McKinsey’s 2026 State of AI survey asks leaders to hold two findings together. Eighty per cent of respondents said AI had improved their individual productivity, yet only 37% reported that it had contributed positively to their organisation’s EBIT, a proportion that was essentially unchanged from the previous year.
The findings come from an online survey of 1,719 respondents across 97 countries, conducted between 4 May and 8 June 2026 and weighted according to each country’s contribution to global GDP. They represent respondents’ perceptions and attributions rather than independently measured productivity or audited financial performance, but the contrast remains instructive: reported individual benefits are considerably more common than reported financial impact at organisation level.
Marketing provides a particularly useful view of this gap because the function combines activities that AI can accelerate quickly with workflows that depend on coordination across the organisation. Research, analysis, drafting, campaign planning and content production can all become faster before anything changes in the way evidence is assembled, decisions are made or work moves between teams.
Where individual productivity gets absorbed
The individual benefits reported by McKinsey are substantial. Alongside the 80% who said AI had improved their productivity, half of respondents believed it helped them make better decisions and around half said it had supported the development of new skills. Adoption is also becoming unremarkable: nearly nine in ten respondents said their organisation regularly used AI in at least one business function, while 44% reported that AI was scaling across the enterprise, up from 38% the previous year.
For marketers, the practical gains are easy to recognise. Research can be summarised more quickly, an initial proposition can generate multiple channel treatments and campaign performance can be interrogated without waiting for a specialist analysis. These improvements are valuable, but they concern particular tasks rather than the complete route from a commercial objective to a campaign in market.
Much of the apparent gain can be absorbed during that route. When a team produces five campaign variants in the time previously required for one, it may also create five variants for product, brand, legal or compliance colleagues to consider. If the brief is incomplete, the evidence is scattered or ownership of a decision is unclear, time saved during production is redistributed across coordination, review and rework. The organisation sees more output at the beginning of the process without necessarily improving the rate at which effective work reaches the customer.
This distinction helps explain why individual productivity can rise much faster than organisation-level performance. A person can measure the time saved on a draft or analysis, whereas the organisation experiences the combined effect of every hand-off, unresolved dependency and competing priority surrounding it. Faster work in one place may shorten the overall cycle, but it may also reveal that production was never the main constraint.
McKinsey’s workforce findings reinforce the point. In its 2025 survey, 32% of respondents expected AI-related reductions in overall employment during the following year; in the 2026 research, only 14% said AI had actually contributed to a workforce decline over the previous year. Reported reductions in marketing and sales were around half the level respondents had predicted. The difference suggests that introducing AI does not automatically remove the work surrounding a task, particularly when people are still needed to assemble context, resolve exceptions and take responsibility for the result.
Workflow redesign is the clearest distinction
About 6% of McKinsey’s respondents qualified as AI high performers, a category covering those who attributed at least 5% of EBIT to AI and described the value created by the technology as significant. The most instructive difference between this group and the rest of the sample concerns the extent to which they had changed the organisation around the technology.
Nearly three-quarters of high performers said their organisation had fundamentally redesigned workflows because of AI, compared with approximately one-quarter of other respondents. High performers were also around twice as likely to report that senior leaders demonstrated commitment to AI initiatives and that the organisation had defined processes for measuring their impact. Most high performers reported pursuing growth or innovation alongside the efficiency ambitions common to both groups.
The research establishes an association rather than a causal relationship. Organisations reporting stronger returns may also have better data, larger budgets, more capable leadership or greater operational maturity, any of which could contribute both to workflow redesign and to financial performance. Even with that caveat, the difference between the groups is large enough to challenge the assumption that widespread access to AI will, by itself, produce enterprise value.
Redesign begins by examining the complete piece of work. Instead of adding an AI tool to one stage, the organisation has to consider which information should be available at the outset, where work repeatedly stops and which decisions are being reconstructed unnecessarily. It also has to establish when work can proceed within agreed boundaries, when an exception requires specialist attention and who owns the eventual outcome.
These decisions are harder than selecting a tool because they expose ambiguities that organisations have often managed informally. A slow process can conceal unclear ownership because people have time to negotiate each case as it arises. Once AI increases the speed and volume of production, the cost of those ambiguities becomes more visible.
Marketing makes the gap visible
McKinsey found that marketing and sales was the function where respondents most commonly reported AI-related revenue gains. That commercial potential is understandable: much of marketing involves language, imagery, research, analysis and variation, all areas in which general-purpose AI is already useful.
The function also depends on decisions that extend beyond the marketing team. A campaign has to reflect the product accurately, make a proposition that matters to its audience and support the commercial objective without damaging trust in the brand. In regulated organisations, those familiar marketing requirements sit alongside formal obligations concerning claims, evidence, customer treatment and specialist approval.
