Automation

Automate with agents (and keep control)

Browser agents + visual workflows—grounded in Operational Knowledge.

Create browser-based agents from screen recordings or voice. Run them with MCP compatibility and integrated prompt engineering—then wrap them into reliable, auditable flows.

Browser-based agents, benchmarked among the best
MCP-compatible tool access
Integrated high-class prompt engineering
Transparent, controllable automation
Live run
From screen recordingOr voice inputAuto prompt engineeringMCP compatible
Workflow Canvas · browser agent run
Workflow canvas with browser-based agent steps

Agent thinking (simulated)

“I will open the claims portal, locate case CLM‑1042, verify policy coverage, then update the payout status and attach proof of submission.”

Voice input

“Update claim CLM‑1042 to paid and attach the confirmation screenshot.”

Open portal

Navigate to the right system and page.

Process automation

Learn, evaluate, and implement—without losing governance

Use our 2026 guides to understand process automation (and agentic automation), then move from knowledge to governed execution with approvals, exceptions, and evidence artifacts—guided by HEIDI.

Start here

Process automation platformThe operating model: gates, evidence, drift loops.Process automation guide (2026)Definitions, patterns, and implementation playbook.Process Designer vs UiPathCompare operating models—breadth vs governed proof.

Governed execution primitives

Approval gates, exception paths, evidence artifacts, version logs.

Agentic automation, safely

Narrow guardrails + mission oversight (Command Center).

Less blue, more readability

Neutral surfaces with on-brand accents for comfortable reading.

Visual flow builder

Transparent steps. Clear approvals. No black boxes.

Browser-based workflow canvas with MCP
MCP Compatible
LLM & API Tools

Control the complexity

Business-friendly by default. Technical when needed.

Start simple with clear steps and approvals. Dial up control with MCP tools, agent policies, and reusable blocks—without turning work into a black box.

SimpleBalancedTechnical

Simple mode

  • Clear steps & ownership
  • Approvals & exceptions
  • Human-in-the-loop controls

Technical mode

  • MCP tool access
  • Agent policies & guardrails
  • Reusable blocks & versioning

Build workflows visually

Create automations with a clear, visual builder. Make steps explicit so business and IT can agree on what happens next—before anything runs.

Standardize execution across teams

Reduce rework and manual follow‑ups

Make handoffs visible and reliable

Scale best practices, not tribal knowledge

Built on Operational Knowledge

Automations are grounded in your Operational Knowledge so they reflect how your business actually works—and so agents can execute with context instead of guessing.

Shared loop

Document → validate → automate → improve. One loop across teams.

MCP-compatible agents

Use MCP compatibility to connect tools and actions to your agents—so they can operate in the browser and across systems while staying controllable.

Tool access with clear contracts

Audit-friendly execution trail

Reusable tool sets across agents

Fits into workflows and approvals

Example (technical view)
tool: "browser.navigate"
input: { url: "https://portal.company/" }

tool: "browser.type"
input: { selector: "#caseId", text: "CLM-1042" }

Prompt engineering—built in

Turn a goal into reliable execution. Prompts, tool instructions, and guardrails are generated and refined so your teams don’t have to become prompt specialists.

Goal

Update claim CLM‑1042 to paid

Context

Operational Knowledge SOP + policy rules

Tools

Browser actions via MCP

Guardrails

Approvals, logging, and safe defaults

Regulated operations

Auditable automations (evidence by design)

In banking and insurance operations, automation must be safe: clear approvals, exception paths, timestamps, and evidence records. Treat every agent action as a process step that produces proof.

DORA controls mapping

Control → decision point → evidence trail.

Open

MiFID II record keeping

Make timestamps and records part of the workflow.

Open

GDPR evidence & lineage

Collect → use → share → store → delete with proof.

Open

Use cases

Automations that business and IT both trust

Start with Operational Knowledge. Add browser agents for execution. Wrap everything in clear workflows and approvals.

Claims processing

Faster cycle times with fewer mistakes

  • Agent updates portals
  • Approvals for exceptions
  • Evidence captured automatically

Customer onboarding

Consistent steps across teams and regions

  • Checklists from SOPs
  • Browser data entry
  • Audit trail for compliance

Procurement & approvals

Fewer follow-ups and clearer ownership

  • Route approvals
  • Validate documents
  • Write back to systems

Support operations

Less copy/paste, faster resolution

  • Retrieve knowledge
  • Draft responses
  • Update tickets in the browser

Compliance checks

Repeatable controls you can prove

  • Run checks consistently
  • Log decisions
  • Escalate when needed

Back-office reporting

Reliable data gathering without manual work

  • Collect from web apps
  • Normalize fields
  • Hand off for review

Start with one workflow. Scale to a fleet.

Capture how work is done, generate agents from voice or screen recordings, and operationalize them with visual workflows and controls.

Q&A

Frequently asked questions

Learn more about how Process Designer works and how it can help your organization.

How is this different from RPA?+

RPA (Robotic Process Automation) is brittle; if a button moves, the bot breaks. Our browser agents use AI vision and DOM understanding to adapt to UI changes. Plus, our automation is grounded in Operational Knowledge, so it understands the process, not just the clicks.

What makes it 'auditable'?+

Every step, decision, and tool execution is logged. You can see exactly why an agent took an action, which policy it followed, and who approved it. No black boxes.

Do I need to know how to code?+

No. You build workflows visually. For agentic steps, you describe the goal in plain English (or voice), and HEIDI handles the prompt engineering and tool selection.

Can it integrate with my internal APIs?+

Yes. Process Designer is MCP-compatible (Model Context Protocol). You can securely connect your internal APIs and databases as tools that agents can use under strict access controls.

How do you handle human approvals?+

Approvals are first-class citizens. You can drag-and-drop an Approval step into any workflow. The automation pauses, notifies the right person, and only proceeds when they sign off.