Main site

Strategy AI agent · Partner diligence

Partnership Development AI Agent

Interrogate the counterparty before the contract is executed — the ownership, the screening hits, the commission basis and the rationale — and hold the dated file for the approver who onboards.

Typical delivery
4–6 weeks
Deployment
Your stack
Named approver
Nothing signed
After launch
Agent Care

What this agent does

Constructs the file, not the clearance

In

  1. 01

    A partner is proposed, and the ECCP of September 2024 wants the business rationale written down.

  2. 02

    A percentage is agreed, and the ECCP asks whether it is commensurate with the services rendered.

Reason

  1. 03

    A red flag surfaces, and the Resource Guide of July 2020 makes deliberate ignorance knowledge.

  2. 04

    A commission is booked as something else, and 15 U.S.C. § 78m(b)(2)(A) bites with no bribe proved.

  3. 05

    An intermediary acts on your behalf, and Bribery Act s.7 leaves only an adequate procedures defence.

Decide

  1. 06

    A diligence step is designed, and Principle 4 of the March 2011 guidance governs third parties.

  2. 07

    A large organisation is in scope, and failure to prevent fraud is in force from 1 September 2025.

Out

  1. 08

    A contract nears execution, and after signature no honest record of the diligence can be made.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

Assembly and re-screening belong to the agent. The statute is unchanged and only enforcement policy moved; a named approver, never the agent, onboards the counterparty.

Example workflow

One counterparty, sources to onboarding

  1. 1Counterparty proposedSponsor request, questionnaire, registry extract or an introduction made by the customer
  2. 2Diligence context assembledThe counterparty, the people who own and control it, the territory and scope proposed, and the day the deal team wants it live
  3. 3Rationale evidence draftedThe counterparty, the sources under it, the flags and completeness
  4. 4Controls appliedScreening checks, ownership-resolution checks, benchmark checks and completeness confidence

No human action required

Stages 1 to 4 run unaided, and nobody is cleared at any of them — the agent builds the file, and the compliance read begins at the completeness gate.

5DecisionSplits at the completeness gate
Evidence sufficient

Reaches the compliance approver, who onboards.

Anything thin

Adds an ethics and compliance read first.

Compliance review

The file is held with its flags, its gaps and where each source came from.

Onboard · Append evidence · Send to compliance review

Onboarded — by the named compliance approver

  1. 6Partner and diligence records updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedSource completeness, flag disposition, compliance corrections and what review found

Corrections

Each compliance correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Clearing a counterparty for onboarding.
    • Deciding that a person is a foreign official.
    • Calling a red flag resolved or immaterial.
    • Judging a commission commensurate with services.

  • Automation boundary

    Agent acts unaided

    • ✓Build each counterparty file from the registry.
    • ✓Record the business rationale and the compensation basis, dated.
    • ✓Screen the counterparty, then re-screen on cadence.
    • ✓Escalate a red flag to a named human rather than scoring it away.

    No counterparty is cleared and none is onboarded except by a named person, on approval.

    • Declaring procedures adequate or reasonable.
    • Treating a payment as a permissible courtesy.
    • Setting whether the UK size thresholds are met.
  • Changes to the partner record or the flag register.

Example output

One counterparty flag, annotated

Our forced-labour import evidence agent assembles UFLPA evidence at the United States border; this is bribery and corruption risk inside a commercial relationship, before it is signed.

Diligence file · single counterparty

Illustrative example
Flag
Introduction at customer request
Recorded as
Raised at diligence, open and unresolved
Counterparty
Referral commission
Evidence of record
Sponsor memo, 3 August 2026
Confidence
Held open
Held for
The compliance approver, by name

As received

Drawn from the registry extract and the sponsor memo, and it reaches exactly as far as they do.

What the record holds

  • Registry extract
  • Sponsor rationale memo
  • Screening hits reviewed

Why no flag is closed here

Whether this flag may be closed is a judgement for a named human.

What the score decides

Below the threshold the file picks up a compliance read before the approver sees it.

Action

OnboardAppend evidenceSend to compliance review

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 counterparty

From the sponsor who wants it

03Evidence

Where the evidence is used

Nothing on this page is signed or certified under either regime: both ask instead for a dated record of what was found, what was concluded and why the relationship went ahead.

