Enterprise Computer Use Agent (CUA)

You click Approve, WalkMe clicks the rest.

Ask in plain words and approve the plan. Enterprise computer use agents then do the work in any app with tested code, and check every result.

The agent reality check

Your agents can’t reach most apps.

It looks great in the demo. Then you hit a supplier portal with no API, a homegrown tool, and the security review. Your people need an agent that works with or without an API.

What is a computer use agent?

AI can now use any app,
so the real question is who’s driving.

A computer use agent (CUA) uses an app’s screen the way a person does, even with no API. For company work, it has to act as one real user, with no more access than they have.

Generic CUA

AI drives.

An AI call and a screenshot on every click

Asks a model what to do next

Clicks or types

Repeats, and small errors add up

WalkMe Enterprise CUA

AI plans. Verified code drives.

One AI call to plan · then AI only if stuck

AI plans the task once, from actions you approved

The whole plan is checked before the app is touched

Verified code does each step, as the user

Each result is confirmed in the app before the next step

Why WalkMe’s Enterprise CUA

AI agents that work beside your people in any app, within the rules you set.

Predictable by design

Verified code runs each step, so results hold up and costs stay predictable.

Governed as the user

It does only what you approved, within each user’s access, and logs every step.

Beyond the API

SAP or not, API or not, it works through the same screens your people use.

How it works

Every step takes the most reliable route first.

If the app can hand over the data directly, it uses that. If not, it runs tested steps on screen. AI steps in only when nothing else works.

First
Data straight from the app

When the app allows it, no AI, no clicks

Second
Tested steps

Tested code clicks through the screen, no AI

Third
DeepUI recognition

Finds the same button after a page change, no AI

Last
AI, only if stuck

Handles one stuck step, capped per task

The enterprise standard

Six things every enterprise agent has to get right.

Hold every vendor to all six, including us.

Reliability

Run the same task twice and both runs follow verified, checked steps.

Governance

You approve what it can do up front and see every step it took.

Compliance

It works within what each user is already allowed to do.

Speed

Most steps run on tested code, so nobody waits on an AI model.

Cost

Before a task starts, you know the most it can spend on AI.

Maintainability

If apps change, it finds the same button or stops; fixes need your OK.

WHO IT’S FOR

Built for the people who own the work and the people who do it.

CIO / Chief AI Officer

Run agents in every app you use, even with no API, and know the max AI cost up front.

CISO / IT governance

No new identity. It runs as the user, does only what you approved, and logs every step.

Business ops / process owner

Agents run your process end to end, API or not, following only the steps you approved.

Business builder / automation CoE

Describe the task once, approve it, and it’s tested before it goes live.

Walkme + Sap

Coming to SAP,
built for every app you run.

WalkMe is an SAP company. Custom Enterprise Computer Use Agents are coming to SAP Joule Studio, where you’ll manage them next to Joule Agents and use them in apps outside SAP, with or without an API.

Next step

Help shape how your agents will work.

Get launch news first, plus a shot at the beta and our design partner program.

Faq

Questions,
answered

What is an Enterprise CUA?

An AI agent that operates an organization’s applications through their user interface, as a specific user, with behavior that is approved in advance, verified as it runs, and recorded end to end.

How is it different from a generic computer use agent?

A generic CUA has AI decide every click from screenshots. An Enterprise CUA has AI plan once, from actions you approved. Then verified code does the work as the user, and every step is checked and recorded.

Is Enterprise CUA available today?

Not yet. It’s in development, and we’re building it with early design partners now. Custom agents are coming to SAP Joule Studio. Request early access to hear first and get a shot at the beta.

Do I need SAP to use it?

No. It works in any enterprise web app, SAP or not, including supplier portals and government sites. Agents are built and governed in SAP Joule Studio.

Does it send our screens to an AI model?

Not on the normal path, and it’s your call. You decide, agent by agent, whether AI may see the screen at all. You can also bar AI rescue for sensitive actions like payments, payroll, or personal data.

Isn’t this just RPA?

No. RPA was predictable but brittle, and it took specialists to build and fix. Enterprise Computer Use Agents start from a plain-language task. DeepUI keeps them working when screens change, and a bounded rescue handles the surprises.

Why not just use APIs?

We do, first, wherever one exists. Everywhere else, the agent works through the same screens your people use, and the app’s own business rules still apply.

What does requesting early access commit me to?

Nothing. It’s free, and you can opt out anytime. If there’s a fit for the beta or the design partner program, we’ll reach out.

Sources:

MuleSoft, 2026 Connectivity Benchmark Report.
Gartner, June 2025.