IESE AI Club

Role playbook

AI for operations.

Operations may have the most to gain. It runs on unread documents, undocumented processes and day-eating exceptions; wins are measured in hours, errors and stock. It also touches real money and safety, so confident wrong answers cost more.

This assumes you have the base. Start there if not

The job

What this role actually does.

Six things every week. Model releases do not change them; this page should still hold in three years.

01Process 02Planning 03Suppliers 04Exceptions 05Compliance 06Improvement

Most examples are vendor customer stories. Their self-reported, unaudited numbers show something was built and used—not a benchmark to expect.

01

Process mapping and documentation

Drafting
The use case
Turning how the work is actually done (recordings, walkthroughs, transcripts, existing scraps) into procedures somebody can follow.
Job to be done
When a process lives in one person's head, I want it written down accurately without a two-week documentation project, so we are not one resignation away from losing it.
Value to the business
Faster onboarding, less key-person risk, and the base layer any later automation needs.
How to evaluate it
Coverage of critical processes, time to produce a procedure, and time to competence for a new joiner. The process owner signs off: a plausible SOP that is subtly wrong is worse than none.

What AI does here

AI turns a spoken process or recorded walkthrough into a first procedure, exception paths and training material. Ask it which steps are ambiguous; that is where new joiners fail.

Workflow map

  1. InputRecord a walkthrough, collect existing forms and name the process owner.
  2. AI-assisted processTranscribe the walkthrough; draft steps, decision points, exceptions and training notes.
  3. Human checkpointOwner tests the draft against live work and corrects unsafe or obsolete steps.
  4. OutputA versioned SOP with clear owners, hand-offs and escalation paths.

What is still yours

Decide how the process should work, not how it does. Efficient documentation of a bad process just spreads it. Understand each workaround before removing it.

02

Demand planning and forecasting

Analysis
The use case
Forecasting demand from history, seasonality and leading signals, and putting the result in front of the people who place orders.
Job to be done
When I plan next month's stock, I want a forecast built from all the data rather than last month plus instinct, so I hold less inventory without running out.
Value to the business
Working capital released and fewer stockouts.
How to evaluate it
Forecast error against the previous method, stockout rate and inventory holding cost. Measure against the old baseline, not against perfection.

What AI does here

AI accelerates scenario comparison, seasonality checks and stale-assumption tests on your exports. Use it to interrogate the forecast, not produce it; make it state data assumptions.

Workflow map

  1. InputBring sales history, inventory, promotions, calendar events and supply constraints together.
  2. AI-assisted processGenerate a baseline forecast, flag outliers and compare demand scenarios.
  3. Human checkpointPlanner challenges assumptions, approves overrides and checks service and cash trade-offs.
  4. OutputA documented demand plan feeding purchasing and capacity decisions.

What is still yours

Own the assumption. A planner knows which input is fragile and what moves downstream. A chat interface is not automatically better than the forecasting model you already run.

03

Supplier and vendor management

Extraction
The use case
Reading the vendor paper (contracts, quotes, invoices, certificates) and extracting the terms that matter into one consistent record.
Job to be done
When vendor documents arrive in twenty formats, I want the key terms extracted the same way every time, so I can compare, renew and dispute from a record instead of a folder.
Value to the business
Renewals caught before they auto-renew, price and penalty terms made visible, and re-keying removed.
How to evaluate it
Documents processed without manual intervention, extraction accuracy on a sampled set, and value recovered on renegotiated or disputed terms. Sample the extractions: commercial terms are exactly where a quiet error is expensive.

What AI does here

AI extracts from contracts, quotes, invoices, specifications, certificates and delivery notes. Compare like terms, pull renewal and penalty clauses into one table and flag delivered specs that differ from contract.

Workflow map

  1. InputCollect quotes, contracts, invoices, specifications and certificates from each supplier.
  2. AI-assisted processExtract standard fields, compare terms and flag missing, conflicting or expiring items.
  3. Human checkpointProcurement validates exceptions and seeks legal or supplier clarification where needed.
  4. OutputA comparable vendor record and a prioritised negotiation or renewal list.

What is still yours

The relationship, negotiation and legal review. Extraction finds the clause; it does not decide acceptance. Liability goes to someone authorised to sign.

