IESE AI Club

Role playbook

AI for marketing.

Marketing production is nearly free: posts, landing pages, ad variants and product photos. Production no longer differentiates. Knowing what to say, to whom, and what is merely adequate does.

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.

01Positioning 02Research 03Content 04Creative 05Distribution 06Measurement

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

01

Positioning and messaging

Synthesis
The use case
Mapping the competitive claim landscape and the gap between the language customers use and the language you use.
Job to be done
When I set positioning, I want the market's claims laid out and the empty spaces named, so I choose a position deliberately instead of by default.
Value to the business
Messaging that actually differentiates, and one argument the whole company can repeat.
How to evaluate it
Message-test results and whether sales can repeat the positioning unprompted; unaided recall in the target segment over time. An empty space may be empty for a reason.

What AI does here

AI reads competitor sites, analyst notes, reviews and win/loss calls, then contrasts their language with customers’. Map claims, find unoccupied space and ask why it may be empty.

Workflow map

  1. InputCompetitor claims, win/loss notes, reviews and brand strategy.
  2. AI-assisted processClusters claims and contrasts market language with customer language.
  3. Human checkpointMarketing leadership tests the space against product truth and strategy.
  4. OutputA positioning hypothesis, message hierarchy and evidence to validate.

What is still yours

The choice. Positioning means deciding what you are not and accepting the customers you lose. A model cannot know what your company can deliver.

02

Audience and market research

Research
The use case
Interrogating a corpus too large to read: reviews, community posts, support history, every interview the company has ever run.
Job to be done
When I need to understand a segment, I want the recurring language and complaints pulled out with quotes, so I start from evidence rather than a persona document.
Value to the business
Research that took weeks becomes an input to a normal week.
How to evaluate it
Time from question to insight brief, and how often a claim in the brief survives being traced back to source. The public corpus is available to competitors too; the advantage is what you do with it.

What AI does here

AI reads reviews, community posts, support history and interviews at once. Find recurring complaints, language patterns and customer words missing from your copy—often a better headline than a week of workshops.

Workflow map

  1. InputInterview transcripts, support tickets, reviews and survey responses.
  2. AI-assisted processTranscribes, groups themes and retrieves supporting quotes at scale.
  3. Human checkpointResearcher checks samples, speaks to customers and interprets exceptions.
  4. OutputA prioritized insight brief with traceable evidence.

What is still yours

Talk to people. Synthesising existing material is not asking something new. Every competitor has the same public corpus; insight from it alone is not an advantage.

03

Content production

Drafting
The use case
Producing the channel variants of one idea: the post, the email, the captions, the script outline, the SMS.
Job to be done
When I have one thing to say, I want it shaped for each channel without writing it five times, so the calendar stops being the constraint.
Value to the business
Publishing volume a small team could not previously reach.
How to evaluate it
Cost and time per published asset, and engagement per asset. The second matters more. If volume rises while engagement per asset falls, you are producing noise faster.

What AI does here

AI drafts, outlines, variants, repurposes, translates and refreshes. Give it a voice guide and real examples of your best work; otherwise volume makes the brand sound like everyone else.

Workflow map

  1. InputBrief, source material, voice guide, audience and channel constraints.
  2. AI-assisted processCreates outlines, first drafts and channel-specific versions.
  3. Human checkpointEditor verifies facts, voice, distinctiveness and final claims.
  4. OutputA publishable asset package with an accountable owner.

What is still yours

Edit, and delete. Default output is competent, safe and forgettable. Someone must hold an opinion and cut the paragraph that says nothing.

04

Creative and assets

Image
The use case
Generating and varying visual assets (variants, resizes, localisations) inside brand constraints.
Job to be done
When a campaign needs fifty variants, I want them produced on brand, so testing is not capped by design capacity.
Value to the business
More tested variants per campaign, and designers moved onto work that needs judgment.
How to evaluate it
Assets per campaign, cost per asset, and the lift from the winning variant. Brand review stays human, and rights and likeness get checked before anything is published.

What AI does here

AI creates concepts, storyboards, mood boards, ad variants, product shots, video and voiceover. Small teams can test directions cheaply, then spend on the one that works.

Workflow map

  1. InputCreative brief, approved brand system, product imagery and media requirements.
  2. AI-assisted processGenerates concept directions and adaptable asset variations for review.
  3. Human checkpointCreative lead checks brand, rights, representation and production quality.
  4. OutputA selected campaign direction and production-ready asset set.

What is still yours

Brand consistency, and knowing when generated is not good enough. It has a look and audiences spot it. Check commercial rights and required disclosure.

05

Distribution and channel operations

Automation
The use case
The plumbing: routing, syncing, tagging, opt-outs and handoffs between the tools that hold your funnel.
Job to be done
When a lead or campaign event happens, I want it to reach the right system and the right person without anyone remembering to do it.
Value to the business
Fewer leads lost between systems, and hours returned from manual list work.
How to evaluate it
Sync error rate, lead response time, and manual data-movement hours removed. Test opt-out and consent handling deliberately. That is where this kind of automation causes legal problems.

What AI does here

Marketing operations is repetitive multi-step work: reformatting, scheduling, briefing, tagging, chasing, reporting and lead routing. Automate it; tool wiring is valuable on a small team.

Workflow map

  1. InputApproved asset, calendar, audience rules, CRM fields and channel access.
  2. AI-assisted processFormats copy, routes leads, tags records and drafts routine reports.
  3. Human checkpointOperator approves triggers, monitors errors and handles exceptions.
  4. OutputConsistent execution with an auditable handoff to sales or support.

What is still yours

Channel strategy. Automating a channel buyers do not use wastes time faster. Every flow breaks quietly; someone must own it.

06

Measurement

Analysis
The use case
Joining channel and CRM data into an answer about what actually drove revenue.
Job to be done
When I decide where next quarter's budget goes, I want the funnel joined end to end, so I optimise toward customers rather than toward form fills.
Value to the business
Budget moved to what converts, with a shorter loop between spend and evidence.
How to evaluate it
Time to a reliable channel read, and the downstream conversion rate of the events you optimise toward. A faster wrong attribution model is worse than a slow honest one.

What AI does here

Ask your own data what moved, which channel gets false credit and how attribution windows change the number. Make it state its assumptions before you believe it.

Workflow map

  1. InputClean campaign, CRM and conversion data plus agreed metric definitions.
  2. AI-assisted processExplores segments, explains changes and proposes follow-up questions.
  3. Human checkpointAnalyst checks attribution, confounders and business significance.
  4. OutputA decision-ready performance readout with documented assumptions.

What is still yours

Decide what is worth measuring. Marketing lacks honesty about vanity metrics, not numbers. A model computes anything, including the metric designed to flatter you.

Start this week

  1. Write your voice guide, add three examples of your genuinely best work, and put it in one project that you use for every draft from now on.
  2. Take one strong asset and repurpose it into every format you support, in a single sitting.
  3. Automate your weekly report so that you never assemble it by hand again.

Where it fails in this role

  • Strategy. It will generate a marketing plan for any business in thirty seconds and it will be generic, because it is assembled from generic plans.
  • Being interesting. Safe and average is the default setting, and average is invisible.
  • Owning the brand. Consistency requires someone with authority to say no. Nobody has automated the ability to say no.

Other playbooks

All nine playbooks Product management Sales Venture capital Operations 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