The question arrives in every planning meeting now, usually from somebody who doesn't own the budget. What's our AI strategy? And the honest answer, for most B2B marketing teams operating on flat resourcing, is three off-the-shelf tool subscriptions, a shared prompt document, and a quiet hope that productivity will materialise before the next target increase does.

I've watched this from an unusual angle. I grew up on DOS, hunting for a way into Windows so I could get at more programs, and I built Flash games at university before bringing that curiosity into a SaaS-focused marketing agency. Every wave of technology since has arrived with the same promise and the same catch.

The catch, in one statistic

According to Gartner's 2026 CMO Spend Survey, CMOs are now allocating around 15% of marketing budgets to AI, while only about 30% describe themselves as ready to scale those capabilities. Read the two numbers together and the shape of the problem appears: substantial spending, limited readiness.

And that gap isn't primarily a skills gap, whatever the vendor training decks suggest.

Andrew Ng, the AI thought leader and co-founder of Google Brain, framed the opportunity in terms that most marketing teams would happily sign up to.

"AI is the new electricity. Just as electricity transformed industries 100 years ago, AI is now poised to transform every major sector, from healthcare to manufacturing to B2B marketing. The ability to automate complex decisions and personalise experiences at scale is what will drive the next wave of growth."

Notice the two capabilities he names there: automating complex decisions, and personalising at scale. Both of them run on first-party data, and neither one works on data your organisation hasn't already organised.

Unrefined data is worth nothing

Clive Humby, the mathematician who coined the phrase, made this point long before the current wave and it has aged well.

Data, he argued, is just like crude oil. It's valuable, and it's useless in its raw state. It has to be refined into gas or plastics or chemicals before it drives any profitable activity, and data must likewise be broken down and analysed before it carries value. Data may well be the new currency; unrefined data is worth nothing at all.

Which brings the AI question back down to earth. Ask a generative tool to personalise your outreach at scale, and it will produce fluent, confident output built on whatever you happen to have. If what you have is a CRM carrying three competing source fields, a contact database where half the job titles are eighteen months out-of-date, and no account-level engagement history worth the name, the tool will personalise beautifully and inaccurately.

Faster wrong is still wrong, and it's a good deal more expensive.

Where the returns actually show up

The evidence here is unusually consistent, and slightly deflating. The Content Marketing Institute's research across 1,015 B2B marketers found the reported gains from AI collapse as you move down the value chain: 87% report improved productivity, 80% improved efficiency, and 39% any improvement whatsoever in content performance.

The tools reliably make you faster, then. Whether they make you better is a separate question, and the current answer is mostly no.

That pattern points somewhere useful, though. The returns concentrate in the structural, day-to-day work: predictive lead scoring, segmentation, campaign variants, meeting summaries, first-pass research and workflow automation. Platforms such as HubSpot's AI features, Marketo Engage and Salesforce Einstein have been doing behavioural recommendation and lead scoring for years, and that's where the hours genuinely come back.

The returns thin out sharply wherever the output needs a point of view.

The three things worth doing before buying anything else are set out at the foot of this piece.

The uncomfortable version

Here's what I suspect is happening in a lot of organisations right now. AI has become a convenient answer to a resourcing question that nobody especially wants to answer honestly.

A four-person team cannot deliver the plan of a nine-person team, and saying so in a planning meeting is politically expensive. Committing to close the gap with AI is free, plausible, and defers the conversation by two quarters. Then the gap doesn't close, since the constraint was never typing speed in the first place.

Where this leaves you

None of this is an argument against the tools. AI amplifies what a small team can do, and the productivity findings are real enough. But it amplifies whatever is already in place: your data quality, your positioning clarity, and the day-to-day coordination between your teams.

So the practical starting point isn't a tool evaluation. Take one workflow you were planning to automate, and trace the data it depends on all the way back to its source system. If you can't establish where a field comes from, who maintains it, and when it was last verified, you've found the actual project, and it's the one that makes every subsequent tool decision cheaper.

Which brings this series back to where it started. Prove the contribution, understand the buyer, name the accounts, say something distinctive, grow the base, and do it all with what you have. The technology changes every eighteen months. That list doesn't.

Related listening on SaaS Stories: Data That Doesn't Lie, brand tracking for the AI era