PT Era Roshd Nusantara

Case study examples

Agentic AI for procurement: problem, recommended solution, results

Six worked examples showing how ERN frames a procurement problem, designs the agentic AI solution, and transfers the capability so the team can run it. Each example follows the same discipline: the agent handles the busywork, people keep the judgement.

Menghadirkan Generasi Cerdas, Unggul dan Kompeten.

Book an Executive DiscussionIndustry 4.0 case studiesProcurement & Supply Chain

The shift

From chasing documents to deciding with evidence

Procurement and supply chain teams still spend most of their week on quotes, emails and spreadsheets. Agentic AI does not replace the buyer — it removes the retrieval, comparison and reconciliation work that hides the real decision.

Yesterday

Procurement teams chased quotes, emails and spreadsheets, and decisions waited on whoever finished consolidating first.

Today

AI agents handle the busywork continuously; buyers focus on judgement — trade-offs, relationships and final terms.

Six worked examples

What each engagement actually looks like

Every example states the problem as the client experienced it, the solution ERN recommends across advisory and training, and the results a comparable programme can be expected to produce.

DiscoveryERN scope: Advisory + Training

Supplier discovery beyond the known contact list

The problem

Sourcing was limited to vendors the team already knew, so every new requirement started from a short and often outdated list.

  • Manual outreach to known contacts only
  • New categories took weeks to build a credible vendor pool
  • No consistent qualification criteria between buyers

Recommended solution

An agentic discovery workflow that scans vetted vendor sources, screens them against qualification criteria and returns a documented longlist.

  • Qualification criteria formalised before automation
  • Agent scans and screens thousands of vendor records
  • Evidence attached to every candidate for audit

What stays human: Buyers pick the final shortlist based on fit, risk appetite and relationship considerations.

Results to expect

  • Longlist ready in minutes instead of weeks
  • Wider, more competitive vendor pool per category
  • Consistent qualification standard across the team
RFQ automationERN scope: Advisory + Training

RFQ generation without rewriting templates

The problem

Buyers spent hours per request rewriting templates, and specification errors surfaced only after suppliers replied.

  • Hours of template editing per request
  • Inconsistent specifications across similar purchases
  • No tracking of who responded and when

Recommended solution

A specification library plus an agent that drafts, distributes and tracks RFQs against the correct template and commercial terms.

  • Specification and clause library cleaned up first
  • Agent builds, sends and tracks RFQs automatically
  • Reminders and response status handled without chasing

What stays human: Category owners review the specification and approve before anything is sent.

Results to expect

  • RFQ preparation time cut to a review step
  • Fewer clarification rounds caused by spec errors
  • Complete, timestamped response trail per event
AnalysisERN scope: Advisory

Quote comparison on a single, honest scorecard

The problem

Bids arrived in different currencies, incoterms and lead times, so comparison spreadsheets were rebuilt by hand every event and rarely compared like for like.

  • Apples-to-oranges data across bidders
  • Manual normalisation prone to error
  • Award rationale hard to reconstruct later

Recommended solution

A normalisation model — currency, freight, duty, lead time, payment terms — feeding an agent-generated ranked scorecard with the assumptions shown.

  • Total-cost model agreed with finance and users
  • Agent normalises bids and ranks them on one scorecard
  • Assumptions and sensitivities exposed, not hidden

What stays human: The committee validates trade-offs and makes the final award decision.

Results to expect

  • Evaluation cycle shortened for every tender
  • Award rationale documented and auditable
  • Savings identified that flat price comparison missed
NegotiationERN scope: Advisory + Training

Negotiation support with benchmarks and risk flags

The problem

Contract review took days line by line, and negotiators entered discussions without a reliable internal benchmark.

  • Days of manual clause-by-clause review
  • Risky clauses noticed late or missed entirely
  • No shared reference for what terms are achievable

Recommended solution

An assistant that drafts counter-offers, flags risky clauses against policy and surfaces internal benchmark terms for the category.

  • Contract policy and red-line rules codified
  • Agent highlights risks and suggests counter-terms
  • Negotiation playbook taught through Dojo role-play

What stays human: The buyer owns the relationship, the conversation and the final terms.

Results to expect

  • Review effort reduced from days to hours
  • Fewer non-compliant clauses reaching signature
  • More consistent commercial outcomes between buyers
TrackingERN scope: Advisory + Training

Order tracking that warns before the delay lands

The problem

Delays were discovered from supplier emails after the fact, leaving planners to react rather than reroute.

