UiPath

What it is

UiPath is the robotic process automation (RPA) incumbent that has spent the last eighteen months rebuilding itself as an orchestration layer for AI agents, software robots and people. The robots are still the core: they drive the screens of systems that have no usable API — mainframe green screens, Citrix-published ERPs, insurance and banking back-office tools — and they are why so many large ops teams already hold a UiPath contract. What is new is everything that sits above them.

It is a public company (NYSE: PATH). In the quarter to 31 July 2026 it reported revenue of $410 million, up 13% year over year, ARR of $1.938 billion, up 12%, and a dollar-based net retention rate of 109%. Net new ARR in the quarter was $37 million. Those are the numbers of a large, profitable, slow-growing installed base — which is the honest frame for everything below.

What changed for the agent era

Three pieces define the current product:

• Maestro is the orchestrator. It is built on Temporal’s durable-execution engine, so a long-running process keeps its state at every step and can be paused, resumed and audited end to end. It coordinates UiPath agents, robots, third-party agents and human approval steps in one process model. By its Q2 fiscal 2027 results UiPath had added Maestro Case for exception-heavy processes and, on 19 August 2026, Maestro Flow, a developer canvas that takes automations written by coding agents into governed production.
• Agent Builder builds the agents themselves, against UiPath-hosted models or a model you manage. UiPath Orchestrator can expose robots, agents and processes as MCP servers, so an external agent can call a governed UiPath process as a tool.
• UiPath for Coding Agents, announced 12 May 2026, lets Claude Code and OpenAI Codex author and run automations inside the same governance layer, with Cursor, GitHub Copilot and Gemini supported as Maestro Flow authoring tools.

Two moves point the same way. UiPath acquired WorkFusion on 6 February 2026 for its financial-crime compliance agents (AML, KYC, sanctions screening), and on 5 May 2026 it shipped most of its agentic features on Automation Suite, the self-hosted edition, so regulated and public-sector buyers can run agents on their own infrastructure with self-hosted LLMs. Gartner placed UiPath as a Leader in its Magic Quadrant for Business Orchestration and Automation Technologies, published 14 September 2026.

A use case worth buying it for

A claims or order-exception queue where half the systems have no API. An agent reads the inbound email and attachments, extracts the fields with document understanding, and decides the route; an unattended robot keys the result into the legacy policy or ERP screen; anything above a threshold goes to a human approver in Maestro; the whole run lands in one audit trail. n8n or Workato can do every step except the screen, and the screen is the step you cannot skip. That is UiPath’s defensible ground: agent judgment plus UI-level execution on systems nobody is going to modernise this year, under one owner.

Pricing

The published tiers are Community (free), Basic (from $25 per month, Europe-hosted, limited users and robots, no Maestro), and Standard and Enterprise, both quote-only. Standard is where agents, document extraction and unlimited users and robots start; Enterprise adds self-healing UI automation, process simulation and customer-managed keys.

Actual contracts are the better anchor. Vendr’s data through February 2026 shows a median of $45,338 per year across 42 purchases, ranging from $22,440 to $223,821. Its guide puts unattended robots at roughly $8,000–$10,000 per robot per year at list, attended robots at $3,500–$5,000 per user, and volume or multi-year commitments at 20–35% below list.

Agents meter separately. Under Flex licensing a production agent run consumes Agent Units per LLM call: 1 unit per call on standard models, 2 on premium models, 0.8 on basic models, charged in 64k-input-token increments — a call with 100k tokens of context bills twice. Bring your own model and it drops to 1 unit per run regardless of call count. On the credit-based commercial plan the same logic applies in Platform Units.

Best for

Ops and automation leaders at enterprises that already run UiPath robots, or that must automate across legacy systems with no usable API, and who want agents added under the same governance, audit and change control rather than as a second platform. It is the default pick if your automation Centre of Excellence already exists and is staffed on UiPath.

Alternatives, and when to pick them instead

• Microsoft Power Automate — bundled into the Microsoft estate most enterprises already license, with Premium at $15 per user per month, unattended bots at $150 per bot per month, and hosted bots at $215. Pick it when the estate is Microsoft 365 and Dynamics, the automations are departmental, and the robot count is small enough that per-bot pricing stays cheap.
• Automation Anywhere — the other long-running RPA Leader, named a Leader in Gartner’s 2026 RPA Magic Quadrant for the eighth year, with the Aisera acquisition (November 2025) adding IT, HR and support agents. Pick it when you are negotiating a competitive RPA renewal and its cloud-native control room fits your deployment model better.
• n8n — the fastest-growing entrant in automation, with a $5.2B valuation in May 2026. Pick it when every system you touch has an API, you have engineers, and you want agent workflows without a five-figure floor.
• Workato — pick it when the problem is integration between systems of record, not screen-level automation.

Watch-outs

• Robots and agents meter on different units, so the agent bill is invisible in a robot-based quote. Guard: before signing, take your three highest-volume agent processes, count LLM calls per run including retries and any prompt that crosses 64k input tokens, and price them on the model tier you will actually use. Negotiate the Agent Unit or Platform Unit pool against that number, not the vendor’s pilot estimate.
• Screen-level robots break when the target application changes. Guard: put every production robot under UiPath Test Cloud or an equivalent regression suite, and write the vendor’s UI-release calendar into the Centre of Excellence runbook. If a process has an API, move it to the API rather than paying twice for robot upkeep.
• Standard and Enterprise are quote-only, and the feature line between them carries real cost. Guard: get written confirmation of which edition includes Maestro, self-healing, customer-managed keys and the number of environments you need before the automations are built, not at renewal.