Main site

Insurance AI agent · Submissions

Submission-Intake & Appetite-Triage AI Agent

Read the broker's submission and its attachments, clear it against what you already hold, and triage it against your written appetite — an underwriter decides whether to quote or decline.

Typical delivery
4–6 weeks
Deployment
Your stack
Quote or decline
Underwriter
After launch
Agent Care

What this agent does

Clears and triages, declines nothing

In

  1. 01

    Take the submission as it arrives — the broker's email, the ACORD forms, the schedule of values and the loss runs.

  2. 02

    Extract every field with the document and page it came from, and total the schedule against the values on it.

Reason

  1. 03

    Clear the account against open submissions, bound policies and the broker of record, and name every duplicate.

  2. 04

    Read the appetite guide, the binder or treaty scope and the licensing in force on the day the submission arrived.

  3. 05

    Take occupancy, class and total exposure from the schedule itself, not from the description in the covering email.

Decide

  1. 06

    Where the account clears, sits inside written appetite and is complete, it reaches the named underwriter.

  2. 07

    Outside appetite, over authority, unclear or conflicted — it reaches an underwriter with the reason named.

Out

  1. 08

    Hand over a triage view — exposure as extracted, the appetite version read, the gaps and the desk it routed to.

  2. 09

    Retain the documents read, the fields extracted, the appetite version, the routing and what happened next.

Product statement

The agent extracts, clears and triages against written appetite. Whether to quote or decline is the underwriter's — and nothing the agent produces is a declination, or may be sent to a broker as one.

Example workflow

One submission, end to end

  1. 1Submission receivedA broker's email with the ACORD application, the schedule of values, the loss runs and any slip or questionnaire attached
  2. 2Documents read and extractedNamed insured, entity, locations, occupancy, class, values, limits and loss history, each with the document and page behind it
  3. 3Cleared against the bookMatched to open submissions, bound policies, prior submissions and the broker of record, with duplicates and conflicts named
  4. 4Appetite and authority checkedThe appetite guide, binder or treaty scope, licensing and capacity in force on the date the submission arrived, with the version stamped

No human action required

Stages 1 to 4 run without a person in the loop — the reading, the clearance and the appetite check are finished before an underwriter opens the submission. Nothing has been answered.

5DecisionSplits on whether the account clears, sits inside written appetite and is complete
Clears, in appetite, complete

Reaches the named underwriter, triaged.

Outside appetite or unclear

Reaches an underwriter with the reason named.

Underwriter or team leader

Reads the submission itself, decides whether to quote, and writes anything the broker is told. A declination is theirs alone.

Quote · Decline · Reassign

Routed — handed back

  1. 6Routed to an underwriterPlaced on the named underwriter's desk with the triage view, the appetite version applied and the open items
  2. 7Outcome evaluatedExtraction accuracy, clearance precision, appetite-version correctness, underwriter agreement and turnaround by broker and class

Underwriter changes

Every re-route and every corrected field is counted against the triage.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Declining, refusing to quote or telling a broker no.
    • Quoting, rating, pricing or binding any risk.
    • Issuing anything a broker could file as a declination.
    • Treating out of appetite as an answer.

  • Automation boundary

    Agent acts unaided

    • ✓Read the submission and extract each field with its source.
    • ✓Clear the account against open submissions, policies and brokers.
    • ✓Apply the appetite guide version in force on the date it arrived.
    • ✓Route to the named underwriter and name every open item.

    Write actions run only inside the approval boundaries agreed during implementation. A declination, a quote and a bind are not among them, on any tier, in any configuration.

    • Reassigning or changing the broker of record.
    • Releasing a cleared account to another broker.
    • Changing the appetite guide, authority or routing rules.
  • Setting the order a broker's submissions are worked.

Example output

One broker submission, annotated

Everything the agent puts in the triage view is attached to the document and the page it came from.

Triage output · single submission

Illustrative example
Submission
Commercial property, new
Attachments
ACORD, SOV, loss runs
Locations
Fourteen, all read
Triage
In appetite, routed
Confidence
92%
Quote or decline
None — underwriter's

As received

The broker's email and its attachments as they arrived, and the appetite guide version in force on that date.

Read from

  • ACORD 125 and 140
  • Schedule of values
  • Five years of loss runs

Why it routed here

Class and total exposure come off the schedule, not the covering email.

What the score decides

How much of the extraction an underwriter checks back at source. Never an answer.

Action

QuoteDeclineReassign

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every submission

From the mailbox, portal or broker platform

03Appetite & authority

Triage on the appetite as written

The appetite guide, binder scope and authority in force on the day the submission arrived — read at triage time and stamped with the version applied.