Regulation sharpens the consequences, but the underlying problem applies more broadly. Greater production capacity increases the importance of reliable inputs and clear decision-making because every additional version can introduce another interpretation of the product, audience or proposition. If the relevant context enters only during final review, AI helps the team reach that point sooner while doing little to reduce the work waiting there.
Marketing leaders therefore need to distinguish between requirements that are already settled and decisions that genuinely require interpretation. Routine checks can often be incorporated earlier, while AI-assisted work can proceed within defined boundaries when the source material and permissible changes are clear. Specialists remain necessary for novel claims, conflicting evidence and decisions whose consequences justify accountable human judgement, but applying the same level of review to every item would consume expertise precisely when production is becoming cheaper.
This boundary becomes especially important when the product being marketed contains AI. CAP’s guidance on advertising AI capabilities states that objective claims about what a product can do require appropriate evidence and that advertising must remain socially responsible. It identifies risks including unsupported claims about replacing qualified therapists, the accuracy of image analysis and unrealistic earnings or performance.
Marketing teams are consequently dealing with AI in two different contexts: as a means of producing work and as a capability that may itself need to be described to customers. Both depend on stronger connections between the product evidence available, the claims being developed and the people authorised to resolve uncertainty. Without those connections, faster production can increase the distance between what marketing wants to say and what the organisation can support.
Measure the work rather than the tool
Many organisations can describe how much AI activity is taking place. They know how many licences have been issued, how often tools are used and how much content has been produced, but those measures show participation rather than whether the organisation is working more effectively.
McKinsey’s finding that high performers are around twice as likely to have defined processes for measuring AI’s impact points towards a more demanding approach. For marketing, measurement should follow work across the workflow, connecting the time saved during production with what happens during review, launch and customer response.
A useful assessment might consider the time between brief and launch, the number of avoidable revision rounds, how long work waits for missing information and how often issues reach specialists that could have been resolved earlier. Those operational measures then need to be read alongside commercial outcomes such as incremental revenue, conversion, acquisition cost and customer response, as well as any material effect on risk or quality.
Taken together, these measures prevent a local improvement from being mistaken for an organisational one. Drafting time may fall while review time rises, or content volume may increase without improving campaign performance. A team may reduce production costs while transferring additional work to product, legal or compliance colleagues. Each result can appear positive when viewed through a narrow productivity measure, even though the organisation as a whole has gained little.
Shared measures also create a more useful basis for collaboration. Marketing cannot treat effort transferred elsewhere as productivity saved, while specialist functions cannot demonstrate effective control merely by increasing the volume of review. The relevant outcome is the quality and commercial effectiveness of the work that reaches the customer, together with the time and organisational effort required to get it there.
The operating-model challenge
McKinsey’s research does not offer a simple formula for AI returns, but it shows that reported individual productivity is far more widespread than reported financial impact. It also shows that the small group reporting the strongest returns is much more likely to have redesigned workflows and established ways to measure what AI is contributing.
Marketing leaders can already see the first half of that pattern as employees produce, analyse and explore more quickly. Whether this becomes organisational value depends on what happens between creation and outcome: how information enters the work, where decisions are made, how exceptions are handled and whether performance is measured across the entire process.
As individual AI productivity becomes easier to achieve, the source of advantage moves into the organisation itself. The lasting gain will come from designing workflows capable of using that productivity, rather than allowing it to disappear into the same hand-offs, queues and unresolved decisions that constrained the work before.
What teams need to know
What does McKinsey’s 2026 State of AI survey say about productivity?
Eighty per cent of respondents said AI had improved their individual productivity, while 37% reported a positive contribution to organisational EBIT. These are respondents’ perceptions and attributions rather than independently audited measures.
How are AI high performers different?
Nearly three-quarters of high performers reported fundamentally redesigning workflows because of AI, compared with approximately one-quarter of other respondents. The survey shows an association rather than proving that redesign caused higher returns.
Where can individual AI productivity be lost in marketing?
Time saved during production can be redistributed into coordination, review and rework when briefs are incomplete, evidence is scattered or ownership of decisions is unclear.
How should marketing teams measure AI’s impact?
Measures should follow the complete workflow, combining cycle time, avoidable revisions, waiting time and specialist escalation with commercial outcomes such as incremental revenue, conversion, acquisition cost and customer response.
Does workflow redesign require human review of every AI output?
No. Settled requirements can be incorporated earlier and work can proceed within clear boundaries, while specialists focus on novel claims, conflicting evidence and consequential decisions that require accountable judgement.