UiPath launched the Agentic Automation Builder Professional on 22 September 2026, and it sits above the Agentic Automation Associate we reviewed earlier. The exam description is version 1.0, dated September 2026. The basics: $150, 120 minutes, multiple choice plus drag and drop and sequencing questions, 70% to pass, three years of validity, delivered at a test centre or online proctored. UiPath's minimally qualified candidate has six months or more building automations on the platform and is familiar with LLMs, prompting and agent behaviour.
Search interest for "uipath certification" is around 480 a month in the US, and this is the exam those searches will increasingly land on. The syllabus has seven domains. The first two are foundations and platform: the difference between automation and agentic orchestration, robots against agents, reading process, solution and agent design documents, choosing a Studio Web project type, and licensing at an awareness level. Domain three is classic deterministic automation, including variables, control flow, DataTables, modern UI automation, ScreenPlay's LLM-based UI navigation, Integration Service connectors, API workflows, Action Center for human in the loop, and debugging and exception handling. Domain four is Maestro: BPMN orchestration for long-running processes, state and callbacks, chaining agents, case management with SLAs and escalation, plus a quick pass over Document Understanding, Communications Mining and the AI Trust Layer. The last three domains are where it gets interesting.
UiPath for Coding Agents covers which third-party coding agents are supported, installing and authenticating the uip CLI, installing skills, what AGENTS.md is for, reviewing AI-generated workflows against Workflow Analyzer rules, and governing coding-agent use in enterprise repos. The agents domain covers grounding, memory, tools, MCP, agents as tools, and building, debugging and deploying Python agents with LangChain and LangSmith. Then the evaluations and governance domain asks you to build evaluation sets, assert on agent output fields, validate JSON against a release gate, read Agent Score and enforce a minimum score through Automation Ops policies, configure guardrails and entitlements, and handle custom or on-prem LLMs through the AI Trust Layer. I like this exam more than I expected to. Most vendor agent certifications I have looked at stop at "write a good prompt and attach a tool".
This one treats agents as software that has to be tested, versioned and promoted through controlled releases, which is the part teams actually get wrong. The coding-agent material is also ahead of the market; it accepts that developers are already generating automations with external coding agents and asks how you govern that. The downsides are what you would expect from a first version. UiPath has not published domain weightings, so there is no way to know whether you will spend half the exam on DataTables and selectors. The scope is huge, and with roughly two hours to cover all of it, a fair amount will be identifying the right feature rather than solving a problem.
There is also a layer of product marketing in the licensing and Autopilot topics. And because the exam is new, third-party prep is either missing or, worse, braindump sites claiming to have the questions already. UiPath's own exam description warns about those, and I would take the warning seriously. Prepare with the recommended UiPath Academy learning plans, which are free, and get real time in Maestro and Agent Builder. If you have Python, build at least one coded agent with LangChain and run it through an evaluation set, because that is the skill in this syllabus most worth having whether or not you sit the exam.