  • Reactive updates after the delay had happened
  • No early signal on at-risk shipments
  • Escalations depended on individual follow-up

Recommended solution

Shipment and milestone data consolidated into a monitoring agent that predicts delays and proposes rerouting or expediting options.

  • Milestone data model defined across carriers and suppliers
  • Agent predicts delays and alerts affected teams
  • Exception playbooks defined per criticality level

What stays human: Planners decide the exception response and when to escalate.

Results to expect

  • Delay warnings ahead of the promised date
  • Fewer emergency freight and line-stop events
  • Shared visibility between procurement and operations
MatchingERN scope: Advisory + Training

Three-way invoice matching without the month-end crunch

The problem

Accounts payable spent days each cycle reconciling purchase orders, goods receipts and invoices, and disputes surfaced only at payment time.

  • Manual reconciliation across three document sets
  • Exceptions found late in the payment cycle
  • Supplier trust eroded by avoidable payment delays

Recommended solution

Automated three-way matching with tolerance rules, exception routing and clean master data as a precondition.

  • Vendor and item master data cleaned before go-live
  • Agent matches in seconds and flags exceptions only
  • AP team trained on exception handling and controls

What stays human: AP resolves disputes and authorises payment.

Results to expect

  • Matching effort concentrated on genuine exceptions
  • Shorter cycle time to payment approval
  • Cleaner audit trail for every transaction
Risk monitoringERN scope: Advisory

Continuous supplier risk monitoring instead of quarterly reviews

The problem

Supplier risk was reviewed quarterly, leaving long blind spots between assessments on financial, compliance and sanctions exposure.

  • Quarterly reviews with blind spots in between
  • News, financial and sanctions signals checked ad hoc
  • No prioritisation between critical and minor suppliers

Recommended solution

A monitoring agent watching news, financial signals, ESG and sanctions lists against a tiered supplier register.

  • Supplier tiering and risk appetite defined first
  • Real-time alerts on every material risk signal
  • Mitigation ownership assigned per supplier tier

What stays human: Risk owners investigate signals and decide mitigation or exit.

Results to expect

  • Risk events surfaced days or weeks earlier
  • Attention concentrated on critical suppliers
  • Documented evidence for governance and audit

These are illustrative examples written to show the shape of an engagement — problem, solution and expected results. They are not attributed to any named client, and the percentage ranges are indicative industry benchmarks rather than guaranteed outcomes.

Before vs after

Process, timeline and cost at a glance

The same procurement cycle, compared before and after an agentic AI programme. Timeline and cost figures are indicative ranges based on typical mid-size procurement operations.

Process

Supplier discovery

Wider pool, one standard

Bid evaluation

Auditable award rationale

Invoice matching

Exception-driven work

Risk monitoring

Earlier warning

Timeline

RFQ preparation

Hours to minutes

Quote comparison

Days to hours

Contract review

~70% faster

Delay awareness

Reactive to predictive

Cost

Procurement productivity

25–40% gain

Purchase cost

5–15% savings

Process cost

~30% efficiency

Expedite & penalty cost

Avoided cost

Market context

Where organisations stand on procurement AI

Adoption intent is close to universal; production-grade deployment is not. The gap between the two is exactly where advisory plus training decides the outcome.

Use generative AI weekly

94%

Piloted generative AI

49%

Have agentic AI in production

21%

Achieved large-scale deployment

4%

Reported procurement productivity gain

25–40%

Reported process efficiency improvement

30%

Indicative industry figures (Hackett Group, McKinsey, Wharton), used for context only.

How we would run it

Four steps from use case to working capability

The same sequence used in our Industry 4.0 work applies to procurement: decide first, assess readiness, integrate, then certify the people who run it.

1. Pick one decision

Choose a recurring procurement decision with measurable cost — supplier selection, counter-offer, exception handling — and define what good looks like.

2. Assess data and process readiness

Check master data, PO/GRN/invoice quality, approval rules and policy coverage. Agentic AI amplifies data problems; it does not fix them.

3. Integrate with guardrails

Connect ERP, e-procurement and document sources; set approval thresholds, audit logging and human-in-the-loop checkpoints before go-live.

4. Certify the team

ERN Academy modules plus Practical Dojo sessions so buyers, AP staff and planners can supervise the agent, read its evidence and override it correctly.

See how we workERN Academy & Dojo

Map one of these examples onto your own procurement

Bring one category or one recurring bottleneck. We will size the advisory scope, the data work required and the training path that keeps it running.

Book an Executive Discussion