You’ve been working in AI for around three years now, and a lot seems to have changed in that time. We’ve gone from marketers trying to understand what AI even is, to claims that it’s now “embedded” everywhere. What’s the reality - myth or maturity?
I don’t think it’s embedded at all. The stat I saw recently was that fewer than 5% of organisations have embedded AI effectively. There’s so much misinformation online and a lot of hype. There are a lot of vibes around what AI can do.
The reality is you can’t just give people these tools and expect good work back. They don’t work like that. The people who get the most value have spent a lot of time understanding how the technology works and testing it.
Where we are now is the proper implementation phase. That means technology, infrastructure, and change management. It takes time for people to get used to using these tools. It’s a new way of working and it signals a new wave of marketing.
Many organisations feel pressure to “get AI” without knowing where to start. What tools would you actually recommend for marketers beginning this journey?
The tools I personally get the most value from are Claude, as a strategic partner and for content creation. Perplexity, which is excellent for research - that’s what it’s built for. Gemini, where Google has really stepped up in the last six months with 2.5 Pro and their broader suite. Those three are a strong starting point.
And what should marketers avoid?
This might be a hot take, but I’m cautious about anything labelled as an agent in marketing.
An agent is essentially another layer of software sitting on top of a large language model. A true agent makes decisions autonomously, without a human in the loop. In marketing, that strips out innovation and nuance.
Because these models are predictive they guess what comes next - agentic AI risks accelerating more of the same ideas instead of creating new ones. In other disciplines, agents can work well. But in marketing, creativity matters.
Do you think marketers are ready to use multiple LLMs for different purposes?
It took me about four years of curiosity to get to the point where I can confidently move between tools. Each one has different quirks and behaviours.
GPT is more subservient, it does what you tell it. Claude is more inquisitive and asks better questions. But getting comfortable takes time and curiosity. Most marketers aren’t there yet.
There’s also a tendency to focus purely on content. Should AI be doing more than that?
Absolutely. Content shouldn’t be the sole purpose of AI. It should be workflows and processes.
From a business perspective, the first question shouldn’t be “what tools should we use?” but “what value are we trying to create?” Every business is different. AI should support an objective, not exist for its own sake.
Are organizations actually seeing ROI when AI is implemented properly?
Yes, when it’s done well. Businesses using AI effectively are cutting acquisition costs by around 50% and improving revenue by 10–15%. That can mean a 20–30% increase in ROI.
But that only happens when AI is used as an extension of the team, not a replacement. Too many companies took the “cheap” route last year: giving everyone ChatGPT and hoping for the best. That’s not actually cheap when you factor in wasted time and poor outputs.
There’s also confusion between individual AI use and enterprise-level AI. How big a problem is that?
It’s huge. AI has been treated as a catch-all. There’s personal AI, helping individuals with ideation, decks, analysis and then there’s technical AI, the actual builds and systems. They’ve been lumped together as if AI can magically solve everything.
That oversimplification causes a lot of frustration.
Looking ahead, where should marketers actually be experimenting next?
My advice is simple: try every task with AI first and assess the output. Even when it doesn’t work, you learn something.
I’m sceptical about AI-generated video and voice. They often feel dishonest and are easy to spot. I don’t mind people being transparent about using AI as long as there’s human oversight.
What excites me more is predictive AI: spotting where markets are heading, identifying emerging interests, and shaping messaging or even products proactively rather than reactively.
Many teams are already “bringing their own AI” into organisations. What should CMOs do about governance?
If you block AI entirely, people will just use it anyway and that’s riskier. Without training and governance, you lose control completely.
I know plenty of people who pay for their own AI subscriptions because their company won’t allow it. They work faster and get better results but without oversight.
The better approach is enablement with guardrails.
Final question: will roles like “prompt engineer” or “Head of AI” still exist in a year?
I think we’ll still see Heads of AI, but not prompt engineers. Prompting will become business as usual.
Watch the full interview on the B2B Marketing United YouTube channel.