AI-03 · AI case study · Trade promotions & claims · FMCG distribution

Automating distributor claim settlement

Concept case study · 2026 · Based on my promotions and claims process work on a DMS rollout

ConceptA design based on my project experience, written during my 2026 sabbatical. It is not a deployed client solution.

Aditya BankarAditya BankarSenior Business Analyst / Product Owner · 3 min read
2 claim typesOn-invoice and off-invoice
Photo checksAI reviews display proof
Every lineChecked against scheme rules
Exceptions onlyReach a reviewer

The context

In FMCG, distributors fund trade promotions first and claim the money back from the brand. Until a claim is settled, the distributor is lending the brand money. On-invoice schemes leave a trail in the invoice. Off-invoice schemes, like display incentives, need proof such as photos.

On my last DMS rollout I mapped as-is and to-be processes for promotions and claims and configured six promotion types. This concept applies AI and automation to the settlement steps that stay manual in most set-ups.

The problem

Most of the settlement time goes on manual checking, not decisions:

  • Claims are compiled by hand from invoices and paper proof.
  • Display and visibility claims depend on photos that someone checks one by one.
  • Sales and finance check every line, routine or not.
  • Disputed lines hold up the whole claim, and the distributor's money with it.

What I did

  • Mapped the claim journey for on-invoice and off-invoice schemes from my project experience.
  • Designed the automated steps: claim generation, rule checks and AI photo checks.
  • Defined the validation rules and what counts as an exception.
  • Wrote user stories with Gherkin acceptance criteria.

The process, before and after

Each column is a team or system. Numbered badges point to where time is lost today, and how the design saves it.

Scroll sideways to see the whole diagram →

Claims process before: manual, sequential checksDistributorSales teamFinanceERPPromotionstartsGive schemeat the shopCompile claimwith documents1Check everyline by hand2Check againand approve3Raise creditnote4Weekslater
  1. 1Claims are built by hand after the sale, from invoices and paper proof.
  2. 2Sales checks every claim, routine or not. Disputed lines go back and forth.
  3. 3Finance repeats the checks before approving.
  4. 4The credit note is raised late, after every approval.
Start Step Decision End Each column is one team or system. Arrows show hand-offs.

What changed, step by step

StepBeforeAfter
Creating the claimCompiled by hand with documentsGenerated from invoice data
Proof for off-invoice schemesPhotos checked one by oneAI checks photos and flags doubtful ones
CheckingEvery line checked by sales, then financeRules check every line; people see only exceptions
Credit noteRaised late in the ERPCreated in SAP once approved

The approach

  1. 1
    Build claims from invoicesOn-invoice claims are generated from invoiced sales in the DMS, so the evidence is already in the system.
  2. 2
    Check photos with AIFor off-invoice schemes, AI checks each proof photo for the right display, shop and date, and flags doubtful ones.
  3. 3
    Validate by ruleRules check eligibility, budget, validity dates and duplicates on every line.
  4. 4
    Review by exceptionOnly lines that break a rule or fail the photo check go to a reviewer, with the reason shown.
  5. 5
    Connect to the ERPApproved claims create a credit note in SAP without re-keying.

Sample work: user story

Story · AI03-US-04

Flag doubtful display photos

As a claims reviewer, I want display photos to be checked automatically against the scheme, so that I only look at the doubtful ones.

Acceptance criteria

  • Given a display claim line has a photo, when the check runs, then the line is marked 'Photo passed' or 'Photo doubtful' with the reason.
  • Given a photo is dated outside the scheme period, when the check runs, then the line is marked 'Photo doubtful: wrong date'.
  • Given the same photo is used for two shops, when the check runs, then both lines are flagged as possible duplicates.
  • Given a line is marked 'Photo doubtful', when I open the claim, then I see the photo, the reason and the options to approve or reject.

Expected outcomes

  • Distributors get reimbursed faster, because routine claims need no manual checks.
  • Reviewers spend their time only on doubtful photos and rule breaks.
  • Fewer duplicate or invalid claims get paid.
  • Every decision has a recorded reason.

Design principles

  • Most claim delays start at promotion set-up, so fix that first.
  • Review by exception works only if finance and sales agree the rules before build.
  • AI flags, people decide.

Concept case study written during my 2026 sabbatical. It is not a deployed client solution. Examples are generic.