Two years of platform transition, revenue pressure, and changing expectations around automation and AI and now an improvement in performance – UiPath's CFO and COO Ashim Gupta attributes this less to a shift in product direction and more to changes in execution, operating cadence, and customer engagement.

UiPath recently achieved its first quarter of net new ARR growth in two years and reached GAAP profitability. Gupta positions those results as the outcome of operational discipline rather than a response to competitive pressure from generative or agentic AI.

Execution discipline – not repositioning

Gupta argues UiPath’s recent performance improvement is driven “80%–90%” by execution changes rather than new product launches. He points to earlier renewal planning, tighter account management, and increased attention to customer deployments after contracts are signed.

According to Gupta, UiPath now plans renewals further in advance and spends more time reviewing top accounts and upcoming quarters. He also describes greater focus on what happens after the initial sale – the period between software purchase and value realisation – which he characterises as the point of highest risk for both vendor and customer.

Gupta frames these changes as operating discipline rather than restructuring. He's careful to note that the company aims to avoid becoming satisfied with early progress in a market that continues to evolve, describing a shift toward consistency in internal rhythms and expectations, alongside ongoing pressure to improve:

“Repeatability means improvement. It doesn’t mean staying the same.”

UiPath generates approximately $1.8 billion in ARR, holds around $1.5 billion in cash with no debt, and has crossed 20 % operating margins. Gupta says the company now evaluates capital allocation inclusive of stock‑based compensation, rather than relying solely on non‑GAAP measures.

Deterministic and agentic automation in the same process

Gupta describes UiPath’s automation strategy in terms of complete process transformation rather than individual technologies. He explains that enterprise workflows typically contain a mix of repetitive, rules‑based steps and tasks that require reasoning or contextual judgement, explaining:

“When you combine deterministic and agentic automation capabilities, that allows for the most breadth of a process to be transformed.”

He illustrates the point with a customer workshop focused on financial reconciliation. The exercise generated roughly 100 automation ideas, which ultimately split evenly between traditional RPA and agentic approaches. Gupta explains that the customer did not distinguish between robots and agents, but instead measured how many steps of the process could be automated.

UiPath’s position is that its existing RPA platform provides the foundation for agentic capabilities, rather than being displaced by them. Gupta says the company no longer sees core RPA deals in isolation, but as entry points into broader platform discussions that include orchestration, testing, and agentic automation.

Human oversight and agent reliability

If asked about fully autonomous processes, Gupta offers a measured view, explaining that UiPath evaluates agents based on accuracy and outcome consistency, and that different thresholds determine whether human oversight remains necessary.

An agent operating at 70 % accuracy may still deliver productivity gains, Gupta says, but requires human review. Even at higher accuracy levels, autonomy depends on the risk profile of the task. He notes:

“I don’t think there’s ever a day where I’m going to let an agent go and put an SEC filing in if there’s a 1 % chance for an inaccuracy.”

Lower‑risk workflows, such as approving purchase orders within budget limits, may be suitable for unattended operation at sufficiently high accuracy. Gupta adds that some processes are unlikely to ever be fully unattended, while others may evolve toward autonomy as confidence increases. He links this progression to verticalisation, arguing that deeper integration into specific domains and systems can improve agent performance by embedding more contextual knowledge.

Earlier in 2025, UiPath acquired Peak to deepen its industry‑specific capabilities in areas such as pricing optimisation and inventory management. Gupta also emphasises time to value as a key driver of ROI, describing the period between purchase and deployment as the most vulnerable phase for enterprise software adoption.

Open ecosystem and data control

UiPath has announced partnerships across the AI and data ecosystem, including with NVIDIA, Google, Microsoft, Snowflake, OpenAI, and Anthropic. Gupta says the company’s open architecture is intended to accommodate customer choice rather than enforce standardisation.

“No customer wants to be locked into a finite set of vendors,” Gupta says, pointing to uncertainty around model evolution and enterprise AI governance requirements.

He describes meeting a large financial institution that permits only specific models across its environment for security reasons. A narrow partnership strategy, he explains, would exclude customers operating under such constraints. He differentiates between partnership roles, noting that model providers enable access to new capabilities for agent development, while Snowflake supports data integration without transferring control. Gupta describes UiPath as a “zero‑copy” platform, meaning customer data remains under customer governance rather than being ingested into UiPath systems.

Process orchestration versus agent orchestration

With confusion around orchestration terminology, Gupta acknowledges that UiPath has contributed to ambiguity in the market. He draws a distinction between agentic orchestration – triggering and governing individual agents – and process orchestration, which spans deterministic automation, probabilistic tasks, and human activity.

Process orchestration, he explains, is concerned with observability and governance across an entire workflow. Gupta describes UiPath’s Maestro platform as providing a live view of processes in motion, rather than static process maps. In this framing, orchestration functions as an operating layer that allows enterprises to monitor document ingestion, human review queues, and downstream execution in real time. Gupta contrasts this with data‑observability tools that focus on pipelines rather than business processes.

Gupta characterises UiPath’s federal business as variable rather than categorically strong or weak, noting differences in procurement scrutiny and initiative maturity across agencies. He says the company views current conditions as a “new normal” similar to commercial markets, where customer‑specific dynamics influence timing and deal composition.

On margins, Gupta says cloud adoption created a 200‑basis‑point gross‑margin headwind in fiscal 2024, and that SaaS‑related pressure is expected to continue as agentic functionality drives further cloud usage. He adds that UiPath does not currently see margin risks that would alter its broader operating model.

My take

Gupta’s remarks describe a company prioritising execution over reinvention. Rather than presenting agentic automation as a replacement for earlier technologies, he consistently frames it as part of a broader effort to automate entire processes while maintaining control, governance, and predictability. Enterprises are not choosing between RPA and agentic automation, is his pitch; they are trying to automate as much of an end‑to‑end process as possible, while managing risk, governance, and time to value.

His comments on human oversight are particularly notable for their restraint. Instead of treating autonomy as an inevitable endpoint, Gupta describes it as conditional – dependent on risk tolerance, accuracy, and context. That distinction matters for buyers assessing where automation can realistically deliver unattended ROI and where oversight will remain permanent.

What remains challenging is communication. Agentic language may open conversations, but the substance of UiPath’s position sits in hybrid automation, orchestration, and observability – concepts that require explanation rather than slogans. That tension is likely to shape how UiPath presents itself next. I’ll be returning to that question in a separate follow‑up conversation with UiPath’s CMO, Michael Atalla, who has been focused on translating customer readiness and trust into how the company talks about agentic automation.