Transportation AI agent · Broker security
Broker Financial Responsibility AI Agent
Watch the security itself — which instrument sits on file, which institution holds it, whether that trustee is still eligible — and hold the record for the principal who answers the notice.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Named answer
- Two sevens
- After launch
- Agent Care
What this agent does
Assembles the security record, never the sufficiency
In
- 01
A claim is made, and § 387.307(e)(1) opens a business-day window to address its validity.
- 02
A trustee changes status, and § 387.307(c) dropped loan and finance companies on 16 January 2026.
Reason
- 03
A trust is tested, and § 387.307(b) asks it to liquidate to cash inside seven calendar days.
- 04
A payment leaves the instrument, and § 387.307(e) puts the FMCSA notice on the provider, not you.
- 05
A suspension notice is served, and § 387.307(e) allows error, restoration or claims satisfied.
Decide
- 06
A provider is proposed for suspension, and § 387.307(g) gives it a calendar-day window to contest.
- 07
A default goes uncured, and § 387.307(f) calls that insolvency — a bankruptcy filing alone is not.
Out
- 08
A forwarder files security, and 49 CFR 387.403(c) applies the same broker rules to it.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent assembles and tracks each window. A named person at the brokerage answers the notice and restores the security; the surety or financial institution notifies FMCSA.
Example workflow
One notice, security to answer
- 1Notice or claim receivedFMCSA notices, trustee statements, surety correspondence and claim papers
- 2Security context assembledThe instrument filed, the institution holding it, the paragraph it answers to and when the window opened
- 3Security state draftedThe instrument, its assets, the claims and completeness
- 4Controls appliedEligibility checks, window checks, asset-quality checks and completeness confidence
No human action required
Stages 1 to 4 run unaided, and nothing is filed or contested at any of them — the agent is assembling, and the brokerage lane opens at the completeness gate.
Goes to the brokerage principal to answer.
Adds a compliance-desk read first.
Compliance review
The record is held with its claims, its windows and the instrument it rests on.
Answer · Append document · Send to compliance
Answered — by the brokerage principal, dated
- 6Security and registration records updatedOnly where write access and records policy allow it
- 7Outcome evaluatedWindow coverage, trustee eligibility, desk corrections and what review found
Corrections
Each compliance-desk correction is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Contesting the validity of a claim on the bond.
- Answering an FMCSA notice of proposed suspension.
- Certifying that the security on file is sufficient.
- Lodging a replacement instrument with FMCSA.
Automation boundary
Agent acts unaided
- ✓Hold the security instrument and its institution.
- ✓Time every claim against the response window that service opened.
- ✓Chase the trustee statement that never came, and log it.
- ✓Flag a trustee that § 387.307(c) will no longer treat as eligible.
Nothing is filed or contested except by a named person, inside the agreed boundaries.
- Judging whether the security on file is enough.
- Telling FMCSA a notification was sent in error.
- Setting the instrument a brokerage is filed under.
- Changes to security, surety or registration records.
Example output
One security record, annotated
Our carrier onboarding and fraud vetting agent pulls authority, insurance and registration facts on the carrier you are about to load; the registry that shows a carrier its standing shows yours, and this record is your own.
Security state · single instrument
Illustrative example- Instrument
- Trust fund, § 387.307(b)
- Recorded as
- Assets liquidate inside the test
- Institution type
- Insured bank
- Source of record
- Trustee statement, 3 August 2026
- Confidence
- Held unanswered
- Held for
- The brokerage principal, by name
As received
Taken from the trustee statement and the registration record — it reaches as far as those sources do.
What the record holds
- Instrument on file
- Trustee statement
- Notice, dated
Why no sufficiency call here
Whether the security suffices is a § 387.307 judgement, not a model output.
What the score decides
Below the configured threshold the record picks up a desk read before the principal sees it.
Action
AnswerAppend documentSend to complianceValue
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 claim
From the institution holding it
Where the record is used
Our freight audit and dispute agent works invoices against contract and tariff, and a claim on the bond is a different instrument on a different clock. Our rate negotiation agent stops at acceptance; the authority to broker at all is what this record protects.
Two sevens, measured apart
§ 387.307(b) sets a seven-calendar-day liquidation test on trust assets. The cure period in paragraph (e) runs in business days from written notice, and trade coverage has repeatedly merged the two.
What was checked, and not found
No reinstatement standard or timeline appears in § 387.307, in the rule overview, in the FAQ or in the January 2026 industry presentation — the only cure described is the one taken before suspension, which is the reason never to arrive there. No enforcement statistics or suspension counts were published, and no extension beyond 16 January 2026 was found: verified to January 2026.
Build an evidence trail
The instrument, the institution holding it and the person who answered stay on the record.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Surety and trust providers
- BMC-84 surety
- BMC-85 trust
- Trustee statements
FMCSA registration
- Registration and authority records
- Notices and public postings
Brokerage systems
- TMS
- brokerage ledger
- Settlement and claim records
Agent
Broker financial responsibility
- Reads the record
- Times each window
- Holds for the principal
Correspondence
- document store
- Claim papers and demands
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six rings between the model and the principal
Six rings, and the innermost is tightest of all. What clears them all is drawn in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a record may stand
- L3
- Puts the answer in a person's hands
- L4 – L5
- Keep the instrument and the institution behind it
- L6
- Holds the response undrafted when signals degrade
How Nestack evaluates it
Evaluate the whole assembly — not only the security record that comes out.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the record an auditor reads
- E1Final-output evaluationDid the record hold what each paragraph actually asks for?