01Approved path

Start the file already read

The documents are read, the schedule totalled and the account cleared before an underwriter opens it, so the desk starts sorted rather than raw.

02Human review

Send the doubt to a desk, not the broker

Anything outside written appetite, over authority or unclear goes to an underwriter with the reason named — never back to the broker as an answer.

04

Build an evidence trail

Retain the documents read, the fields extracted, the appetite version applied, the routing and the underwriter's own decision — on both paths.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Submission channels

    • Shared mailboxes
    • broker portals
    • Ivans
    • PPL and Whitespace

  • Documents & extraction

    • ACORD applications
    • schedules of values
    • Loss runs
    • slips and questionnaires

  • Policy admin & clearance

    • Guidewire
    • Duck Creek
    • Sapiens
    • Majesco
    • clearance and account tables

Agent

Submission intake & triage

  • Extracts and clears
  • Checks written appetite
  • Routes to a desk

  • Appetite & authority

    • Appetite guides
    • binder and treaty scope
    • Licensing and capacity
    • exposure data

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the mailbox and a desk

Each control wraps the one inside it. A submission clears every layer before it reaches a queue, and the decision to quote or decline sits outside all six.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeReturn triage to the intake desk if extraction or routing quality degrades.Roll back
L5Routing recordBroker, appetite version, routing, reason and waiting time all kept.Record
L4No-declination ruleNo refusal, no quote, no wording a broker could file as a declination.Withhold
L3Appetite versioningAppetite guide, binder scope and authority pinned to the submission date.Pin
L2Clearance gateNo account routes until it is cleared against submissions, policies and brokers.Clear
L1Extraction provenanceNo field enters the triage without the document and page behind it.Cite
Model coreTriage proposed — exposure as extracted, the clearance result, the appetite version read and the desk it routes to
L1 – L2
Decide whether an account may route
L3
Fixes which appetite version applied
L4
Keeps a declination out of the agent
L5 – L6
Keep the trail and pull automation back

How Nestack evaluates it

Evaluate the triage — and what the underwriter did with it.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the triage an underwriter opens

  1. E1Final-output evaluationDo the extracted values match the documents they came from?
  2. E2Step-level evaluationWas the appetite version the one in force when it arrived?
  3. E3Tool evaluationDid it clear against the right account, policy and broker?
  4. E4Underwriter agreementWould an underwriter have triaged and routed it the same way?
  5. E5Slice evaluationHow do accuracy and waiting time change by broker and class?
  6. E6Business outcomeHow much of the triage did the underwriter have to redo?

Floor — the account reaching a desk that can actually write it

Failure modes

Where each failure originates in the agent

Seven failure modes plotted against the five stages of the agent lifecycle.

Agent lifecycle

Direction of processing →

  1. 01 · Extraction

    2 modes

    • SA-01

      Schedule read short

      Locations dropped or a valuation date missed; exposure reads low.

    • SA-02

      Occupancy taken from the email

      Class derived from the broker's description, not the schedule.

    Stage extracts

    Fields from the ACORD forms, schedule and loss runs

  2. 02 · Clearance

    1 mode

    • SA-03

      Cleared onto the wrong account

      A near-identical name, so a duplicate or a conflict survives.

    Stage clears

    The account against policies, submissions and brokers

  3. 03 · Appetite check

    2 modes

    • SA-04

      Stale appetite version

      Triaged against a guide the carrier has already moved off.

    • SA-05

      Triaged out on a proxy

      A postcode or a name stands in for the class the guide names.

    Stage checks

    Written appetite, binder scope, authority and licensing

  4. 04 · Routing / handover

    1 mode

    • SA-06

      Reads as a declination

      Wording a broker can file as a refusal, from no underwriter.

    Stage routes

    The triage view an underwriter opens and works

  5. 05 · Drift / Version

    1 mode

    • SA-07

      The queue sorts by broker

      One broker's or one class's work waits longer, every week.

    Stage tracks

    Appetite versions, extraction quality and queue order

  • Sev-1 · a broker gets an answer no one gave
  • Sev-2 · the triage misstates the risk or queue
  • Sev-3 · a field is wrong, review catches it
See the complete Insurance failure-mode catalogue

Affected slices

The worst reading lands on the smallest brokers

A cohort that is read worse is asked more questions and waits longer — that is how a broker segment, and the insureds behind it, is deprioritised without anyone choosing it. Nestack reports performance by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Scanned and photographed schedules8.1%3.7×Review
Schedules over a hundred locations5.9%2.7×Review
Low-volume broker offices4.2%1.9×Watch
Clean renewal submissions1.5%0.7×Normal

Bar: extraction-error lift vs. clean-renewal baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The submissions that waited are evidence too

When a broker has to chase, the triage already failed — wrong desk, or a cohort that always sits behind. Both belong in the evaluation.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Extraction accuracy, routing agreement or waiting time moves in one cohort.