01Approved path

The record is the defence

Neither regime takes a filing: s.7 and the fraud offence are defences proved in court, on the record made while the relationship ran.

02Human review

What was checked, and not found

Enforcement since the pause is sparse, and that is the finding: DOJ charged Scoular on 17 July 2026 over customs brokers although the company did not know. No instrument extending or revoking EO 14209 was found, no corporate resolution of that year could be named, and no pause-era dismissal was verified.

04

Build an evidence trail

The counterparty, the rationale recorded for it and the person who approved stay on file.

Integrations

Typical integrations

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

  • Counterparty and registry data

    • Company registries
    • ownership sources
    • Legal identity and control chains

  • Screening and watchlists

    • Sanctions
    • PEP
    • adverse media
    • Screening hits and versions

  • Deal team and sponsors

    • CRM
    • sponsor questionnaires
    • Business rationale and proposed scope

Agent

Partner diligence evidence

  • Reads the counterparty
  • Builds the record
  • Holds for the approver

  • Contracts and payments

    • Contract repository
    • ERP
    • Commission terms, payees and audit rights

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

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

Agent controls

Six ratchets between the model and the approver

Six ratchets turned one way only, the last the tightest. What holds at the end is drawn in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to evidence assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and diligence rules; EO 14209 paused enforcement on 10 February 2025, and the DOJ guidelines of 9 June 2025 set the terms since.Track
L4TraceabilityRecord each fact, its source, the date it was retrieved, the human who read it and every read of the file.Record
L3Approver releaseRelease the file only to the named approver who onboards; the ECCP also asks whether audit rights exist and were used.Gate
L2Screening guardrailsTest the file against the Resource Guide red flags of July 2020, where conscious disregard, wilful blindness and deliberate ignorance all count as knowledge.Restrict
L1Confidence thresholdsRoute a thin file to a compliance read first; the March 2011 guidance is not prescriptive and is not a safe harbour.Require review
Model coreEvidence assembled — the counterparty, its sources, the flags and completeness
L1 – L2
Test whether a file may stand
L3
Hands the onboarding to a person by name
L4 – L5
Keep the counterparty and the rationale behind it
L6
Holds the onboarding open when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the diligence file that comes out.

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

Surface — the file an investigator reads

  1. E1Final-output evaluationDid the file record what the sources actually show?
  2. E2Step-level evaluationDid the agent read the right registry, the right list version and the live ownership chain?
  3. E3Tool evaluationDid it read and write the correct counterparty record and the correct flag?
  4. E4Confidence calibrationDo low-confidence files actually attract more compliance corrections?
  5. E5Slice evaluationHow does performance change across specific counterparty classes?
  6. E6Business outcomeHow many files needed a correction before the approver onboarded?

Floor — the record the company answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed at the stage where it first bites.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • LU-03

      Stale ownership read

      The ownership chain is not the one now on the register.

    Stage gathers

    The parties, the owners, the hits and the dates

  2. 02 · Reasoning

    2 modes

    • LU-04

      Rationale asserted, not shown

      A rationale is recorded with no source under it.

    • LU-06

      Pause read as repeal

      Lighter enforcement is worked as a safe harbour.

    Stage proposes

    The flags, their sources and completeness

  3. 03 · Tool / write

    2 modes

    • LU-02

      Thin file passed forward

      A file moves on without the compliance read.

    • LU-05

      Flag bound to wrong party

      A hit is filed against the wrong counterparty.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • LU-01

      Onboarded, evidence unrecorded

      The file shows an approval but not what was found.

    Stage returns

    The file a court or a regulator reads later

  5. 05 · Change / Version

    1 mode

    • LU-07

      Silent screening drift

      A list version moves while the file keeps the old clearance.

