Skip to content
Hendl
PLATFORM  /  DIGITAL WORKERS/ the always-on workforce

The workforce that never sleeps.

Hendl's agents are autonomous workers that pick up cases the moment something happens — gathering documents, checking cover, setting reserves and moving each case forward, around the clock. You set the playbook and keep the final call.

2.4M
Agent actions / week
across a mid-size book
71%
Of steps run without a human
the rest go to approval
<2 min
From notice to first action
median, all hours
24/7
Always-on coverage
overnight and weekends

Not a bot. Not a script. An operator.

RPA breaks when the form changes and a chatbot waits to be asked. An Hendl agent understands the case, decides what to do next, and does it — within the limits you set.

Autonomous

Agents pick up work on their own — no one launches them. They run continuously, overnight and at weekends.

Contextual

Every agent works from the full live case record — documents, history, policy and prior decisions — not one isolated task.

Accountable

Every step is logged to source, every action is permissioned, and a human approves anything consequential.

ANATOMY OF A RUN

Watch an agent think, act, and hand off.

An agent plans the work, uses tools to do it, checks itself against the record — then stops at the gate you set and asks a human to approve.

Triage agent · Run #70Executing
New claim submitted
Motor · CLM-4821
Triggered
Extracted incident documents
3 files · scope parsed
Done
Confirmed cover at date of loss
Cited to policy admin
Done
Set opening reserve
£12,400 · from similar history
Done
Checked the SLA clock
Acknowledgement due in 4h
Done
Recommended route: fast-track
Awaiting handler approval
Gate
THE AGENT'S PLAN
Read the notice and open the claim
Gather and parse incident documents
Confirm cover at the date of loss
Set an opening reserve
Recommend the next step for a handler
AWAITING YOUR APPROVAL
Route CLM-4821 to fast-track
Low value, cover confirmed, no fraud signals.
ApproveAdjust

A team of agents for every step.

Start from a template, or build your own.
Intake agent

Opens a claim the moment a notice arrives.

On new notice
Triage agent

Scores severity and routes the claim.

On claim open
Document agent

Chases, collects and reads missing documents.

On missing docs
Coverage agent

Confirms cover at date of loss, with citations.

On triage
Reserve agent

Recommends and updates reserves from the facts.

On assessment
SLA agent

Watches every deadline and escalates before breach.

Always on
Fraud agent

Screens for anomalies before any payment.

Before payment
§
Settlement agent

Drafts the settlement and prepares payment.

On decision
ALWAYS-ON OPERATIONS

See every agent at work, in one place.

Agents run continuously across your whole book. The operations view shows what's running, what's waiting, and what needs a human — so nothing slips.

Agents12 running · 3 need attention~ 40 actions / min
Triage agentCLM-4830 · PropertyRunning
Document agentCLM-4821 · MotorAwaiting documents
Coverage agentCLM-4815 · LiabilityNeeds approval
Reserve agentCLM-4808 · MotorRunning
Settlement agentCLM-4802 · TravelDone
SLA agentPortfolio watchRunning
+ 6 more agents running

Compose your own agents. No code.

Build an agent the way you'd brief a colleague — what starts it, what it can use, what it must check with you, and the playbook it follows.

Triggers

A new notice, a stage change, an SLA risk, a schedule — or another agent.

Tools

Search, read documents, draft, email, set reserves, request payment.

Guardrails

Permissions, autonomy level and approval gates, set per action.

Playbook

The steps and conditions the agent follows on every case.

Autonomy you can dial in.

Give agents room where the risk is low and keep them on a tight leash where it isn't — and change your mind whenever you like.

Approval gates

Nothing leaves the building without a human — payments, settlements, external messages.

Full provenance

Every action, tool call and source is logged and replayable.

Autonomy levels

Run fully autonomous, suggest-only, or step-by-step — per agent.

Tool permissions

Grant each agent only the systems and actions it needs.

WHAT THIS LOOKS LIKE

Agents pick up new claims overnight and have them triaged before the team logs in. People arrive to decisions, not a backlog.

Questions teams ask about agents.

What happens if an agent gets something wrong?

Agents recommend; a human approves anything consequential. Every action is logged and reversible, and you set where an agent must stop and ask.

Do agents run without us launching them?

Yes. Agents are triggered by events — a new notice, a stage change, an SLA risk — and run continuously in the background.

How much can we trust them to do alone?

You decide per agent: fully autonomous for low-risk steps, suggest-only for the rest. Autonomy can widen as you build confidence.

How do we build our own agents?

Compose them from triggers, tools, guardrails and a playbook — no code. Start from a template and adapt it to your process.

Which model runs the agents?

Hendl runs on frontier models with claims-specific scaffolding. Your data stays in your tenant and is never used to train shared models.

Can agents act in our other systems?

Yes — through governed tools they read and write to policy admin, document stores, email and payment rails, within their permissions.

Put an agent on your busiest claim type. See what it clears in a week.

© 2026 Hendl Labs Ltd. All rights reserved.
Services AgreementPrivacy PolicyTrust Center