EXCLUSIVE: UiPath CMO Michael Atalla talks AI in action

Good morning {{first_name| Artificial intelligence lovers}}. Artificial intelligence has changed the world we live in. Every CEO is talking about AI, and most of their employees are wondering what it means for their paychecks.

In the middle of this transformation sits UiPathwhich just celebrated the fifth anniversary of its IPO – has evolved from a company that was once in the business of automating tasks to one that now orchestrates how AI agents, automation, and people work together.

We sat down with the company’s marketing director, Michael AtallahTo understand what’s really happening inside organizations: why AI promises often fail, who wins, and what it means for everyone whose job is changed by this technology.

Rundown: UiPath’s pitch five years ago was simple: automate the task. Today, it’s even more ambitious – coordinating how every AI agent, robot, and human works together in a workflow.

Cheung: Five years ago this week, UiPath rang the IPO bell as the leader in robotic process automation. Today you are promoting the “agent business format.” What’s the biggest thing that’s changed in betting, and what’s stayed exactly the same?

Atallah: Five years ago, the promise was simple: automate the task, free the person. You succeeded. It still works. But if you look at most organizations today you’ll find dozens of automations running in parallel with no real way to connect them to each other, or to what the company is actually trying to achieve.

The question customers used to ask us was, “Can we automate this?” The question now is, “How can we make all of these things work together?” Coordination is the answer. AI agents, automation, people, and systems work end-to-end, with visibility across everything.

Atallah added: The unchanged bet: Technology should remove friction from people’s work, not add new types of friction.

Cheung: I spent 15 years at Microsoft leading the marketing of Office through its transformation from Office 365 to cloud-based. What has this taught you about moving organizations through a paradigm shift that most AI leaders today are missing?

Atallah: In 2011, I was demonstrating Exchange features like Conversation View and the Don’t Reply All button to customers who were skeptical about moving their email from a server they could physically touch. This was my education in Office 365.

You could have the right product and lose the customer if you can’t help them rethink how work gets done. We weren’t selling cloud software. We were asking people to change how they collaborate, where they store information, and whether they trust a system they can’t see.

Today’s AI conversations focus on the model. What it could do in theory. Companies don’t care about theory. They care about whether it works reliably under real conditions, within a real workflow, and with real accountability. The companies caught up in the cloud transformation were not lacking in ambition. They just lifted and moved without redesigning anything. This same pattern is happening with artificial intelligence right now.

Why it matters: If your team is evaluating AI tools right now, the question to ask at the next vendor meeting needs to be rephrased from “What can this model do?” to “What should our workflow look like for this to work?” If you get this wrong, the company could end up being among the 70% to 80% of pilots that never outperform AI pilots.

Rundown: Atallah says the main problem behind lagging AI initiatives is a lack of coordination. Whether in beta or deployment, tools that operate in isolation and disconnected from each other and business goals are where costs pile up and ROI disappears.

Cheung: Between 70% and 80% of effective AI initiatives never make it out of the pilot phase. What are the honest reasons why most companies fail?

Atallah: AI pilots always work in isolation. One agent in one corner of the business. Automating one into another. No vision between them. The pilot succeeds, leadership asks what’s next, and no one has a real answer. Costs accumulate. It is difficult to measure results. Eventually, someone decides it’s not worth it.

Organizations that have passed this stage are not doing anything radical. They have stopped treating AI agents as deployment tools. They began to treat them as components of a larger, controlled workflow. This is the whole game.

Cheung: A survey found that nearly half of organizations describe AI as a “huge disappointment” despite huge investment. What goes wrong after publishing?

Atallah: No one intends to fail at this. Ambition is there all the way. From the CEO to the person whose Tuesday is supposed to get easier. So when you see numbers like that, it’s not a motivation problem at all.

What I hear from clients is a coordination issue. They performed automated tasks. They have artificial intelligence tools in operation. But they are certainly not connected to what the company is trying to achieve. The return on investment disappears in that gap.

Clients who break through start with a different question. Not “What AI tool should we buy?” But “where does the work start, where is it delivered, and where are decisions made?” Start here, and the technology options will become clearer.

Why it matters: Coordination is an important piece of the AI ​​adoption puzzle. Redesigning AI workflows is important, but the next step is ensuring that the tools running within those workflows remain aligned with business goals. Once this is done, the value compounds – each tool becomes more useful because it works as part of the system.

