Insurance automation that starts with how claims and underwriting run.
A claim or a submission changes hands many times before the core system records a decision. Fluency observes that work across claims and underwriting teams, ranks it by hours saved, and puts AI agents on the steps your team approves.
Request a Demo
One path through the work
Claims intake and triage
- Phone and email
- Customer portal
- Policy administration
- Document store
- Claims systemNotice keyed in by handSystem of record

A claim passes through more hands and systems than the claims system shows.
The claims system holds the reserve, the notes and the payment. The calls to a repairer, the supplier portal, the email to a broker and the second look at a document all happen outside it, and that is where a claim spends most of its life.
Handoffs
Each time a claim moves between intake, assessment, suppliers and payments, it waits. Fluency shows every handoff and how long the work sits at each one.
Rework
Missing documents and incomplete submissions send work back to an earlier step. Fluency counts how often each loop happens and what it costs in handling time.
Variants
The same claim type is handled along many different paths. Fluency maps each path with the share of work that takes it, so the common one and the costly one are both visible.
The claims and underwriting work Fluency finds.
Four workflows that carry most of an insurer’s manual effort. Each one leaves the core system several times before a decision is recorded.
Claims intake and triage
First notice arrives by phone, email, portal or broker, and someone keys it into the claims system and decides where it goes. Fluency shows how much of intake is re-keying, how claims are routed, and which ones come back for missing information.
Runs through
- Phone and email
- Customer portal
- Policy administration
- Document store
- Claims system (system of record)
Assessment and settlement
Assessing a claim means checking cover, appointing suppliers, chasing reports and approving payment. Fluency maps each variant of that path, shows where claims wait on a third party, and sizes the steps an agent could prepare for a handler to approve.
Runs through
- Claims system
- Policy administration
- Supplier portal
- Spreadsheet
- Payments (system of record)
Underwriting submissions
A submission arrives as an email with attachments, and an underwriter or assistant extracts the details, checks appetite and enters the risk for rating. Fluency shows how long a submission takes to reach a quote and how much of that time is data entry.
Runs through
- Broker email
- Submission inbox
- Spreadsheet
- Rating engine
- Policy administration (system of record)
Policy servicing and renewals
Endorsements, cancellations and renewals are high in volume and governed by rules. Fluency counts the requests of each type, shows the steps each one takes across systems, and ranks them by the hours a change would save.
Runs through
- Broker portal
- CRM
- Document generation
- Policy administration (system of record)
Ranked by hours and cost before anything is built.
Fluency observes how claims and underwriting teams do their work across every application, then ranks what it finds by the return on changing it.
- Observed where the work is done
- A desktop agent observes work during the hours you set, and Fluency groups it into workflows, processes and steps. No integration with your process systems is needed.
- Numbers you can check
- Annual value is the observed runs, multiplied by the time saved on each, multiplied by the labor rates you set. Every estimate links to the runs behind it.
- A method you can follow
- The ranking weighs volume, handling time and how hard the change is. Our guide on how to identify automation opportunities sets out the method.

Opportunity · Claims intake
Claim details are re-typed from the portal
Proposed: pre-fill the claim from the portal.
ProcessAI assist
- Annual value
- $48,200
- Capacity
- 1,240 h/yr
An estimate from observed runs, not a realised saving.
Explore opportunities by their potential outcome. Some can support more than one goal.
Estimates, not realised savings. 24 of 33 annual estimates included.
- Claim details are re-typed from the portalPre-fill the claim from the portalProcessAI assistEst. $48,200/yr1,240 h/yr
- Missing documents are requested by handRequest missing documents when the claim arrivesProcessAI agentEst. $31,590/yr810 h/yr
- Teams triage on different rulesPublish one triage guideProcessPolicy / process--
- Submission data is keyed into the rating enginePass submission data to ratingProcessIntegrationEst. $33,500/yr860 h/yr
- Declines are written one by oneSend declines that are outside appetiteProcessRules automationEst. $21,900/yr560 h/yr
Agents on the work your team approves.
For the opportunities you approve, Fluency generates the agent from the observed process and runs it in your systems.
- People keep the decisions
- Where a step needs a person’s judgment, the agent has a review gate. The run stops there, shows what it proposes and waits.
- Measured against the baseline
- The work was being observed before the change, so the result is compared with how it ran before.
- Then the next change
- Observation continues after an agent goes live, so the ranking stays current and the next project starts from current evidence.

Agent · Claims intake
- Read the claim notificationAction
- Check cover and extract detailsAI step
- Create the claimAction
- Approve reserves over $5,000Review gate
The run stops at the review gate, shows what it proposes, and waits for a person to approve or reject.
See a claims operations example.
In the demo, we walk through a claims workflow, show how opportunities are ranked, and demonstrate an AI agent with a human review gate.
Request a Demo“Fluency is foundational in how we understand where to focus, where we get the best impact, and where AI will actually create value.”
Nick Carnell
Chief Executive Officer, Johns Lyng Group

“Post implementation of Fluency, the impact was instantaneous. Clear visibility, increased productivity, and measurable improvements.”
Mark Saliba
Chief Operating Officer, BoardRoom Australia

Secure and private by design.
Fluency is independently audited and built with the controls enterprise teams expect: single sign-on and SCIM, the option of your own encryption key, personal data redacted before any AI model sees it, and observation hours you set.
More about securityQuestions claims and operations leaders ask.
- Where should an insurer deploy AI agents first?
- On high-volume work with clear rules that crosses several systems. In most insurers that is claims intake, document chasing during assessment, submission entry in underwriting and routine policy servicing. Fluency observes this work, ranks each workflow by the hours and cost a change would save, and shows the evidence behind each estimate before anything is built.
- How can an insurer find where claims handling time goes?
- Measure the work outside the claims system as well as inside it. Fluency observes claims work across every application handlers use, separates active handling time from waiting, and lists the waits by step. The result shows whether time is going to data entry, supplier follow-up, rework or approval queues.
- Does Fluency measure individual claims handlers?
- No. Fluency analyzes how work happens across workflows and processes, not individual performance. It reports on a process, such as how many paths a claim type takes and where it waits, so the conversation is about the work and how to change it.
- Can AI agents make claims or underwriting decisions?
- Your team decides which steps an agent runs. Each step can be automated, automated with a review, or left with the team. Where a step needs judgment, such as a coverage decision or a settlement approval, the agent has a review gate: it stops, shows what it proposes with the files behind it, and waits for a person to approve or reject.
- How does Fluency protect privacy?
- Personal data is redacted before any AI model analyzes the activity, and the unredacted original is destroyed. You choose the hours and days the agent may capture, and outside them it captures nothing. People can pause their own agent, and excluded apps and URLs are never captured. Fluency analyzes how work happens across workflows and processes, not individual performance.Security at Fluency
- How is Fluency deployed?
- Fluency runs as a desktop app of about 15MB for macOS and Windows, rolled out through the device management tool you already use, such as Intune, SCCM, Jamf or Kandji. No integration with your process systems is needed. Useful maps of the work typically appear after about a week of observation.
Useful AI starts with understanding the work.
More industries
- Financial ServicesMap lending, servicing and operations work, variant by variant.
- Legal & Professional ServicesSee fee-earner and support work without timesheets.
- BPO & Customer OperationsCapture every variant of customer-service work across sites and clients.
- Healthcare & Life SciencesObserve operational work, not patients.
- Private EquitySee how work runs inside each portfolio company, and where AI could move margin first.
- RetailMap the head office work behind every product, price and supplier.
- ManufacturingSee the office work between the customer, the plant and the supplier.