The State of AI Adoption in Growth Functions: What the Data Is Actually Telling Us

AI adoption figures are everywhere. Survey after survey reports that some significant percentage of companies are "using AI in their marketing" or that AI is a "top priority" for growth teams in 2025. These numbers are cited in vendor decks, conference keynotes, and analyst reports as evidence that AI transformation in marketing is well underway.

The problem is that "using AI in marketing" covers an enormous range of activity, from a founder prompting ChatGPT once a week to write social captions, to a company running fully integrated Claude API workflows across their entire growth stack. These are not the same thing, and treating them as equivalent produces a distorted picture of where actual adoption stands.

The more interesting and practically useful question is not "how many companies are using AI in marketing?" but "where are growth teams actually deploying AI in ways that change what is achievable, and where is adoption still largely performative?"

This piece attempts to answer that question based on patterns we have observed across client work, industry research, and the original survey work we have conducted on AI adoption in knowledge work.

Where Adoption Is Genuine

Genuine AI adoption in growth functions — meaning deployment that has materially changed the output, quality, or cost structure of the growth operation — is concentrated in a relatively small number of use cases.

Content production assistance is the most widespread genuine adoption area. A large proportion of marketing teams are now using AI tools at some point in their content production workflow — most commonly for first-draft generation, for repurposing long-form content into shorter formats, or for generating variations on existing copy. The quality and depth of this adoption varies enormously: some teams have integrated AI into a systematic content pipeline that produces consistent, high-volume output; most have added AI as an ad hoc tool that individual team members use at their own discretion when they think to.

Email and outreach drafting is the second most common genuine adoption area. AI-assisted email writing — using tools like Claude, GPT-4, or dedicated email AI tools — has become common enough that it is no longer a differentiator for most teams. The adoption is real, but the strategic advantage of doing what everyone else is doing is minimal.

Data summarisation and reporting is increasingly adopted, particularly in teams with data analysts or growth operations staff. Using AI to summarise campaign performance data, generate weekly reports, or produce structured analysis from raw data reduces the time burden of the reporting function significantly. This is adoption that genuinely changes the cost structure of the analytics function.

Prospecting and lead research is an area where genuine, high-impact adoption is growing but still far from universal. Teams using AI to enrich prospect data, generate personalised research notes on target accounts, or automate lead scoring are seeing meaningful improvements in the quality and efficiency of their prospecting operation. But the majority of outbound teams are still running largely manual processes, or are using AI only at the surface level of writing templates rather than at the deeper level of research synthesis and personalisation.

Where Adoption Is Performative

Performative AI adoption — the "we use AI in our marketing" claim that does not translate to meaningful operational change — is far more common than genuine adoption in several areas.

Strategy and positioning. Many marketing teams report using AI for strategic tasks — market analysis, competitor research, positioning development — but in most cases the actual use is surface-level: prompting a general AI tool for broad strategic input, receiving generic output, and using it as a starting point for thinking that remains largely human. This is not adoption that changes what is achievable; it is AI as a brainstorming partner, which is valuable but not transformative.

Social media content. AI-generated social media content is extremely widespread, and in most cases it is AI that is making the output worse, not better. Generic AI social captions that sound like every other generic AI social caption are not a competitive advantage. This category of adoption is real in volume and largely counterproductive in outcome.

AI in title, not in practice. A growing number of companies are claiming AI-native positioning without the underlying systems to support it. The claim is in the marketing; the AI is in the marketing content claiming AI use. This is a credibility risk for any company that is not building genuine AI infrastructure — the gap between claim and capability is increasingly visible to sophisticated buyers and investors.

The Gap That Matters: Systems vs. Tool Use

The most important observation from both our own client work and the broader adoption picture is the gap between AI tool use and AI systems.

AI tool use means individual team members using AI applications — Claude, ChatGPT, Gemini, Midjourney — to assist with tasks they were already doing. This is adoption, but it is fundamentally additive: you get some productivity improvement in the tasks where it is used, but the underlying process has not changed.

AI systems means building AI into the process architecture: automating workflows so that AI reasoning is applied consistently and at scale, regardless of whether a specific human team member remembers to use a tool. The output improves not because individuals are using AI ad hoc, but because the system is designed to apply AI at every relevant step.

The companies that have made the shift from tool use to systems deployment are the ones where genuine competitive advantage is emerging. They are not just more productive — they are operating with a fundamentally different cost structure and output ceiling. Their content volume is higher. Their outreach personalisation is better. Their analytics are more current and more integrated. And the gap is compounding: every cycle the system runs, the prompts improve, the integrations deepen, and the advantage increases.

What the Research on Cognitive Redistribution Tells Us

At GrowthBoxx, our published SSRN working paper examined how AI tools are redistributing cognitive load in knowledge work across a sample of 150 professionals. Several findings from that research are directly relevant to the AI adoption conversation in growth functions.

The first and perhaps most important finding is that AI tool adoption does not automatically redistribute cognitive load toward higher-value work. In many cases, the time saved by AI assistance is absorbed by other low-value tasks rather than redirected toward the strategic and creative functions that AI cannot perform as well as humans. This suggests that the productivity gains of AI adoption are highly sensitive to how adoption is structured — whether the organisation has actively redesigned workflows to capture the freed capacity, or whether it has simply added AI tools and hoped the benefit would materialise.

The second relevant finding is that professionals with higher AI literacy — those who understand not just how to use AI tools but how to construct effective prompts and integrate AI into their workflow — see materially better outcomes than those using the same tools with lower literacy. The tool is not the differentiator; the skill in using the tool is. This has direct implications for growth teams: the ROI of AI adoption depends significantly on whether the team using the tools has the skills to use them effectively.

The third finding is that the areas where humans feel AI assistance improves quality are often different from the areas where objective quality metrics show improvement. There is a tendency to perceive AI assistance as improving quality broadly, including in areas where the AI output is actually mediocre — which can mask the cases where AI assistance is making the output worse.

Implications for Growth Leaders

For founders and growth leaders trying to separate signal from noise in the AI adoption conversation, three practical implications emerge from this analysis.

Audit your actual adoption, not your stated adoption. Map every point in your growth process where AI is currently being applied and characterise whether that application is systematic (it always happens as part of the workflow) or ad hoc (it happens when someone remembers). The systematic applications are the ones that are changing your cost structure. The ad hoc ones are the ones that feel like progress but may not be.

Prioritise systems over tools. The marginal value of adopting another AI tool is lower than the marginal value of systematising the use of tools you already have. If you are using Claude for outreach personalisation ad hoc and not systemically, the priority is building the workflow that applies that capability at scale — not evaluating another AI tool.

Invest in AI literacy before AI tooling. The research is clear that the ROI of AI adoption is gated by how effectively the tools are used, not just whether they are adopted. An investment in prompt engineering skills, workflow design training, and systematic quality evaluation will generate more value than adding more AI subscriptions to a team that does not yet know how to use the ones they have.

The state of AI adoption in growth functions is genuinely encouraging in some areas and significantly overstated in others. The companies that are honest about the gap between their stated and actual adoption — and that invest in closing it through systems design rather than tool acquisition — will find themselves in a substantially stronger competitive position as the market matures.

GrowthBoxx offers AI Adoption Benchmarking as a service: mapping your current growth function against AI-native best practice and identifying the highest-priority systems to build. Book a discovery call to start with a growth stack assessment.