Rundown: With advances in artificial intelligence, Atallah admitted that anxiety in the labor market is real. However, he rejected the idea that human participation has become optional, saying that roles change, not disappear.

Cheung: Three-quarters of AI experts are optimistic about the impact of AI on jobs, but only 23% of the public agrees. Who is closest to reality, and why this disconnect?

Atallah: genuinely? Both groups see something real. They just look at different parts of the picture. Experts see what the technology can do. The end user sees what the technology is being delivered and wonders what that leaves them with. It’s a reasonable reading of the signals.

What I will reject is the idea that human participation becomes optional as AI becomes smarter. The LLM student cannot ask “Should we?” She has no drive, no taste, no instinct for risk. Every system we deploy at UiPath still needs humans to oversee it, make decisions, and apply it in ways that add value. The role is evolving. The need does not disappear.

Cheung: Junior developer jobs have fallen by approximately 20% since 2024 as more senior roles grow. The UiPath CEO himself has said that the goal is to “grow without increasing headcount.” Is anxiety in the job market justified, or are people worried about the wrong thing?

Atallah: The concern is real, and it deserves to be taken seriously. A large number of entry-level roles are now being reshaped. This is nothing, especially for people who have built their career expectations around a different set of circumstances.

The redistribution process is more subtle than the headlines suggest. Routine and structured work is absorbed. But the work itself does not disappear. Its shape changes. New roles are emerging around workflow design, AI governance, and end-to-end process ownership. The request is there. The skills required are different.

My daughter is 13 years old. When you apply to colleges in five years, the jobs you’ll be competing for may not have been named yet. That’s cold comfort if you’re 24 now. But it’s also not the same thing as replacement.

Why it matters: Every worker watching AI transform their industry wonders: Is my job next? The answer is yes and no. Artificial intelligence will absorb the routine and structured parts of work. What it will not replace is judgement, taste, instinct, or the parts of a human being that require us to ask “Should we?” Now, it’s all about improving skills for the remaining work.

Rundown: UiPath deploys agents for tasks that involve ambiguity – such as interpreting an invoice that doesn’t fit the standard mold – while keeping humans out of anything that carries real responsibility, Atallah says.

Cheung: What does AI and automation working with people inside UiPath look like for someone who works in finance, HR, or operations?

Atallah: Imagine a finance team reconciling data across five systems and seeking approvals via email. Automation handles the structured and repeatable parts, such as data pulling, record matching, and routing requests. An agent steps in when there’s ambiguity – reporting an anomaly, interpreting an invoice that doesn’t fit the standard mold. The person in this role stops reconciling and begins reviewing exceptions, actually making decisions that require judgment.

A person’s time is diverted towards only the things he can do. This changes the way the job feels day to day, and it’s a bigger deal than it seems.

Cheung: With autonomous agent tools gaining significant attention, what types of tasks do you think AI agents will handle on their own in the next year or two?

Atallah: The “full autonomy” conversation precedes what actually happens. What I see in UiPath is more specific and, frankly, more interesting.

Agents are very good to deal with Unstructured data, context-aware decision making within a defined process, and exception management. Think document understanding, fraud detection, and customer service triage. The type of business where the inputs are not clean, and the rules-based system either fails or needs constant babysitting.

Deterministic, rule-based work still works better over traditional automation. Decisions that carry real responsibility, approvals, escalations, anything with consequences, stay with the people.

The near-term model is agents operating within coordinated courses of action. More cognitive responsibility. He still rules. It’s still noticeable.

Why it matters: While frontier AI giants continue to talk about full autonomy, UiPath’s approach is more consistent – deploying agents only where they do a really good job, and humans on higher-value tasks. It’s less exciting, but it keeps the mechanism moving with the results and is a version of AI adoption that actually holds up.

The one thing companies get wrong about AI more than anything else?

Atallah: We expect him to fix a broken process. AI makes good workflows faster and makes bad workflows more expensive.

What mistakes do companies make in how they deliver AI to their teams?

Atallah: Frame it as something that happens to people rather than something they will build. This is where anxiety begins and is usually avoidable.

If you weren’t at UiPath, what AI problem would you like to work on?

Atallah: The gap between what organizations believe AI will do for them and what it is actually set up to do. This is a clarity issue, not a technical issue. I find it really fascinating.