  2. 02

    Diagnose

    Traced to the document, the clearance match, the appetite version or the routing rule.

  3. 03

    Improve

    The extractor, match rule or appetite mapping is changed, re-approved by underwriting and version-linked.

  4. 04

    Verify

    Re-run against held-out submissions, including every one an underwriter re-routed.

  5. 05

    Learn

    The re-routed submission becomes a regression case, with the appetite version it was triaged on.

  6. Learn → Detect

    The return edge. Every cycle re-reads waiting time by broker and by class — a segment that reliably waits longer is something the triage did to it.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, channels and extraction, clearance and appetite, evaluation, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Workflow discovery and automation-boundary definition.Week 1
  2. 02Submission channel and mailbox inventory.Week 1, Week 2
  3. 03Appetite, binder scope and authority sourcing.Week 1, Week 2
  4. 04Extraction for ACORD, schedule and loss runs.Week 2, Week 3
  5. 05Schedule and loss-run reconciliation rules.Week 2, Week 3
  6. 06Clearance matching and broker conflicts.Week 3
  7. 07Appetite versioning by submission date.Week 3, Week 4
  8. 08Routing rules, authority limits and the no-declination constraint.Week 4
  9. 09Underwriter triage-view sign-off.Week 4
  10. 10Extraction, clearance and routing evaluation.Week 4, Week 5
  11. 11Turnaround and broker-segment slice testing.Week 5, Week 6
  12. 12Observability, deployment and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne line, one channelProductionProduction intake integrationAdvancedMulti-line / multi-market
Introduced at Pilot
Document extraction with field provenance✓✓✓
Schedule and loss-run totalling✓✓✓
Clearance against accounts and brokers✓✓✓
Written-appetite triage✓✓✓
Baseline evaluation✓✓✓
Introduced at Production
Appetite versioning by submission date—✓✓
Authority, binder and licensing checks—✓✓
Underwriter routing and triage view—✓✓
Observability and audit trail—✓✓
Introduced at Advanced
Turnaround and broker-segment reporting——✓
Multi-market and enterprise controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on the lines and markets in scope, the submission channels and document formats in play, policy-admin and clearance integrations, submission volume, appetite and authority complexity and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01The lines, classes and markets in scope

    Workflow discovery and automation-boundary definition

    Week 1

  2. 02Your written appetite guide, and how often it changes

    Appetite, binder scope and authority sourcing, versioned

    Week 1

  3. 03Access to the submission mailbox, portals and policy admin

    Channel, clearance and policy-admin integration

    Week 2

  4. 04Real submissions, including the scanned and the disordered

    Extraction for ACORD forms, schedules and loss runs

    Week 2

  5. 05Your clearance rules and what counts as a broker conflict

    Clearance matching against accounts, policies and brokers

    Week 3

  6. 06Your written rule on who may decline and who may quote

    Routing rules, authority limits and the no-declination constraint

    Week 4

  7. 07Named underwriters and a team leader

    Triage-view sign-off, routing agreement and turnaround slices

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Phases are drawn over the weeks they actually occupy. The overlap in week 5 is where live submissions first reach underwriters, and re-routes start counting.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Lines and markets in scope, and who may decline

W2Mailbox, portal, clearance and policy-admin integration

W3Extraction, schedule totalling and clearance matching

W4Appetite versioning, routing rules and the evaluation suite

W5First live triage, underwriter re-routes and targeted fixes

W6Production validation, turnaround slices and handover

Reading the band
Re-routes in week 5 are counted on live submissions, not a retrospective sample. Each bar covers its own weeks only.
At the end of W6
Underwriters are picking work off triaged desks, and every submission can name the appetite version it was triaged on.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Insurance AI agent

Sort the submissions before you answer any of them.

Show us how submissions reach you, the appetite guide you triage against and who may decline. We'll read a week of your real submissions beside your own triage, and write down the wording the agent may never send to a broker.

Typical delivery
4–6 weeks
Deployment
Your submission stack
Quote or decline
An underwriter, always
After launch
Agent Care available

Nestack Agents · Submission intake & appetite triage

AGT-INS-10 · Agent Care available after launch

Build a submission-triage agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your setup

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.