On 7 October 2026 UiPath announced that it had been named a Leader in The Forrester Wave™: Adaptive Process Orchestration Software, Q3 2026. Forrester evaluated thirteen vendors whose platforms run complex processes across AI agents, robots, people and systems. UiPath was assessed on UiPath Platform and UiPath Maestro. The evaluation is framed around trust: Forrester looked at whether each platform offers the architecture and functionality to apply AI to complex processes in a way users can rely on.
In brief:
1/ Forrester evaluated 13 adaptive process orchestration vendors, and UiPath is among the Leaders,
2/ the strengths cited for UiPath are its context layer, orchestration testing, process observability and governance,
3/ Forrester accepted vendor material until 17 July 2026, and Maestro Flow and Maestro Automate reached general availability in September, after that date,
4/ whether an agent can be trusted in a given process is settled by how that process is designed - below are five decisions we make on every UiPath Maestro deployment.
Adaptive Process Orchestration Software as Forrester defines it
The category covers platforms that run a process end to end and hand individual steps to AI agents, robots, people or systems. Forrester opens with the buyer’s problem. Many organisations doubt that AI will produce accurate results reliably and repeatedly, and the analyst names this as one of the main reasons they hold back from autonomous operations with agentic AI. Some of that doubt is inherent in how generative models work, but the right architecture reduces unreliable output.
Vendors were scored on current offering, strategy and customer feedback. To qualify, they needed at least USD 70 million in annual revenue from the product category and a presence in at least three regions. Forrester accepted vendor material until 17 July 2026 and did not consider anything submitted after that date.
How Forrester assessed UiPath
The passages UiPath quotes in its announcement concern the design environment and the control layers. Forrester describes a design environment with superior collaboration and adaptivity and an advanced context layer, and says that comprehensive orchestration testing, proven process observability, governance and optimisation enable sophisticated process automation. It sees solid capability in autonomous design, agent library and discovery, hosting, process complexity and volume handling. Customers interviewed by Forrester stressed the breadth of integrated functionality and of out-of-the-box connectors to third-party systems.
UiPath makes the Forrester Wave™ graphic, showing all thirteen vendors, and the full report with every assessment available on its website: The Forrester Wave™: Adaptive Process Orchestration Software, Q3 2026 on uipath.com.
The evaluation cut-off and Maestro Flow and Maestro Automate
Maestro Flow became generally available on 17 September 2026 and Maestro Automate a day later, while building BPMN and Case processes with coding agents has been in Preview since September and October. We covered these releases in UiPath Maestro explained. Both GA releases came after 17 July, when Forrester stopped collecting material. The report does not say which pre-release capabilities vendors showed the analysts, so we read it as an assessment of the platforms as they stood in mid-2026, with anything any vendor shipped later sitting outside it.
Trust in an AI agent is a property of the process design
Forrester assesses whether a platform provides the mechanisms on which trust can be built. In a specific process, trust depends on how you use them. A context layer cannot help an agent that reads an outdated version of a procedure. Platform governance cannot tell you which action in your process is irreversible. Orchestration testing proves nothing if nobody has written down the expected results.
Defaults are no substitute for decisions either. Since 16 October 2026, new UiPath AI Trust Layer policies are created with input and output saving for the audit log switched off, while existing policies keep their current setting. Anyone who creates a policy without checking that switch gives up part of the evidence that the model was supervised.
AI-generated illustration.
On Maestro projects we settle five questions before an agent handles its first production case. Two of them have their own articles on our blog; here they sit side by side.
Autonomy limits live in the process. Any argument that defines the scope of an action or makes it irreversible - company code, amount, payee - is never left to the model. In a Maestro Flow agent node we set it as a fixed value or bind it to a process variable fed from a controlled source, such as the purchase order in SAP, which the agent cannot overwrite. The agent decides what calls for judgement, and the process decides how much is paid and to whom.
A human gate on low scores and on high-risk actions. The process pauses and routes the task to a person when the agent’s confidence falls below a set threshold, or when the action is high-risk whatever the confidence. A change to a supplier’s bank account always goes to a person. In our architecture reviews, a Maestro process with no escalation to a human at a high-risk stage is rated critical and does not go live. How these gates work across Maestro, Action Center and AI Trust Layer is covered in our article on HITL gates.
The process is tested before every release. Agent steps are non-deterministic, so each change to a prompt, model or tool is checked against a set of expected outcomes. In Maestro BPMN this is what evaluations are for, currently in Preview, using exact matching or a model acting as judge. Guardrails get their own tests: a guardrail wired up incorrectly can silently do nothing, with no error and no warning, so we try to get round every control, prompt injection included.
A trail that reconstructs the decision. Every model call governed by an AI Trust Layer policy appears in the audit log. The prompt and the model’s output are stored only if the policy has input/output saving enabled; otherwise the log shows them as “Blocked by Policy”. Whatever the agent established in earlier stages reaches the Action Center task as decision context, and the person’s decision flows back into the process as input to the next stage. Months later you can reconstruct the decision and the context the reviewer had in front of them, provided log retention covers the period. That is evidence of oversight an auditor may ask for, including in organisations in scope of NIS2 or DORA.
Autonomy is granted in stages. An agent starts as an assistant whose every decision a person approves. Its remit widens only where the rate of human corrections stays consistently low, and each widening is a formal, data-backed decision recorded in a register. A model change or a new document type moves the agent back a level until evaluations confirm the quality has held.
What an analyst ranking does not settle
Leader status says the platform’s trust mechanisms are mature compared with the thirteen platforms evaluated. It says nothing about whether your process is ready for an agent: whether the data the agent works from is current, whether decision roles are staffed, and whether the turnaround on human tasks holds at real volumes. A pilot on a single process answers that, with the map of gates drawn up before anything is configured.
For companies running SAP, invoice handling in purchase-to-pay is a good first candidate. A robot collects the data, an agent assesses discrepancies against the purchase order, ambiguous cases go to an approver, and the result is posted back to SAP S/4HANA. The path is known in advance, model and human decisions separate cleanly, and every stage leaves a trace.
UiPath Maestro at SNOK
SNOK is a UiPath Platinum partner and part of the Agentic Fast Track programme. Automation, security and SAP expertise sit in one team, so we design Maestro processes alongside the security review of the agents. Our services are described on the UiPath Maestro and AI Security pages, and the partnership on our UiPath page. We discussed an earlier analyst recognition of UiPath in orchestration in our piece on the Gartner Magic Quadrant for BOAT.
If you would like to find out which of your processes suits a first Maestro pilot, and where its gates should sit, get in touch.
Sources
- The Forrester Wave™: Adaptive Process Orchestration Software, Q3 2026, Forrester Research, Inc., September 2026, by Bernhard Schaffrik, report download on uipath.com (accessed 25.10.2026)
- UiPath, announcement of its Leader position in The Forrester Wave™: Adaptive Process Orchestration Software, Q3 2026, Business Wire, 07.10.2026, press release on ir.uipath.com (accessed 25.10.2026)
- UiPath, Automation Cloud release notes, October 2026, entry of 16.10.2026 on input/output saving in AI Trust Layer policies (accessed 25.10.2026)
Forrester does not endorse any company, product, brand, or service included in its research publications and does not advise any person to select the products or services of any company or brand based on the ratings included in such publications. Information is based on the best available resources. Opinions reflect judgment at the time and are subject to change. This report is part of a broader collection of Forrester resources, including interactive models, frameworks, tools, data, and access to analyst guidance. For more information, read about Forrester’s objectivity here.
UiPath™ is a trademark of UiPath, Inc. and UiPath SRL, and is used here with permission.