marketing leadership
WHAT DOES AN AI-AUGMENTED MARKETING DEPARTMENT LOOK LIKE?
Marketing has been organised around the production and movement of assets. That constraint is disappearing. The change that matters is that the function becomes a continuous sensing and response system, with people setting the point of view and agents preparing the work around it. Judgment stays with the person. Attention moves toward customers, evidence and decisions.
Production absorbs
the function.
A typical Nordic mid-market marketing and communications function has 4 to 15 employees. Multibrand or international companies may have 15 to 25. The team usually includes a CMO or marketing director, a brand or communications lead, one or two content specialists, a demand-generation or digital marketer, a marketing operations specialist, and sometimes events or internal communications. Creative, media, PR, web development and production often sit in two to four agencies.
The function sits between company strategy, customers, sales and the public market. It drives demand, conversion, retention and pricing power. Stronger functions connect marketing activity to sales progression, customer behaviour, pricing and retention. Weaker ones remain centred on impressions, posts and email opens.
Eight processes- 01Positioning, messaging and brand governance
- 02External content production
- 03Campaign planning, execution and agency management
- 04Lead generation and nurturing with sales
- 05Website management, SEO and answer visibility
- 06Events and field marketing
- 07Internal communications
- 08Analytics, attribution and budget allocation
Information is scattered across the CRM, marketing automation, the CMS, GA4, Search Console, advertising platforms, document stores, agency folders and spreadsheets. Brand judgment, customer nuance and the reasons behind past decisions remain disproportionately in people's heads or agency relationships.
Production has been the organising constraint. Drafting, translation, resizing, versioning, tagging, research synthesis and first-pass reporting consume attention that could go to the market.
Dubé and Xu's three randomised Wine Access experiments found that model-written and human-edited copy took at most ten minutes to prepare and check, compared with about three hours for a writer. Human fact-checking stayed in the process. Pew's March 2025 browsing study found conventional-result clicks on 8 percent of Google visits containing an AI summary, versus 15 percent without one. Different changes, the same demand: the function must rethink both production and discovery.
Marketing becomes
a response system.
The operating model shifts along two linked dimensions.
Production stops organising the work. People set the point of view, supply evidence, choose ideas and edit. Governed agents draft, translate, resize, tag and adapt the working material. The released capacity goes into named-account research, local-language variants, continuous experimentation, reuse of expert knowledge and faster support for sales. Publishing more generic content is not the prize.
Periodic campaigns give way to continuous sensing and response. Customer behaviour, sales conversations, website activity, search visibility, service issues, market news and campaign results are reviewed together. Work starts when signals change, not only when the annual calendar says it should.
Discovery changes with the operating model. Owned content serves buyers and the systems that summarise information for them. Useful original content, sound SEO, crawlability and accurate structured data remain the foundation. Direct relationships through opted-in email, events, communities, customer programmes and proprietary research matter more when platforms can provide visibility without traffic.
An AI-augmented marketing department does not wait for the next campaign cycle. The function sees a change, assembles the evidence, prepares a response and puts the consequential choice in front of a person.
Where it is weakestcategory position, distinctive codes, major brand work, executive statements, crisis communication, regulated claims and novel creative concepts. These depend on accountable human judgment, and should.
A day, a week,
a month.
A function that invests heavily still has recognisable marketers, communicators, agencies and systems in 2028. Its cadence and attention are different.
The dayThe operating layer assembles a briefing from CRM activity, website behaviour, campaign performance, sales-call summaries, customer-service themes, media signals and approved market sources. Exceptions come first: new interest from a valuable account, a campaign moving outside its expected range, an outdated product claim, a competitor entering an important answer, or an internal announcement waiting for approval.
Content begins with evidence and intent from product experts, executives, salespeople or customer teams. A workflow develops draft narratives and channel variants under encoded brand rules. Editors verify claims, select the strongest expression and decide what should be published. Routine translations and low-risk adaptations move after sampled quality checks.
Connected workflows prepare segments, drafts, landing-page variants, tracking plans and performance summaries. A marketer approves audiences, claims, publication and material budget changes. Tagging, task creation, draft metadata and pausing an asset that breaches a preset rule can run within defined limits.
The weekMarketing and sales review named accounts, buying signals, content needs and nurture recommendations together. The discovery review covers conventional search, available visibility inside generated answers, conversational referrals and manual tests of important buyer questions. The campaign review treats work as an experiment portfolio, with agreed hypotheses, audiences, claims, stop conditions and spend limits.
The monthThe function produces a decision memo rather than a dashboard tour. It explains what changed, what caused it with stated confidence, which experiments support the conclusion, what sales is seeing and how the budget should move. Platform attribution remains an operational signal. Incrementality tests and marketing-mix models inform allocation.
Continuous attention to market signals, governed adaptation across channels, and evidence behind every material choice.
