.