Process engine

Run document-heavy processes with AI agents and people

The Vertesia process engine runs business work that mixes clear rules, AI agents, and human review. The engine decides what happens next. Agents do the reading and judging. Every step is logged, and work picks up again after a restart.

  • Work survives restarts and long waits
  • Agents only change what each step allows
  • Every step is logged
Why it matters

Rules alone are too rigid. Agents alone are too loose.

Workflow tools force fixed code, and chat agents choose their own route. Flip the switch to see what changes when the engine owns the route and agents do the thinking.

Documents

Treated as attachmentsRead, understood, and acted on at every step

Routing

Hard-coded rules or an AI's guessClear conditions the engine checks

Agent actions

Can change anything they reachCan only change the fields each step allows

People

Approvals lost in emailHuman tasks built into the process

Long waits

Timeouts and lost progressWaits for days, then picks up where it left off

How it works

Follow a contract through the process engine

Describe the process, or start from a diagram

Tell the Studio Assistant what you need, share a screenshot or whiteboard photo, or share a BPMN diagram. The assistant drafts the process and checks it before it can run.

Studio Assistant
Build a contract review process. Pull out the parties and the contract value, send anything over $50,000 to legal, then file approved contracts.
Draft process
Inside the engine

Built so AI can do real work inside your rules

Traditional process engines were built for two kinds of actors: code that runs fixed steps and people who make judgment calls. Vertesia adds a third, AI agents, which can reason like people and run like code, but only within the limits the engine sets.

Typed context and declared writes

Every step in a process declares exactly what information it can add or change. The engine enforces this strictly, so no agent can act outside what the process permits.

  • Each process carries typed data that its steps build up.
  • Agents can only fill in the fields their step declares.
  • Anything outside those fields is rejected.
Step: Extract terms
partiesAllowed
total_valueAllowed
legal_decisionBlocked
Process modes

Choose how much AI steers the process

Run a process on fixed rules, add an AI supervisor that can step in, or hand open-ended work to an agent.

Programmatic

Rules drive every step

The engine follows your process step by step, with no AI steering the route. This is the default, and the right fit for most work.

Supervised

An AI supervisor can step in

An AI supervisor watches the process and can retry, skip, or reroute a step. The engine checks each command before it applies it.

Agents

Or let an agent do the job

For open-ended work, an AI agent can work on its own, outside a process, with its tools and your permissions.

Use cases

Document-heavy processes, automated

Instead of people reading every document, routing it, and filling out forms, AI agents handle the work. People stay in the loop for approvals, exceptions, and judgment calls.

Every step can use your content. Process steps can search, read, and update documents stored in Vertesia, under the same access rules.

Contract reviewPull out key terms, flag changes from your standards, and route to the right reviewer
Invoices and purchase ordersMatch invoices to orders, find mismatches, and approve or escalate
Regulatory submissionsCheck completeness and rules before anything goes out
Customer correspondenceSort requests and route cases with a summary ready
Built-in security

Security built into every process

Agents work within your permissions, and their work is logged so you can audit it. Your content stays under your control and is never used to train public AI models.

SOC 2 Type IISecurity checked by outside auditors
HIPAASupports HIPAA compliance
GDPR and CCPABuilt for privacy rules on personal data
Full audit trailAgent steps, decisions, and handoffs are recorded
FAQ

Questions about the process engine

What is agentic workflow automation?

Agentic workflow automation means AI agents do the steps of a business process for you. People step in only when needed. Unlike rule-based tools, agents can read documents, decide based on what they find, and handle cases that don't fit a script.

How is a process different from an AI agent?

An agent works toward a goal and chooses its own steps. A process has named steps, clear rules for what happens next, and pauses for human review. Inside a process, agents do open-ended work, like reading a document, while the engine decides the route.

Can we import our existing process diagrams?

Yes. Share your BPMN diagrams with the Studio Assistant, and it turns them into a Vertesia process. Choices, parallel paths, and loops map to process steps. You review the draft and confirm before it is published.

What happens if a step fails or a server restarts?

Processes are durable. Each step is saved as it finishes, so a restart, a redeploy, or a long wait doesn't lose progress. A failed step can be retried, and the history shows exactly what happened.

Who answers human tasks?

Each task goes to a specific person or to a team, such as legal or finance. Team members see the task in their inbox, can open the whole process for context, and answer it there.

Can different steps use different AI models?

Yes. Some tasks work best with a different model. You can pick a model for each step in a process, to balance quality and cost.

Get started

Automate the work. Keep control.

See how Vertesia handles your document work with AI agents, clear rules, and human review.