Reflecting on three years of AI in marketing

In 2023, generative AI was still an emerging marketing capability. Capgemini found that 37% of organizations were implementing GenAI in marketing and another 21% were experimenting with it.

Three years later, AI is becoming standard marketing infrastructure. Nearly 60% of marketers now use AI several times a week, while McKinsey reports regular AI use somewhere in the organization at almost 90% of companies.

The more interesting development is how unevenly marketing has progressed. Adoption, content production, individual productivity and technical capability have advanced rapidly. End-to-end workflows, data foundations, organizational capabilities, governance and demonstrable enterprise value have moved much more slowly. At the same time, AI has started changing the customer journey itself.

This is not a clean longitudinal comparison. Capgemini's 2023 study covered 1,800 marketing executives at companies with more than $1 billion in revenue across fourteen countries. The 2026 studies use different populations, definitions and questions. The comparison is therefore directional rather than a precise three-year time series.

Against that background, eight developments deserve CMO attention.

1. Adoption is no longer a useful measure of progress

In 2023, 58% of organizations in Capgemini's study were either implementing or experimenting with GenAI in marketing.

By 2026, McKinsey reports that almost 90% of organizations regularly use AI in at least one function. Forty-four percent report enterprise-scale deployment, 56% use AI in three or more functions, and 40% of companies above $1 billion in revenue are scaling agents. Nearly 60% of marketers use AI several times a week.

Yet almost 90% of CMOs are experimenting with AI use cases while fewer than 10% report value from AI across end-to-end marketing workflows.

BCG finds a similar discrepancy. Ninety-six percent of CMOs discuss end-to-end transformation, yet 42% still use GenAI primarily as an assistant for individual tasks. Only 8% report autonomous execution of some campaigns by multiple agents.

Adoption has largely been achieved. It no longer tells a CMO much about marketing maturity. The more relevant measure is where AI is producing measurable value across the commercial system.

2. Individual productivity has advanced much faster than enterprise value

One of the clearest successes since 2023 is productivity.

Adobe reports that 75% of CMOs see improvements in content production volume and speed, 71% see greater content-production capability among non-creative teams, and 63% report improvements in experimentation and innovation.

There are tangible cases as well. McKinsey describes a Chime case in which campaign lead times fell from around ten weeks to four, testing volume tripled and ROAS increased by 18%. Other agentic workflow pilots have produced substantial acceleration.

The enterprise-level evidence is less convincing.

McKinsey reports individual productivity improvements at 80% of organizations and better decision-making at 50%. Yet only 37% report any EBIT impact, approximately unchanged from 2025. Around 6% qualify as high performers with both material EBIT and broader enterprise impact.

This does not prove that productivity improvements fail to create financial value. It does show that evidence of individual productivity has become considerably stronger than evidence of enterprise-level value realization.

That distinction matters. Saving a marketer an hour creates capacity, not automatically economic value. The business case depends on whether that capacity produces more output, higher quality, shorter lead times, more experimentation, lower external spending, avoided hiring or an actual reduction in internal capacity.

3. End-to-end marketing has barely kept pace

AI now touches insights, planning, content, media, commerce, sales, service and pricing. The breadth of possible applications has expanded enormously since 2023. The connections between them have advanced much less.

Fewer than 10% of organizations in McKinsey's marketing research report AI value across end-to-end workflows. Only 28% of marketers say their organizations are fundamentally redesigning teams and workflows.

This becomes more consequential as agents move from isolated tasks to sequences of work. Optimizing content generation is relatively contained. Optimizing a commercial workflow requires data to move across systems, decisions to have clear owners, exceptions to be handled, performance to be measured and responsibility between people and agents to be explicit.

For the next phase, the gap between task maturity and workflow maturity matters more than the number of AI use cases.

4. Better models have not solved the data problem

Data was already a prerequisite in 2023. Rapid advances in models have not made it less important.

Adobe reports that 78% see data integration and quality as significant barriers to agentic AI. Only 31% say they have a unified customer data foundation.

Salesforce finds that 75% of marketers use AI, yet 84% still run generic campaigns and 69% struggle to respond to customers in a timely way. Access to data across service, sales and commerce remains fragmented.

The same pattern appears in customer experience. Talkdesk and NewtonX found that 98% of organizations use AI somewhere in the customer journey, but only 15% combine agentic AI with cross-functional orchestration. Only 35% preserve customer context when moving between systems.

AI has not circumvented fragmented customer data. More capable systems arguably expose the problem more clearly: they cannot reliably reason or act on context they cannot access.

For CMOs, the data foundation remains unfinished transformation infrastructure, not a legacy issue that can be left behind as attention shifts to agents.

5. Training has scaled, but the skills gap remains

In 2023, Capgemini found that 63% saw demand for GenAI skills exceeding supply and 53% planned to provide training.

By 2026, BCG reports that roughly 80% of CMOs are making substantial investments in AI upskilling. Investment in responsible-AI and ethics training has increased as well.

Yet Adobe still finds that 71% of CMOs identify talent and skills as a major barrier to agentic AI.

There is also a less visible organizational issue. McKinsey finds 87% enthusiasm alongside 57% general role uncertainty. Among CMOs themselves, 96% are enthusiastic, while 71% experience anxiety and 80% see risk to their own role.