    Stage tracks

    Model, prompt, screening rules and file fields

  • Sev-1 · a partner onboarded on no evidence
  • Sev-2 · a red flag closed without reasons
  • Sev-3 · source degrades, file held open
See the complete Strategy failure-mode catalogue

Affected slices

Customer introductions absorb the corrections

A class-level rationale-coverage figure can read clean while customer-introduced intermediaries carry most of the rework. Nestack reports the correction rate by counterparty class, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Customer-introduced intermediaries10.6%3.7×Review
Sales agents and consultants7.6%2.6×Review
Customs and logistics brokers4.7%1.6×Watch
Resellers and distributors2.2%0.8×Normal

Bar: correction-rate lift vs. reseller and distributor baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What an unexplained intermediary costs

The loop shuts when the unexplained intermediary is a regression case. That suite is what the next counterparty cleared is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on customer-introduced intermediaries.

  2. 02

    Diagnose

    The consultant introduced by the customer, paid a percentage of the contract with no explicable role, is read back until one cause remains.

  3. 03

    Improve

    Any change goes out numbered, with the counterparties that caused it attached.

  4. 04

    Verify

    One counterparty case still failing is enough to hold the release back.

  5. 05

    Learn

    One case joins the suite, one line joins the diligence record.

  6. Learn → Detect

    The return edge. The next counterparty 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, diligence assembly, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Red-flag-escalation and approval-boundary work.Week 1
  2. 02Registry, screening and sponsor sources.Week 1, Week 2
  3. 03Rationale-to-source and flag-disposition mapping.Week 1, Week 2
  4. 04Counterparty evidence ingestion.Week 2
  5. 05Counterparty, source and flag binding.Week 2, Week 3
  6. 06Completeness scoring and review routing.Week 3
  7. 07Compliance approver workflow.Week 3, Week 4
  8. 08Screening and CRM integration.Week 5
  9. 09Red-flag and rationale cases.Week 4, Week 5
  10. 10Guardrails and approval controls.Week 4
  11. 11Counterparty-trail instrumentation.Week 5, Week 6
  12. 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 tierPilotOne class, one territoryProductionProduction diligence workflowAdvancedMultiple classes / territories
Introduced at Pilot
File assembly to your risk tiers✓✓✓
Compliance approver release✓✓✓
Intermediary-population baseline✓✓✓
Introduced at Production
Reporting by counterparty class—✓✓
Approver sign-off workflow in your systems—✓✓
Approved write-back—✓✓
Screening-source integration—✓✓
Introduced at Advanced
Multi-party ownership chains——✓
Cross-border evidence packs——✓
Large partner estates——✓
Multi-jurisdiction diligence controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, counterparty volume, 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.

  1. 01Your live counterparties and the sponsors behind them

    Counterparty mapping and source capture

    Week 1

  2. 02Representative registry, screening and contract records

    Source binding, flag logic and the diligence baseline

    Week 2

  3. 03Your standing onboarding procedure and the approvers it names

    Risk-tier mapping, flag binding and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Registry, screening and contract source assessment, then integration setup

    Week 2

  5. 05Counterparties you would not want examined

    Red-flag cases and failure-mode testing

    Week 4

  6. 06What no diligence file may guarantee

    Completeness scoring, review routing, guardrails and release controls

    Week 3

  7. 07A named approver who onboards the counterparty

    Handover to the compliance approver, 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

This is a schedule and not decoration, which is why two of the phases genuinely overlap in week five.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Diligence workflow discovery, risk-tier mapping and the automation boundary

W2Source integration and the counterparty evidence baseline

W3File assembly, flag logic and release controls

W4Evaluation suite, red-flag cases and failure-mode testing

W5Screening integration, pilot counterparties and targeted corrections

W6One partner year run under the compliance approver, then Agent Care handover

Reading the band
Each bar takes the weeks its own work is named for, and week five carries a pair because the work does.
At the end of W6
Once the diligence 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 · Strategy AI agent

Build a partner diligence agent that finishes before the contract is executed.

Nothing here is signed and nothing is filed. Show us your last onboarding and the record it left: the rationale in the words of the sponsor, the screening hits reviewed, the commission basis, and the name of the person who proceeded anyway.

Typical delivery
4–6 weeks
Deployment
Your live partners
Named approver
Nothing signed
After launch
Agent Care available

Nestack Agents · Partner diligence

AGT-CS-03 · Agent Care available after launch

Build a partner diligence agent

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

About you
Your counterparties

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