04

Exception handling

Automation
The use case
The daily queue of things that did not go to plan (the late delivery, the mismatched invoice, the failed payment), triaged and routed.
Job to be done
When an exception appears, I want it categorised, enriched and sent to whoever can fix it, so the queue does not become a backlog nobody owns.
Value to the business
Shorter resolution times, and a team spending its day resolving exceptions rather than sorting them.
How to evaluate it
Time to resolution by exception type, share auto-routed correctly, and escalation rate. Anything touching money or a customer commitment keeps a human approval step.

What AI does here

Most ops work is exceptions: late shipments, mismatched invoices, escalations and failures to reconcile. AI can triage, route, draft and escalate. Start high-volume, low-stakes; retain approval until failure modes are known.

Workflow map

  1. InputCapture the exception, its source documents, account context and service-level deadline.
  2. AI-assisted processClassify, extract facts, draft a response and route routine cases by agreed rules.
  3. Human checkpointOwner reviews high-value, unusual or low-confidence cases and changes the route if needed.
  4. OutputA resolved case or an accountable escalation with an audit trail.

What is still yours

Set the threshold and escalation path. Acceptable error rate is a risk decision. Automate routine cases and deliberately route unusual ones to a human.

05

Quality and compliance

Knowledge
The use case
Checking work against the standard: audit preparation, control testing, and policy questions answered from approved documents.
Job to be done
When I run a check, I want the current rule and the relevant evidence in front of me, so I audit against what the policy says today rather than what I remember.
Value to the business
Consistent audits across regions, and less dependence on a central team for routine questions.
How to evaluate it
Audit planning and cycle time, findings quality on review, and the share of answers traced to a current approved source. Point it at a controlled document set: an assistant reading stale policy is a liability.

What AI does here

Grounded, clause-cited answers over regulation, standards and policy improve on PDF searching. For audits, gather evidence, map it to requirements and find gaps first.

Workflow map

  1. InputLoad approved policies, controls, standards, evidence and current audit requirements.
  2. AI-assisted processRetrieve relevant clauses, map evidence to controls and highlight gaps or stale documents.
  3. Human checkpointQualified compliance or quality staff verify the cited source and decide the response.
  4. OutputAn evidence-backed audit pack, remediation list and named control owner.

What is still yours

The final answer, without exception. Use AI to locate the clause, never as your compliance position. A qualified person makes the call and signs it.

06

Reporting and continuous improvement

Automation
The use case
Assembling the recurring report and doing the first pass on why the numbers moved.
Job to be done
When the monthly pack is due, I want the assembly done and the variances flagged, so my time goes to the explanation rather than the spreadsheet.
Value to the business
Analyst weeks returned, and reports that arrive early enough to act on.
How to evaluate it
Report cycle time, number of manual touches, and whether any decision actually changed as a result. Every figure is reconciled to source before it leaves.

What AI does here

Automate weekly-pack gathering and formatting. Spend the recovered time asking why the number moved; AI can generate candidate causes you are too close to see.

Workflow map

  1. InputDefine the KPI dictionary, source systems, reporting cadence and the decision each metric supports.
  2. AI-assisted processGather and reconcile data, draft commentary and surface candidate drivers or anomalies.
  3. Human checkpointOperations lead validates the numbers, chooses the cause to investigate and assigns action.
  4. OutputA trusted operating review with owners, actions and measures of follow-through.

What is still yours

Actually change the process. Improvement dies between finding and behaviour change; that gap closes on the shop floor or in the warehouse, not in a better report.

Start this week

  1. Record yourself walking through one undocumented process and turn it into a written procedure before the end of the week.
  2. Pull every renewal date and penalty clause out of your supplier contracts into a single table.
  3. Take your highest-volume exception type and automate the triage and draft response, with a person still approving every one.

Where it fails in this role

  • Anything where being wrong is expensive and unchecked. Safety, regulated decisions, payments and anything carrying liability keeps a human signature on it.
  • Replacing a real system. If you need inventory optimisation or statistical forecasting, use software built for that.
  • Fixing a broken process. Automating chaos gives you faster chaos and a considerably harder problem to unpick later.

Other playbooks

All nine playbooks Product management Sales Marketing Venture capital Consulting Forward deployed engineer Founder / entrepreneur AI product manager

Now do it somewhere real.

A playbook is a map. The Industry Fellowship is the terrain: a term spent talking to professionals who are implementing AI in one industry, working out where it actually creates value, and building a working prototype against what you find. Most fellows start as beginners.

Explore the Fellowship Apply for the Class of 2028