UiPath posts first profitable quarter. Here's why it thinks orchestration is the missing piece

There's a lot of speculation right now about whether enterprise AI investments are actually delivering returns. Against the backdrop of the constantly-referenced MIT survey, UiPath just posted its first GAAP profitable third quarter – and says it is on track to be GAAP profitable for the full fiscal year for the first time. For a company that has spent the past year in turnaround mode under returning co-founder and CEO Daniel Dines, that's a meaningful milestone.

The numbers for Q3 fiscal 2026: revenue of $411 million, up 16% year-on-year (14% normalizing for currency). Annual Recurring Revenue (ARR) reached $1.78 billion, up 11%, with net new ARR of $59 million. Non-GAAP operating income hit $88 million, a 21% margin. The company beat guidance across the board.

The orchestration play

Not every problem needs an AI agent, and a lot of enterprises are figuring that out the hard way. The challenge is how to make it all work together, and that's the gap UiPath is going after.

The pitch is that enterprises need three things. First, reliable, rules-based automation for well-defined tasks – what UiPath calls deterministic automation, built on its Robotic Process Automation (RPA) heritage. Second, Large Language Model (LLM)-powered agents for the messier stuff: complex documents, conversations, situations where rigid rules don't cut it. And third, something to co-ordinate between the two, keep humans in the loop where needed, and make sure the whole thing is governed and auditable. That's what UiPath's Maestro platform is designed to do.

Dines explains:

In every conversation with customers and partners, the message is clear. Automation and AI are stronger together. Our deterministic foundation – enterprise-grade RPA and API automation capabilities – delivers the trust, scale and reliability mission-critical processes demand. On top of that, our leading AI capabilities bring adaptability, intelligence and speed. What brings it all together is orchestration.

The numbers suggest customers are buying in – over 950 companies are now developing agents on the platform, with more than 365,000 processes orchestrated through Maestro.

Looking for the ROI

Given how many enterprises are still waiting to see real Return on Investment (ROI) from AI, UiPath leans heavily on quantified customer outcomes on the call. A few examples:

One of the world's largest investment management firms chose UiPath because Maestro can work with different AI models and systems rather than locking them into one stack. The firm has run multiple agentic proofs of concept, already showing a 95% reduction in time to value. It has identified over 40 use cases that it expects to generate more than $200 million in savings over three years.

In another example, a US-managed care provider is using UiPath to tackle a backlog of more than 140,000 provider appeals. Agents classify forms, robots handle the processing, and Maestro orchestrates the workflow, pulling in humans only for exceptions. The target is 80% autonomy in year one.

USI Insurance Services is running a similar multi-agent orchestration deployment and expects over $32 million in savings over three years.

A reality check on agentic AI

Dines acknowledges that customer expectations around agentic AI are starting to come back down to earth:

In all fairness, there are customers that believe that AI agents will do everything. So they think very far-fetched in terms of swarms of agents that talk together. At the same time, I would say that the majority of our customers are starting to realize that their automation programs are actually quite important to power their agentic initiatives... Typically, they can come up with like 100 ideas that they call agentic. And then when we look deeply, we discuss that 50 of them are better suited for automation.

This is the core of UiPath's argument. It's not that agentic AI doesn't work – it's that it works best for certain kinds of tasks. If a process is well-defined and rules-based, traditional RPA is faster, cheaper and more reliable. If it involves judgement, ambiguity, messy documents or conversation, that's where LLM-powered agents add value. The key is knowing which tool to use where – and having something in the middle to coordinate the handoffs.

UiPath announced integrations with OpenAI, Microsoft Azure AI Foundry, Google Gemini, NVIDIA and Snowflake at its Fusion event earlier this year. UiPath maintains that this about interoperability, rather than commercial tie-ups.

The logic behind this is one that has come up in numerous use case discussions – enterprises don't want to be locked into one AI vendor, especially when the model landscape is shifting so fast. UiPath is continuing to position itself as neutral ground, aiming to reassure customers and buyers that it doesn't matter which LLM you've chosen, Maestro will be able to orchestrate it alongside your existing automation.

My take

This is a solid quarter, and the turnaround under Dines is starting to show in the numbers. More interesting is how UiPath is carving out its position. While much of the market is chasing the 'AI agents will do everything' hype, Dines is making a more nuanced case – the future isn't pure agentic or pure deterministic, but a hybrid where orchestration does the heavy lifting.

Rather than trying to out-agent the AI-native players, Dines is pointing to customers learning that half their 'agentic' ideas are better suited to traditional automation, which goes against the grain of vendor messaging in this space. The unspoken point is that UiPath has both tools in its kit, and Maestro to coordinate them. Whether that resonates depends on how many buyers have been burned by the 'agents will do everything' promise - and based on Dines's comments, that number is growing.

At the FUSION event earlier this year, Dines uses the analogy of a hybrid car engine to explain this – the idea that different power sources work together, each optimised for different conditions. I recently spoke with UiPath's Taqi Jaffri, Senior Director, Product Management, to dig deeper into what that hybrid model looks like in practice and where the boundaries fall between deterministic and agentic automation. More on that soon.