The distance between how many employees say they use AI and how many companies can point to real business value has become the defining problem of enterprise adoption. The reason is rarely the technology. Organizations rolled out access, told everyone to use it, and never built the structure underneath, which means the time and money going in has no dashboard, no governance, and no honest accounting of what comes back out.
Dan Slagen is taking a more measured approach. He's the Chief Marketing Officer and Chief AI Transformation Officer at Zapier, where he leads the company's AI transformation strategy across its business and operations. A Forbes Top 50 Entrepreneurial CMO and an AI-transformation instructor, Slagen has spent the past year turning Zapier's own teams into a working model for what enterprise adoption looks like when it's done deliberately rather than declared. His diagnosis of the AI value gap lands on a simple piece of infrastructure most companies never built.
"We have a dashboard for everything else in the world, but almost nobody has an AI measurement dashboard," he says. The dedicated dashboard layers on top of a handful of components he believes transformation depends on.
The four pieces almost no company has in place at once
Slagen organizes transformation around four conditions, all of which he says must be present for it to work. A majority of companies, in his experience, are missing most of them.
The first is leadership that builds, rather than just mandates. When he pushed his team onto developer tools, he acknowledged that he was in the same unfamiliar territory they were. "We're marketers, so we're learning it for the first time, but I was in there learning it with them, building dashboards and products and ideas with them in the weeds."
The second piece is talent and culture, held to a standard and learning out loud. "Training is fine, but hackathons or buildathons accelerate learning a lot faster," he shares. "Getting teams together to go solve a specific problem or giving them a bunch of big problems and having them do a buildathon has worked really well for us."
The third is tooling and access, which he thinks can strangle more transformations than anyone admits. "You see a lot of companies that are stunting growth because they won't let certain tools in the door. They won't implement the governance and security measures that would allow more people to build." For example, some organizations won't allow employees to connect a tool like ChatGPT to their internal workflows because they lack proper controls. "That doesn't make any sense," Slagen asserts.
That leads directly to the fourth condition, governance, which Slagen frames as permission rather than restriction. Done right, it frees people to build while letting the security team sleep. Done wrong, it produces a policy that everyone ignores. "A huge number of people are already building within your company using their own personal email or they've just gone and done it and you don't know about it," he says. "So you have to have systems in place that allow people to build."
The dashboard brings it all together
The aspect of AI transformation Slagen thinks is scarcest is a daily measurement dashboard built so that it actually means something. Most companies, he says, know what AI costs and almost nothing else. "You know how much people are spending, but you have no idea what they're actually producing or if it's ROI positive."
His approach doesn't require a complex new metric, but a new view of metrics companies already trust. "Every SaaS company in the world shares their normal metrics with their employees, like how much ARR we're making or our ARR per headcount. All I'm arguing for is to reshare those exact same metrics through the AI usage lens." How far each department is into the token budget, how long it takes to build a new system, and how long after launch it takes to realize value all are solid examples he offers.
Making the business case a conversation, not a crackdown
Slagen cautions that daily visibility into spend can surface numbers that would alarm most executives. His response is not to shut the spender down, but to walk through the math with them, because usually no one told them what the tool costs.
He describes an employee running every sales call against the product roadmap on a constantly self-updating spreadsheet. "The cost was insane." Rather than kill it, he walked through the numbers: if this works, how many leads, how many opportunities, how much revenue does it produce, then compare that against current spend. "When you talk to people about it, they say, 'Oh. I didn't realize that."
He's careful to put the responsibility on leadership rather than the employee. "We never asked our employees to think about that as an industry. We told them to use AI, and they did, so it's our fault for not having set these guidelines in place."
Let teams know what "next" means
Underneath the equation is a map of where an organization is trying to go. Slagen's teams use a five-rung ladder that runs from casual experimentation (dubbed 'AI curious') to fully autonomous systems, and the value of it is that it turns a vague ambition into a plan a team lead can act on. "AI curious is someone who's opened up any of the LLMs and they're doing some general prompting," he says. From there people build their own AI brain, a digital twin, then start implementing systems, then move to a shared AI working space where teams collaborate. "So you transition from individual AI to institutional AI, and then you go fully agentic," he explains.
The ladder is what makes everything else measurable. "You can then actually go to your team leads with a structure, a plan in place, and a guide for how they move forward," he says. "Most companies don't have those things. No one's giving their employees actual structure and guidance on how to do it. It's all open-ended, which can be very overwhelming."
In Slagen's telling, that's the whole difference between a company that says it uses AI and one that can prove what the AI is worth. The tools are largely the same everywhere. The structure to measure them, govern them, and tell people where they're headed is what almost no one has built yet.