The apparent contradiction is useful. Training is an input; capability is an organizational outcome.

As AI moves from assistance towards execution, marketing needs people who can design workflows, exercise judgment, manage agents, establish controls, interpret performance and determine when human intervention is required. Roles such as AI product owner, workflow designer, agent manager, orchestrator and governance owner are emerging around these needs.

A more demanding measure of progress is therefore what the marketing organization can now reliably do, rather than how many employees have received AI training.

6. Governance is improving, but autonomy is moving faster

Governance has professionalized since 2023. Capgemini then found that 30% had clear AI guidelines and 42% had measures addressing copyright and intellectual property.

McKinsey's 2026 AI Trust research shows improving responsible-AI maturity, but only around 30% of organizations reach maturity level three or above in areas including strategy, governance and agentic controls.

More importantly, the governance problem itself has changed.

In 2023, the emphasis was largely on how people should use systems that generated content, analysis and recommendations. Organizations increasingly need to govern systems that can select actions, optimize campaigns, interact with other systems and potentially transact.

This brings AI governance into operational management. What may an agent decide? Which objectives may it optimize? What data can it use? When must a person intervene? Who owns the outcome? Can actions be reconstructed when something goes wrong?

Governance maturity now needs to be assessed against the level of autonomy being introduced.

7. Creative capacity has increased. Distinctiveness is a different question.

The creative promise of GenAI has partly materialized. Adobe's evidence on speed, production volume, accessibility and experimentation shows that marketing organizations can produce substantially more creative material.

Forrester finds that nine out of ten US agencies use GenAI and half use agentic AI for execution. Employee productivity is the primary GenAI objective for 81%, while efficiency also dominates agentic-AI objectives.

That creates a strategic tension. When every brand and agency can produce competent content faster and more cheaply, the supply of acceptable creative work increases. That does not necessarily make any individual brand more distinctive.

The evidence supports increased creative capacity. It does not establish increased creative distinctiveness.

The implication is not to resist efficiency. It is to avoid confusing production efficiency with marketing effectiveness. Brand judgment, originality and quality control become more consequential as production becomes easier.

8. The customer journey is changing as marketing is still transforming

The largest development was barely visible in the 2023 baseline.

The original GenAI discussion was primarily about AI inside marketing: how marketers could use it to create, analyze, personalize and optimize. AI is now also appearing between brands and customers.

BCG reports that 90% of CMOs believe AI is changing brand discovery and evaluation. Ninety-one percent of B2C CMOs and 76% of B2B CMOs see no-click AI discovery changing the funnel.

Adobe finds that 70% consider conversational AI platforms important to brand relevance, while 52% are preparing content for AI discovery.

Consumer intentions point further ahead. In Accenture research among more than 25,000 consumers, 74% say they would delegate routine tasks to an agent. Thirty-two percent would allow an agent to decide what to buy within defined boundaries, and 37% of loyal consumers would allow an agent to switch from their preferred brand for a better match.

These findings do not establish that autonomous commerce is already mainstream. They do indicate a potentially important shift in how choices are mediated.

Marketing has traditionally optimized how people discover, evaluate and choose brands. CMOs may increasingly need to consider how AI systems acting on behalf of customers find, interpret, compare and trust product information, claims, pricing, reputation and availability.

The timing is significant: parts of the transformation envisioned in 2023 remain unfinished while the customer environment for which they were designed is already changing.

A directional maturity scorecard for 2026

The studies are not directly comparable enough to construct a scientific maturity index. Taken together, however, they support a useful directional assessment.

The asymmetry is more informative than the individual scores. Marketing has become relatively mature in access to AI and its application to individual activities. The capabilities required to convert this into integrated commercial performance remain much less developed.

Seven priorities for the CMO

Measure end-to-end value. More users, pilots and use cases are no longer meaningful evidence of transformation. Trace value through complete commercial workflows.

Separate productivity from enterprise value. Track where AI-created capacity goes and whether it improves output, quality, speed, experimentation, external spending, hiring requirements or internal capacity.

Keep customer data on the transformation agenda. Better models do not compensate for fragmented context. Increasing autonomy raises rather than reduces the value of reliable, accessible data.

Manage capability, not training volume. Assess what teams can independently design, operate, evaluate and improve.

Match governance to autonomy. Controls designed for copilots will not necessarily be sufficient for agents taking consequential commercial actions.

Protect creative distinctiveness. As production becomes abundant, efficiency and differentiation need to be managed as separate objectives.

Design for AI-mediated customer journeys. Understand not only how your own AI operates, but how customers' AI systems may discover, interpret, compare and ultimately select your products and brands.

Where this leaves marketing in 2026

Three years of AI have produced substantial progress. Adoption is dramatically higher, content production is faster, individual productivity gains are increasingly tangible and technical capabilities have advanced beyond what most marketing organizations were considering in 2023.

The harder parts of the transformation have moved more slowly: end-to-end value, integrated customer data, organizational capability, governance and operating-model redesign.

That imbalance matters because the external environment is no longer standing still. AI is beginning to influence how customers discover, evaluate and potentially buy at the same time that marketing organizations are still integrating AI into their own operations.

For CMOs, that creates two agendas that now have to be managed simultaneously: finish the internal transformation required to turn AI capability into sustained commercial value, while adapting marketing to a customer journey increasingly mediated by AI.