Multi agents

Parallel AI workers

Multiple Agent workers process your documents at the same time. Scale your review capacity without scaling your team.

Capabilities

One task, many agents, one answer.

6 subagents120 contracts
Agent 1 · contracts 1–20Done
Agent 2 · contracts 21–40Done
Agent 3 · contracts 41–60Running
Agent 4 · contracts 61–80Running

Parallel processing splits a large job across several agents that work at the same time on separate slices.

Estimate · 120 contracts
Single agent≈ 4 h 10 m
6 subagents≈ 42 min
Rate2.9 documents / min
StatusOn track

Faster throughput what took an afternoon takes the length of a coffee, with the same per-document care.

Agent 3
Reading Vendor MSA §14 · change of control
Running
Agent 4
Flagged change of control in Reseller.pdf
Flagged
Agent 5
Waiting on OCR for scan_017.pdf
Waiting
Agent 6
Writing the summary table
Running

Real-time progress watch each agent's step as it happens, and step in the moment something needs you.

Shared rubric
Definition of change of control
Same clause test across all agents
✓
Citation format
Document · page · clause
✓
Confidence threshold
≥ 0.8, otherwise flag for review
✓

Consistent quality every agent applies the same rubric and citation format, so the merged result reads as one.

Split this due diligence across a team of agents
Plan: I lead. Subagent A takes corporate records, B the litigation search, C material contracts.
I'll merge, de-duplicate and cite when they report back.

Delegate to subagents a lead agent plans, hands out the work, then merges and cites what comes back.

FAQ

Common questions

When does the system use more than one agent?

When a job splits cleanly. A hundred contracts to review, a bundle to chronologise, or a research question with several independent limbs. The work is divided into slices and worked on at the same time, rather than one document after another.

How much faster is it, really?

It turns a job whose length was the sum of its parts into one whose length is roughly its slowest part. A review that ran through an afternoon in sequence finishes in the time it takes to read one of its documents carefully.

Does splitting the work make it less careful?

No. Every agent applies the same rubric and the same citation format, and each document gets the same attention it would have got alone. The merged result reads as one piece of work rather than as several stitched together.

Can I watch what the agents are doing?

Yes. Each worker's current step is on screen as it happens, so you can see which document is being read and where things stand. If one is heading the wrong way, you can step in while it is still running rather than after the fact.

What if two agents reach different conclusions?

They are reconciled before the result reaches you, and where a genuine inconsistency exists across documents the output says so rather than silently picking one. A disagreement between two contracts is usually the finding, not a problem.

Do I have to set this up?

No. The agent decides when a job is worth splitting and how to slice it. You describe the outcome you want, as with any other task.

Does a parallel run cost more?

It uses your AI allowance in proportion to the work done, not to the number of workers. Running six agents for ten minutes is much the same as running one for an hour.

Crafted for both lawyer and firms. Start using today.

Get started