AI Made You Faster. Did It Make Your Company Better?

A conversation with Marko Horvat, SVP of Business Transformation at ELB Learning

Marko Horvat has worked where finance, strategy and organizational change meet. He’s been a CFO and a controller in healthcare and consulting, and a VP at Gartner, where he advised hundreds of CFOs on what to change first. He’s now SVP of Business Transformation at ELB Learning, a 25-year enterprise learning and development company that spends more and more of its time helping organizations figure out AI. He also writes for the Forbes Finance Council and sits on a board.

On a recent episode of Building For Others, he named a problem most large companies are living with and few can describe clearly.

Individual productivity and firm productivity are not the same thing

Individual productivity and firm productivity are not the same thing. Marko keeps returning to this idea, and most enterprise AI strategies fall apart once you hold them up against it.

His example: say he spends time setting up Copilot to write his emails. His natural style is three sentences, the occasional typo. Copilot turns that into polished, multi-paragraph replies, and he’s proud of them. Then he sends one to Sachin Kamdar, elvex’s CEO, who gets hundreds of emails a day and would much rather have the short version. Sachin actually builds his own bot to shrink long emails back to something readable. Marko’s productivity gain needed a second productivity gain on Sachin’s side to cancel it out. Two people did extra work, pulling in opposite directions, on one email exchange.

“Did I get more productive? I guess,” Marko said on the episode. “Did you get more productive? I guess. But we didn’t really solve for a business problem that helps the business get better.”

That’s what goes wrong when everyone uses AI on their own. With no shared context, no common knowledge of how the work gets done, what the output should look like or what the next person needs, the gains disappear at the handoff instead of adding up. The handoff is where value actually moves from one person to the next, and most AI platforms ignore it.

Marko takes it further. AI has made starting things very easy, and it’s much less clear that it helps anyone finish them. A four-hour task now takes 45 minutes, so people have three and a half hours free to pick up something new. That adds more half-finished work, which stalls as soon as it needs someone else. “You can be incredibly individually productive,” he said. “But unless you’re the only stakeholder and sole decision maker in that loop, you can only get so far.” Sooner or later the work needs input from someone else, and that’s always been where things get stuck.

So what do you do about it?

Better prompts and more training won’t fix a structural problem. What one person does alone has to become something the team does together.

When someone finds a better way to do a task, the people around them should see it, instead of it sitting in one person’s chat history. The context that makes AI useful (how your company describes its work, what good output looks like, what the next person in line needs) has to be shared and kept up to date for the whole organization. Nobody should have to rebuild it from scratch each time.

And AI needs to be present where people hand work to each other. The points where one person’s output becomes someone else’s input are exactly where it should be, not only in the stretches where people work alone.

Both of those need a platform that covers the whole company.

Enterprise AI has made people more productive. Those gains aren’t showing up as company results, because the tools were built for individuals and rolled out to companies.

The Model Isn’t the Moat

From his finance background, Marko adds another point: the model itself is quickly becoming a commodity.

He cites MIT research from last summer that compared frontier and open-weight models on a common benchmark. At the start of the study, open models were six months behind. By the end of the summer they were three months behind, and the economics of model pricing suggest the gap will keep closing. “The LLM behind it is rapidly becoming like a commodity,” he said. “It’s all about how you use the tool.”

So a company’s edge is in what surrounds the model: what the organization knows, its own workflows, and the company context that makes AI output match how the business really runs instead of how an average user writes prompts. Send the same prompt to Claude and everyone gets roughly the same answer. The companies that get ahead will be the ones that put their own way of working into their AI.

That’s why Marko doubts the “prompt engineer” idea that was everywhere in 2023 and early 2024. AI gets easier to use with every release, and the technology is mostly solving the usability problem on its own. The harder work, which most organizations keep putting off, is governance and context: deciding what AI should be used for, how it’s organized and which company rules it follows.

The Implied Bet Nobody Is Naming

When a company pays for AI by cutting headcount or freezing hiring, it’s betting that AI productivity will show up fast enough, across the whole workforce, to make up for the people it didn’t hire. Marko calls that an implied bet, and the early results don’t look good.

McKinsey research shows that top performers beat average performers by an exponential margin, not a fixed one. If you plan revenue or output around an AI-boosted skill level that never shows up across your workforce, you end up without the assets you were counting on. “That’s a balance sheet liability,” Marko said, drawing on his Forbes Finance Council work. “It means you do not have the assets you thought you were going to have to achieve the results you thought you were going to achieve.” A $500M AI budget that burns through in a month is a spending problem. An organization that can’t produce what it planned for is a strategy problem.

You Can’t Hire Your Way Out

Most companies try to fix this with consultants and specialist hires. Marko thinks that’s a mistake and says why.

AI moves faster than the job market can. Someone who has built agentic finance workflows for an enterprise, with real production deployments and an audit history, is mostly a hypothetical candidate right now. The few people who have done it are either too valuable to their current employer to leave or busy starting their own company. “You have to train your own people,” he said. “This is not a problem you can hire out of, because the pace of technology is increasing faster than the workforce can keep up.”

ELB Learning’s work is built on that idea. The company has spent 25 years helping companies learn new ways of working. Marko’s approach is to make the organization itself own its AI solutions, building them into workflows, onboarding and career development so the skills keep going on their own and don’t depend on one champion who might leave or a consultant who certainly will.

He’s also seen companies skip that step. Clients come to him asking for use cases and agent builders, and he turns the question around. “What is it exactly that you want them to do?” he asks. Usually the answer is something they heard at a conference. His advice doesn’t change: find the workflow problems first, then work out where AI helps with them.

Stop Obsessing Over Data

Marko ended with a view you might not expect from a CFO: companies pay too much attention to data and too little to decisions.

Moneyball came out in 2003 and The Signal and the Noise in 2012; we’ve been having more or less the same conversation about data-driven culture for two decades. Meanwhile data has become a goal in itself. Organizations build lakes and warehouses full of it with no clear idea of which decisions it’s supposed to support.

AI has made this worse. Because it’s good at querying data and finding patterns, it’s now easy to find data that backs whatever you already wanted to conclude. “You can get the data to fit whatever decision you want,” Sachin said during the conversation. The discipline is to know which decisions you need to make and work back to the data they require.

Marko’s advice: don’t start with the data you have. Start with what you’re trying to decide and what you’d need to know to decide it well. When business changes this fast, data from five years ago says little about what you’re doing today. “You’re generating data every day,” he said. “By the time your initiative goes live, you’ll have six months of data designed for exactly what you’re building.”

Marko Horvat is SVP of Business Transformation at ELB Learning and a Forbes Finance Council contributor. You can find him on LinkedIn and follow his writing on AI, finance transformation and workforce strategy. Learn more about ELB Learning at elblearning.com.

Building For Others is elvex’s podcast about the people building AI things that other people use. No hype, no vaporware: just practitioners talking about what they’ve made and what they’ve learned.

Bring your agent. Bring your team.