What runs, and what
stays with the person.
| Process | What the agent does | What stays with the person |
|---|---|---|
| Positioning, messaging and brand governance | Compares approved strategy, customer evidence, competitor claims and past decisions; checks drafts against explicit verbal, visual and factual rules | Category position, distinctive codes, acceptable risks and approval of exceptions |
| External content production | Turns approved source material into a modular content set; prepares channel variants, formatting, tagging, transcription and routine localisation | Source selection, fact-checking, editorial judgment and final publication |
| Campaign planning, execution and agency management | Develops the brief, work breakdown, variants, QA checklist and measurement plan; brings agency outputs into the same flow | The idea, audience promise, budget and final release |
| Lead generation and nurturing with sales | Assembles buying-group context and recommends the next useful content or action from behaviour, deal stage and consent | High-value outreach, relevance judgment and sales feedback |
| Website, SEO and answer visibility | Monitors buyer questions, citations, search, referrals, content gaps and technical health; proposes inventories, links and stale-content updates | Original research, clear expert answers and editorial priorities |
| Events and field marketing | Prepares audience briefs, speaker packs, invitations, follow-up drafts and post-event content | Relationships, live facilitation, sensitive invitations and sales handoff |
| Internal communications | Converts approved decisions into audience-specific drafts, FAQs, manager talking points and translations | Meaning, political sensitivity and leadership accountability for the message |
| Analytics, attribution and allocation | Validates data, explains movement, identifies anomalies, proposes hypotheses and prepares scenarios | Experiment design, challenge to causal claims and material budget reallocation |
Four stages, mapped to
The Access Ladder.
The sequence follows business value and action risk, mapped in order to The Access Ladder.
Build approved context from brand standards, product facts, customer evidence, strategy, claims and disclosure guidance, and strong examples. Marketers approve agent drafts.
Connect CRM and marketing automation for customer history, consent, campaign exposure, deal outcomes and sales notes. Add CMS, GA4, Search Console and performance data for read-only briefs and recommendations.
Allow CMS staging writes, CRM tasks, draft campaigns, metadata and reversible tagging, subject to marketer approval. Publication, audience changes, sends, deletion, pricing claims and material spend retain named approval.
Allow low-impact tagging, task creation, draft metadata and rule-based pausing within limits. Each workflow needs minimum permissions, an evaluation set, owner and revocation path. People retain sensitive communication, final publication and consequential allocation.
A warehouse is necessary for one customer identity across CRM, web, ERP, service and media, and for historical snapshots, margin-aware attribution and consistent experiment data. Workflows receive governed features and aggregates.
Six things we believe about
the responsive marketer.
Six beliefs follow from the structural shift already underway. They are about how marketing and communications should work when production is no longer the organising constraint.
Faster drafting and adaptation matter, but they do not make the market care. The return sits in noticing the right change, choosing a useful point of view and responding with evidence before the moment passes.
Generic volume only increases noise and review work. Put the capacity into named-account research, local language, expert knowledge, owned audiences, experiments and direct support for sales.
Product experts, executives, salespeople and customer teams supply the source material. Agents can develop narratives and variants, but editors choose the claim, verify it and decide whether it deserves publication.
The team sets hypotheses, audiences, claims, stop conditions and spend limits. The operating layer prepares the variants and reads the signals, while marketers decide what the evidence means and where the next test belongs.
Content now speaks to buyers and to the systems that summarise for them. Original knowledge, clear expert answers and a direct opted-in audience become more valuable than chasing a technical trick on any single discovery surface.
Customer context, brand rules, consent, connector permissions and run quality cannot be left between vendors or agencies. Marketing leadership must own how evidence enters, how work moves and which decisions keep a named person.
Publishing more is not the proposition behind these six beliefs. They describe a function that senses continuously, responds deliberately and keeps judgment accountable.
Roles, rhythm, and
where it fails.
The content writer becomes an editor, interviewer and source validator. The campaign manager becomes a workflow owner. The digital marketer becomes an experiment designer. The brand lead turns tacit judgment into rules, examples and exception decisions. Marketing operations owns data definitions, consent propagation, connector permissions and run quality. The CMO owns customer context and operating architecture, not only the budget and calendar.
The practical skills are plain: frame a useful brief, evaluate evidence, edit generated work, understand data permissions and design tests that can disprove a favoured idea. Technique with prompts has limited value without those disciplines. Agencies remain important, but their work moves toward original ideas, research, specialist craft and independent challenge.
The rhythmOne document-based drafting and review workflow
Campaign work linked to approved context, CRM and performance data
Permissions, evaluation samples, revised roles and employee consultation
Continuous sensing, exception handling and accountable action across the function
Where it failsGeneric content volume. No controlled source of truth. Brand rules that remain vague. Agency ownership of core prompts and data. Isolated tools with no workflow owner. Weak evaluation samples. Personalisation without valid permissions. Review work left on top of the old process. Production targets that rise whenever capacity is released.
Involve the people whose work changes while the workflow is still being designed. Their source knowledge, editorial judgment and understanding of exceptions are what turn a drafting tool into a way of working.