- E2Step-level evaluationDid the agent read the right instrument, the right institution and the live filing?
- E3Tool evaluationDid it read and write the correct security record and the correct claim?
- E4Confidence calibrationDo low-confidence records actually attract more desk corrections?
- E5Slice evaluationHow does performance change across specific claim types?
- E6Business outcomeHow many records needed a correction before the principal answered?
Floor — the record the brokerage answers for
Failure modes
Where each failure originates in the agent
Seven failure modes, each fixed at the stage where it first surfaces.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- JF-03
Stale filing read
The instrument read is not the one now on file.
Stage gathers
The sureties, the trustees, the claims and the dates
02 · Reasoning
2 modes
- JF-04
Record asserted, not shown
A record is called sound without its statement.
- JF-06
Superseded date read as live
The extended January 2025 stagger is worked as law.
Stage proposes
The instrument, its assets and completeness
03 · Tool / write
2 modes
- JF-02
Thin record passed forward
A record moves on without the compliance read.
- JF-05
Claim bound to wrong instrument
A claim is filed against the wrong bond.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- JF-01
Answered, evidence unrecorded
The record shows an answer but not what supported it.
Stage returns
The record a principal answers and an auditor reads
05 · Change / Version
1 mode
- JF-07
Silent window regression
A configuration change moves the window, not the record.
Stage tracks
Model, prompt, window rules and security fields
- Sev-1 · a record sound on no statement
- Sev-2 · wrong instrument reaches the record
- Sev-3 · source degrades, record unanswered
Affected slices
Late answers absorb the corrections
A claim-level response-timeliness figure can read clean while claims answered late carry most of the corrections. Nestack reports the correction rate by claim type, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Claims answered late | 9.6% | 3.6× | Review | |
| Trusts on an ineligible trustee | 7.0% | 2.6× | Review | |
| Judgment-driven payments | 4.3% | 1.6× | Watch | |
| Bonds with no claim on record | 2.6% | 1.0× | Normal |
Bar: correction-rate lift vs. no-claim bond baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
What an ineligible trustee costs
The loop shuts when the ineligible trustee is a regression case. That suite is what the next filing checked is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Correction rate rises on claims answered late.
- 02
Diagnose
The letter that starts a clock nobody in the office recognised is read back until one cause remains.
- 03
Improve
Any change goes out numbered, with the claims that caused it attached.
- 04
Verify
One claim case still failing is enough to hold the release back.
- 05
Learn
One case joins the suite, one line joins the security record.
Learn → Detect
The return edge. The next record is measured against a suite one case longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, record assembly, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Security-record and automation-boundary discovery.Week 1
- 02Surety, trust and registration sources.Week 1, Week 2
- 03Instrument-to-paragraph and claim-window mapping.Week 1, Week 2
- 04Security document ingestion.Week 2
- 05Instrument, institution and claim binding.Week 2, Week 3
- 06Completeness scoring and review routing.Week 3
- 07Principal response workflow.Week 3, Week 4
- 08Registration and surety-system integration.Week 5
- 09Eligibility and claim cases.Week 4, Week 5
- 10Guardrails and response controls.Week 4
- 11Security-trail instrumentation.Week 5, Week 6
- 12Deployment, documentation 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 tier | PilotOne instrument, one security year | ProductionProduction security workflow | AdvancedMultiple instruments / entities |
|---|---|---|---|
| Introduced at Pilot | |||
| Record assembly to your paragraphs | ✓ | ✓ | ✓ |
| Brokerage-principal release | ✓ | ✓ | ✓ |
| Instrument-eligibility baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by claim | — | ✓ | ✓ |
| Response workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Registration-record integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-entity registrations | — | — | ✓ |
| Cross-paragraph security packs | — | — | ✓ |
| Heavy claim traffic | — | — | ✓ |
| Multi-clock response controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, claim traffic, approval controls 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.
You bring
We build with it
01Your instruments and the institutions holding them
→Filing mapping and claim capture
Week 102Representative surety, trust and notice records
→Source binding, window logic and the eligibility baseline
Week 203Your registration issued under part 365
→Instrument mapping, institution binding and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Surety, trust and notice-source assessment, then integration setup
Week 205Notices you would not want unanswered
→Claim cases and the evaluation run
Week 406What no security filing may guarantee
→Completeness scoring, review routing, guardrails and release controls
Week 307A named principal to answer the notice
→Principal response workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Width here is worked time and not layout space, and so two of the bands must share the fifth week.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Security-record discovery, instrument mapping and the automation boundary
W2Source integration and the instrument-eligibility baseline
W3Record assembly, window logic and release controls
W4Evaluation suite, claim cases and failure-mode testing
W5Registry integration, pilot notices and targeted corrections
W6One security year run under the brokerage principal, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its own work is named for. The fifth week carries two because the work does.
- At the end of W6
- Once the security record validates, Agent Care assumes the agent.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Transportation AI agent
Build a broker security agent around the notice your principal has to answer.
Show us the instrument on file and the institution behind it. A named person at your brokerage answers the notice and restores the security; the provider notifies FMCSA. Since 16 January 2026 a loan or finance company is no longer an eligible trustee.
- Typical delivery
- 4–6 weeks
- Deployment
- Your security filing
- Named answer
- Two sevens
- After launch
- Agent Care available
Nestack Agents · Broker security
AGT-TR-19 · Agent Care available after launch