One stage at a time, in a fixed order, with your examiner signing.
Each lifecycle agent does the work of one stage and hands your examiner a package: the recommendation, the rationale, the cited evidence, the draft letter. Sign it or send it back. The agents arrive in a fixed order so each one builds on a file the previous one assembled, and the control plane arrives with the first.
The slowest part of the cycle is waiting. The Intake agent verifies identity and eligibility against the HRIS census, confirms last day worked and job demands with the employer by email, requests the physician statement from the provider, follows up by the cadence you set, assigns the segment, and starts the regulatory clock. No money, no determinations.
Mail Room and Claim File Intelligence. Adds the HRIS adapter, the plan repository, and the clock service.
Identity or eligibility mismatch, claimant unresponsive after the maximum follow-ups, ambiguous document returns.
Eligibility determined with evidence, diagnosis and restrictions captured, last day worked confirmed, job class known, authorization on file, segment assigned.
A file summary citing the evidence behind each exit criterion, and an action plan with every open request.
Where the leakage is. On an approved claim the agent pays on schedule, requests recertification by segment, evaluates every evidence delta, handles extension requests against the guideline, reaches out before the expected return-to-work date, asks the employer about modified duty, suspends on a missed recertification, and flags long-term disability candidates early.
Intake. Adds the benefit engine and the clock service as included add-ons, and the payment and correspondence adapters.
Extension beyond the guideline maximum (nurse, then examiner), any termination recommendation, overpayment, payee change, a call is warranted.
Extensions within the guideline, closure on confirmed return to work, suspension for missed recertification. Your strategy owner can move any of these to a person.
A referral packet with the clinical summary, duration against guideline, and one specific question.
By now your examiners have been signing agent recommendations for months. The Decisioning agent evaluates own-occupation disability against restrictions and job demands, checks exclusions, estimates duration against the guideline, and produces one of three things: an approval with a benefit period, a pend with a named request and the extension notice, or a denial recommendation with the full adverse-determination content. Approvals within the guideline can go straight through for whitelisted simple claims. Denials always wait for a signature.
Intake and Ongoing Management. Adds the straight-through gate and the strategy whitelist.
Every denial, every approval beyond the guideline optimal, every clinical referral trigger. Straight-through only when segment and diagnosis are whitelisted and eligibility was verified deterministically.
Specific reason and plan provision, discussion of any disagreement with the treating provider, internal rules relied on, right to records, appeal rights and the window, draft letter.
The dial: the whitelist, the duration tolerance, the referral triggers. Publish, replay last quarter, compare.
Usually added with Decisioning, or as the last step to the full platform.
Assesses long-term disability candidacy at the strategy's midpoint, recommends the transition to an examiner, tells the claimant, gathers the LTD claim form and a fresh physician statement, assembles the package, submits it, and lines up the last short-term payment against the LTD elimination period. No gap, no surprise.
Acknowledges the appeal, collects new evidence, obtains any new clinical review, shares new evidence and rationale with the claimant before the decision as the regulation requires, and drafts an analysis that presents both sides for a reviewer who was not the original decision maker. The reviewer decides. Always.
These are enforced in code at the workflow and at the write to your system of record. No configuration can relax them. The model cannot see them as something it could change.
Each of these is checked in code at the workflow and again at the write to your system of record, and every check is logged, pass or block. A breach is a fact in the audit store, not an argument. That is what makes them possible to warrant.
| Rule | Enforced | If it is ever breached |
|---|---|---|
| C1 | No adverse determination without an examiner's signature. No denial, termination, reduction, or rescission leaves Claimwell without a recorded human decision. | Re-adjudication of the affected claim at Claimwell's cost, plus a fixed penalty per occurrence. |
| C5 | No payment outside an approved benefit period. The payment activity validates the period; the adapter refuses anything else. | Any overpayment the agent caused is reimbursed by Claimwell, including recovery costs you choose not to pursue. |
| C8 | No regulatory clock missed on a claim the agents own. Decision deadlines, extensions, tolling, and appeal windows come from the clock service and are enforced by the runtime. | Claimwell funds the remediation: notices, re-work, and any penalty attributable to the miss. |
| C4 | No letter stating a decision before that decision is on record. Correspondence with a decision template requires an existing decision reference. | Claimwell pays for the correction and the re-issue, and for the claimant outreach to put it right. |
Contractual warranties on the process Claimwell owns. Claimwell carries technology errors and omissions coverage with the customer named as an additional insured, at limits set in the agreement. The warranties sit alongside the service levels, not inside them. Terms, caps, and penalty amounts are in the agreement. Behind them: every release is replayed against the full scenario library before it reaches your claims, and you see the results.
Extraction accuracy, match accuracy, and recommendation quality. Those are measured by audit sample, not proven per claim, so they live in the service levels with credits. Warranting a probability turns every disagreement into a dispute about the sample.
Constraints are enforced in two places that never consult a model. Strategy is a versioned document your claims leadership owns. Judgment is what's left. Each tier can only narrow the one below it.
Adverse determinations, appeals, payee changes, payment windows, decision letters, clocks, citations on every write, step budgets.
Segmentation, straight-through whitelist, cadences, clinical triggers, duration tolerance, which tunable points are human.
What the evidence supports, what to ask, how to summarize, what to recommend and why. Every step logged and replayable.
| Id | Point | Default |
|---|---|---|
| H1 | Denial, termination, reduction | examiner · locked |
| H2 | Appeal decision | independent reviewer · locked |
| H3 | Payee or banking change | examiner · locked |
| H4 | Straight-through approval | auto when gate passes |
| H5 | Approval beyond guideline optimal | examiner |
| H6 | Extension within guideline max | auto |
| H7 | Extension beyond guideline max | nurse, then examiner |
| H8 | Clinical referral | strategy-triggered |
| H9 | Claimant phone call | examiner, agent-prepped |
| H10 | Payment exception, overpayment | examiner |
| H11 | Unmatched or ambiguous mail | examiner |
| H12 | Identity or eligibility mismatch | examiner |
| H13 | Claimant unresponsive after max follow-ups | examiner |
| H14 | Closure on confirmed return to work | auto |
| H15 | Agent step or time budget exceeded | examiner · locked |
| H16 | LTD transition recommendation | examiner |
Locked points cannot be moved by strategy. Every other point can be set to a persona or to automatic by your strategy owner.
Days from first notice to approval, pend, or recommendation, and separately to final decision. Cut by segment, cause, channel, and strategy version.
Paid duration versus optimal and maximum for the diagnosis and job class, as a ratio and a distribution.
Counted by handoff point and persona. The dial your strategy owner turns, and the number that shows what it did.
All stages owned, the strategy layer on, the dashboard live. Not a different product, just every rung with the dial in your leadership's hands. See the core and runtime it all runs on.
The package, the letter with the regulatory content, the signature, the letter going out, the appeal clock starting. Twelve minutes.