There are two unhelpful positions on AI in marketing. One says it changes everything, the other that it changes nothing. The useful truth sits between them, and the data now makes that unusually easy to demonstrate. Three-quarters of marketers now use at least one form of AI, according to Salesforce’s State of Marketing 2026, a survey of 4,450 marketing professionals across 26 countries. Yet McKinsey’s research finds only 1% of leaders describe their organisation’s AI deployment as mature, and S&P Global Market Intelligence’s 2025 survey of over 1,000 enterprises found 42% had abandoned most of their AI initiatives, up from just 17% the year before.
So nearly everyone is using it, and very few are getting sustained value from it. That gap is the whole story. Used well, AI is a genuine efficiency gain. Used badly, it produces a lot of forgettable content very quickly. Salesforce’s own headline finding says as much, noting that most marketers who’ve adopted AI are still using it to send one-way, generic campaigns. Here’s a measured read on where it actually helps and where it doesn’t.
What AI is genuinely good at
- Getting past the blank page. As a starting point for drafts, outlines, and ideas, it’s fast and useful, provided a human shapes and finishes the work.
- The repetitive middle. Summarising, reformatting, repurposing one piece of content into several, tidying data. These are the tasks that quietly eat your team’s time, and they’re where the efficiency gains are real. Salesforce found its highest-performing marketing teams reclaimed around eight hours a week by putting AI to work on exactly this. If you want the practical detail on where to start, our no-nonsense guide to using AI in marketing covers exactly that.
- Analysis at speed. Spotting patterns across data faster than a person would, which frees your people for the judgement that follows.
- Scaling without losing the thread, if it’s guided by clear direction and a human editor. That condition matters more than the tool does. McKinsey’s research consistently finds the organisations seeing real returns are those that redesigned how they work around AI, whilst those bolting it onto existing processes see far less benefit. What separates the two groups isn’t the technology. It’s the thinking around it.
What AI can’t do, and shouldn’t be asked to
- Own your judgement. What to say, to whom, when, and whether it’s right for your brand. That’s strategy, and it stays with people.
- Understand your business the way you do. It doesn’t know your clients, your market, or your values unless a human brings that to it. Left alone, it produces content that’s competent and completely generic.
- Build relationships. Marketing in businesses like yours runs on trust, and trust is human. It’s also fragile around this technology. Salesforce’s research shows customer trust in businesses to use AI ethically has fallen to 42%, down from 58% in 2023. AI can support the work around a relationship, but it can’t be the relationship. For organisations where trust is the entire asset, we’ve written separately about how purpose-led organisations can use AI without compromising it.
- Guarantee it’s right. It’s confidently wrong often enough that everything it produces needs a human check, and experienced practitioners already know this. Around 78% of marketers apply moderate or extensive editing to AI-assisted content before publishing it (Sociality.io, 2026). The ones who don’t are how brands embarrass themselves.
The sensible way to adopt it
Treat AI as capable assistance, not a replacement for thinking. Keep a human at both ends, with direction going in and judgement and editing coming out. In our experience, the firms getting real value from it are the ones using it to free their people for higher-value work, not the ones hoping it’ll let them stop paying for judgement. The abandonment figures above tell you how that second approach ends. When 42% of companies walked away from most of their AI initiatives last year, the common thread was adoption without strategy.
Adopted with a clear head, AI becomes a meaningful part of digital marketing transformation, delivering efficiency that gives you back time for the strategic and human work that actually differentiates you. Adopted lazily, it just helps you produce mediocrity faster. The technology is the same. The difference is entirely in how you use it.
If you’re working out where AI genuinely fits in your marketing, and where it doesn’t, that’s a conversation we have with clients regularly as part of our digital transformation services. Let’s have a straightforward, no-jargon conversation about it.
Source list
- Salesforce, State of Marketing 2026 (adoption figure, generic campaigns finding, eight hours reclaimed by high performers) – https://www.salesforce.com/news/stories/state-of-marketing-2026/
- McKinsey, Superagency in the Workplace (1% maturity) – https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work
- McKinsey, State of AI research (workflow redesign point, referenced without link in the body) – https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai (standard location, confirm it resolves before use)
- Salesforce marketing statistics page, citing State of the Connected Customer (trust decline) – https://www.salesforce.com/marketing/marketing-statistics/
- S&P Global Market Intelligence 2025 enterprise survey (42% abandonment) – widely reported but no public press-release URL confirmed. Source the S&P page or a high-authority report before linking externally, or leave unlinked as now.
- Sociality.io 2026 survey (78% editing figure) – corroborated across roundups but primary not yet confirmed. Verify before publishing, or cut the citation and soften to ‘most marketers’.
Publishing actions
- Action the agreed redirect or rewrite of the old ‘Is It as Intelligent as We Think?’ post into this piece
- Add inbound links to this article from the no-nonsense AI guide, the AI tokens article, and the Digital Transformation page
- Meta title and H1 match by design, but there is room to test a SERP variant if wanted
