I’ve attended many Gartner conferences. This year was my first Symposium, where CIOs and technology leaders come together to tackle some of the biggest challenges facing their businesses. One session called “ROI or Die” challenged IT leaders to get serious about demonstrating the value they create for the business. One provocative suggestion was to publish a “Revenue Manifesto” that basically stated if an investment doesn’t generate new revenue, it shouldn’t be done.
I lead Enterprise AI and employee services at Zendesk. I spend a lot of time thinking about the value our investments in AI will deliver, and what those changes mean for our employees and the business.
Throughout the week, discussions kept returning to a central problem: how do AI agents fit into workflows? Who owns them, who is accountable for their decisions and how do we manage risks with the explosion of AI usage? Can we really turn faster tasks into better business results and how do we measure those results?
In a one on one analyst meeting, I raised a core question: how do we measure the value of AI for knowledge workers? Not just software developers, but everyone using AI day-to-day to automate workflows and enhance our personal productivity.
What we ended up talking about was what we wanted work to become. Where are we headed? How do companies break out of the race to the bottom? I came away with three takeaways:
1. Be clear about the outcome.
The Revenue Manifesto idea stuck with me because it demanded a clear commitment. It sounds like business 101, but it’s something we can easily lose sight of. If we’re asking for investment in IT, we should be able to explain what will improve.
That might mean more revenue, fewer billing errors, or a faster resolution for a customer. It could also mean an employee getting help they need sooner. Gartner describes this employee-focused perspective as “ROE” or Return on Employee.
Yet we tend to measure AI using what’s easiest to count: logins, licenses, and agents deployed. Those numbers tell us about adoption and activity, but they don’t tell us about outcome or resolution.
Getting work done faster doesn’t change your business, someone still has to decide what happens with the capacity that you create.
2. Follow the work all the way through
A familiar problem for cross department collaboration is that if one team finishes a task faster, the work can fall into another, slower queue. Also, if the work is performed agentically, you may have someone downstream needing to check the work and look for exceptions.
This provides a useful warning about what we call a “productivity gain”. Sometimes when one person’s day gets easier/quicker, someone else’s day gets harder. We still need to account for reviewing the AI output and correcting any mistakes.
We must take responsibility for the handoffs. Giving someone another tool without changing the surrounding process can leave the system with another tool to manage.
Before we claim to have given time back, we should ask the people doing the work if they have actually experienced that benefit.
A company that can demonstrate how AI has made working life better has something substantial to offer a prospective employee.
3. Let employees keep some of the gains
The question I kept coming back to all week was: who decides where the time goes that is saved? There is productivity FOMO in the conversation around AI. Competitors announce efficiency gains and CoWork harnesses and models are introducing new skills weekly, so the pressure to capture more continues to build.
We’re asking employees to adapt to new ways of working while continuing to deliver. If the only benefit they receive from using tooling to gain efficiency is just a longer list of expectations, we shouldn’t be surprised when their enthusiasm wears thin.
I believe companies have an opportunity to make a different commitment: some of the capacity AI creates should improve employees’ lives.
Of course, there will be legitimate demands on that capacity. Businesses need to grow, serve customers, and pay for the technology. Leaders will have to make choices. I want the employee’s share to be an explicit part of those choices. My bet is that companies that figure this out will have an advantage in attracting and keeping good people.
A company that can demonstrate how AI has made working life better has something substantial to offer a prospective employee.
And improving employees’ lives is worth pursuing, even when we can’t assign every benefit a dollar value.
I’d add that commitment to the Revenue Manifesto: employees should be able to point to something in their lives that improved because their company invested in AI.
If AI helps someone finish their work at five, we should be willing to let five be the end of their working day.
