Claims & recovery
Agent: the claims agent (claims in ExecuteLoop), with the claims and
recovery evidence documents.
What it does
It puts a freight claim together from your own shipping records and drafts the paperwork for a person to approve. Its jobs:
- Shortage at delivery. Compares the manifest weight with the weighbridge reading and the counts received, and works out what is short and what it is worth.
- Over, short and damaged receipts. Compares the purchase order, the bill of lading and the receiving count, and prepares the OS&D report.
- Filing windows. Works out the notice and suit dates from the delivery date and the named regime, and orders open claims by the earliest deadline.
- Disputes and recovery. Prepares the notice of intent to claim, the statement of position and the subrogation demand.
- Recovery procedures. Ranks the recovery work with the most money at stake first, for example SLA breach refunds, accessorial and duplicate-charge disputes, and return patterns worth a closer look.
What it reads
Manifests, bills of lading, purchase orders, goods-received records, weighbridge
events, carrier tracking milestones, SLA records, invoices and the business rules
that Rules Recon found in your data. It can run read-only
queries against the connection a run was started with. Other agents can hand it a
sub-task with delegate_to_claims.
What it produces
- Evidence documents from the closed vocabulary:
OSD_REPORT,NOTICE_OF_INTENT_TO_CLAIM,STATEMENT_OF_POSITIONandSUBROGATION_DEMAND. Each figure cites the record it came from, and each deadline names the rule that set it. See Evidence documents. - Action cards on Actionable Twins (claims play), each with one proposed next step, such as “send this claim letter to the carrier”.
- A short recovery plan: two or three procedures, ranked by money recovered, each tied to the records behind it.
Human oversight
- A claims handler approves every claim letter, dispute and demand before it reaches a carrier. Approval seals the evidence document; rejection needs a reason.
- The agent never files a claim, contacts a carrier or withholds a payment on its own.
- A computed deadline is an input for the person deciding, not legal advice.
Model
The deployment’s general reasoning tier.
The figures in the evidence documents (amounts, units, deadlines) are fixed arithmetic over cited records. No model computes them. A model may phrase the narrative only after they are set.
Where it runs and data handling
On the deployment’s own model plane: Runink’s shared machines on a Trial, your own cloud on Cloud and your own servers on On-premises. No outside AI service is called on any plan. Your records reach the model as fenced, untrusted data. See Facts shared by every card.
Guardrails
- Never invents a figure. A missing rate or document is named as missing, and no amount is given in its place.
- Does not decide or score claims on the basis of a claimant’s name, demographics or location, and does not lowball settlements for particular areas.
- Does not fabricate weather events or disruptions to build a force majeure defence.
- Does not build defamatory profiles of carriers, and does not scrape partner portals or private systems without authorisation.
- Stays on claims and recovery work; warehouse routing and macro forecasting belong to other agents.
How to read its results
- It works from your records. Rates come from your contract file, and a rate that isn’t in it is reported as not stated.
- It doesn’t establish liability. A claim document assembles the file; the carrier, an adjuster or a court decides.
- A pattern that looks like return abuse is a lead for a person to check, not a finding of fraud.
- When a source could not be read, the document records the figure as absent with a
reason, such as
NO_BACKING_RECORDorNOT_STATED_ON_DOCUMENT.