When your board asks whether the money going into AI is paying off, you want three honest answers, and most engineering organizations don’t yet have the numbers to give them. In a Gartner survey published in October 2026, 44% of CEOs said they struggle to demonstrate AI-driven financial returns to stakeholders, and inside engineering only 31% of organizations measure the impact of their AI developer tools (LeadDev, 2026).
Why usage numbers won’t answer it
The easiest numbers to take to a board are usage: how many engineers use the AI tools, how much of the code AI wrote, and how many hours people say it saved. The first two show the tools are in use, and the third is people’s own estimate. None of them says whether the business is better off, and usage can climb for months while release dates keep slipping. A team can feel faster without shipping anything sooner, and the usage numbers look the same either way.
Three questions, in order
“Is AI paying off?” sounds like one question, but a good answer has three parts, and the order matters.
The first is whether engineering is building what the business needs. AI makes it cheap to build the wrong thing quickly, so speed only counts once the work is pointed at your goals. The second is whether that work is reaching customers faster without getting riskier. Speed and safety belong on the same page, because AI can make a team faster while making its releases riskier.
The last is whether the AI spend is paying back, and it comes last because a cost means little until you know what it bought. Set beside the first two answers, it tells your board whether the gains so far are worth what they cost.
What you can walk in with
Once the first change to how your team works is live, you take your board a readout built with your CTO, who presents the engineering side. It gives the three answers in that order, each set against where you started.
None of this needs new software. The history already sits in your code, deploy, ticket and incident tools, and the spend sits in your AI tools’ billing. When my team starts with a company, we pin down its goals first and take a baseline from what’s there, and where that history is thin we start with estimates and a short survey of the team. For delivery we use DORA’s software delivery metrics, the standard measures of how fast and how safely changes reach customers, and the Core 4 framework from DX, a company that researches developer productivity. Everything is measured by team and never by person, and targets go on your business goals rather than on these measures, so nobody is graded on them.
As the work gets going, you can see whether the first signals are showing up, with progress on your goals, work waiting less for review and changes reaching customers sooner without failing more often. Early numbers can dip while a team changes how it works, and DORA’s 2026 report on AI returns expects most organizations to see a temporary productivity dip early on. That’s why the first signals are an early read, and the readout is the one to judge by.
That readout is what working with us is built to give you. If AI is on your next board agenda and you’re not sure yet what you’ll show, I want to talk to you.