What has to happen around an AI agent before it pays off

If you’re weighing a tool that builds AI agents for your team, whether those agents pay off will turn on three things around them, and each one needs an owner before the first agent goes live. Your leaders agree what each number means, someone checks the agents’ answers against the systems their data came from, and your people change how they work.

Tools like that are getting good, and they deserve a serious look. McKinsey’s survey on the state of AI, published in August 2026, found that 80% of respondents said AI had improved their own productivity, and that those gains have yet to reach most companies’ results: 37% attributed at least some EBIT impact to AI, about the same share as a year earlier. All three of the things that decide the payoff happen around the agents, so you can start on them before you pick a tool.

Agreeing what each number means

An agent works from whatever your systems tell it. If sales and finance count revenue differently, an agent will answer from one of them, or mix the two, and sound just as sure either way. Gartner warned in May 2026 that without a clear understanding of the rules and relationships in an organization’s data, AI agents “are far more likely to hallucinate, introduce bias and produce unreliable results.”

Finding where your systems disagree is legwork agents can do. Deciding what revenue means is a call about the business, so it belongs to your leaders, and once they’ve made it, their answer can be built into the data every agent reads.

Agreed numbers paid off for me long before agents. At Hideaway Report, a members-only luxury travel guide where I led digital marketing, marketing analytics and technology, teams that read the same numbers, talked to each other and worked from the customer journey paved the way for a 100X increase in the visitor to paid-subscriber conversion rate.

Checking the answers against the source

An agent’s answer is a claim until it has been checked against the system its data came from. The check can be done by a person or by an agent built for it, as long as it goes back to the source and any decision that costs money comes to someone who owns it.

At Lost Horizon, the luxury glamping retreat I designed and own in the Texas Hill Country, an AI agent we set up went back to a year of booking records and found that all eight Expedia bookings had been sold at a discount. It brought me three choices and waited for my call, and you can follow that run step by step.

Lost Horizon is a small business, and the same rule carries into our data work: when our agents build your pipelines, our senior builders check every number against yours, and your leaders sign off before any AI acts on it.

Changing how your people work

In the same McKinsey survey, the small group it calls high performers, who credit AI with at least 5% of their EBIT and call its value significant, stood apart in how they work. Nearly three-quarters of them reported fundamentally redesigning workflows because of their AI use, against one-quarter of the other respondents. Someone has to lead a redesign like that, and it has to be made with the people whose day changes.

At a title research company, the step that needed redesigning was one everybody had stopped seeing: before the software could start, someone had to sort and bookmark every title search by hand. I sat with the company’s experts and made skipping that step a requirement, and once the system we built went into production, the staff stopped bookmarking title files by hand.

Before you choose

Put three questions to any tool or partner you’re considering: who agrees what your numbers mean, who checks the agents’ answers against your own systems, and who helps your people change how they work. We work inside the systems you already run, so a good tool can be part of the answer.

Bring the tool you’re weighing to a thirty-minute call, and we’ll look at where it fits in your business and who should own each of the three.

Read next

When AI makes a few engineers faster, spreading it to the whole team is a leadership job

All writing