AI made execution cheap and connective skills scarce. Most teams are still built for the old economics.
Your martech sits idle for the same reason your team stays busy. Nobody has audited what they can actually do.
I've written before about the half of the average martech stack that never gets switched on. The features exist, the licences renew, and the capability sits unused. The explanation is almost never the software. It's that activating advanced functionality requires someone who can configure a platform, define a data taxonomy, design a lifecycle and interrogate the result - and most marketing teams are not staffed for that. They're staffed for channels and content.
Usage is not capability
The most useful piece of research I've seen on this measured marketers' AI skill rather than asking them to rate it. CXL's 2026 maturity benchmark found that only around one in ten marketers reached native-level capability - designing systems rather than executing tasks - while roughly a third had reached integration. The headline finding was the gap between frequency of use and measured skill: across workflow, research, analytics and operations, marketers used AI far more often than they were good at it. Daily use creates a false signal of competence, because the tool produces plausible output regardless.
That pattern shows up in what leaders report, too. Marketing Week's 2026 Career & Salary Survey found two-thirds of marketers see an AI skills gap in their team, and - for the second year running - marketing effectiveness ranked as the single biggest skills gap, ahead of strategy and research.
A brief word of caution however, since this category is noisy: a great deal of the AI-marketing statistics in circulation are aggregator-recycled and hard to trace. I've stuck to studies that name their method.
What AI actually took, and what it left
I'll be precise about this, because "AI changes everything" is not an operating insight.
What collapsed in cost: first drafts, variant production, resizing and reformatting, summarisation, first-pass analysis, research synthesis, basic segmentation logic, translation. Work that used to justify headcount now takes minutes.
What didn't - and got scarcer as a result: deciding what should be built. Configuring the platform rather than raising a ticket for it. Defining the data model everything else depends on. Designing a funnel rather than a campaign. Noticing that a number is wrong. Judging whether the output is any good, and having the standing to say so.
Notice the shape of that second list. It's all connective work - the joins between tools, between data and decisions, between activity and revenue. AI is excellent inside a task and useless between them. As execution got cheaper, the relative value of everything between tasks went up.
The shape has changed
The T-shaped marketer - one deep specialism, broad awareness across the rest - was a sensible model when breadth was expensive to acquire. Breadth is now cheap. What's scarce is depth in more than one place, and specifically depth in the joins.
The teams that are working well look less like a T and more like a comb: two or three genuine depths per person, sitting on a wide base of working fluency. That isn't a call for generalists. A generalist who can do a bit of everything shallowly is now the most replaceable person in the function, because that's precisely the band AI covers.
The six capabilities
This is the model I use when auditing a team. It deliberately isn't a channel list.
Systems fluency - can they configure the platform, or only request changes to it? The dividing line between a team that uses its stack and one that watches it.
Data and measurement literacy - can they define a taxonomy, build a report, and challenge a number that looks wrong? Not analyst-grade statistics; enough to not be fooled.
Funnel and lifecycle design - do they think in journeys and states, or in campaigns and sends?
AI direction and judgement - prompting is trivial and teachable in an afternoon. Knowing when the output is subtly wrong, and when not to use it at all, is the actual skill.
Commercial literacy - can they connect activity to revenue, build a business case, and hold their own in a conversation with the CFO?
Craft - writing, design judgement, taste. More valuable now, not less, precisely because AI commoditised the average version of it.
Most teams I assess score respectably on craft and poorly on systems, data and commercial literacy. That combination produces a function that makes good-looking things and can't prove they worked - which is exactly the position that gets marketing budgets cut.
What you can coach, and what you have to hire for
This distinction saves a lot of wasted training budget.
Coachable, reliably: systems fluency and data literacy. Both are procedural. Give someone platform access, a real problem, a competent person to ask, and eight to twelve weeks, and they will get there. Funnel design is coachable with reps - it needs live campaigns to practise on, not a course.
Partly coachable: AI direction. The mechanics take a day. The judgement takes exposure to enough wrong answers to develop suspicion.
Rarely coachable in the timeframe you have: commercial instinct and taste. You can improve both at the margins, but if nobody in the team has them, you hire for them or you borrow them. This is where a leadership gap is a hiring problem, not a training problem.
When training won't fix it
Three cases where I'd tell you not to spend the money.
When the structure is wrong. If the team is organised into channel silos with no owner for the joins, upskilling individuals won't produce connective work - there's nowhere for it to live. Fix the operating model first.
When there's no time to practise. Capability is built on live problems. A team at 100% delivery utilisation cannot absorb capability development, and sending them on a course anyway just produces certificates.
When the gap is at the top. If the leadership can't evaluate the work, the team's ceiling is set regardless of training spend. Gartner's 2026 research found a telling asymmetry here: most CMOs expect AI to transform their role within two years, while only about a third think their own skill set needs significant change.
How to run the audit
Score each person against the six capabilities on a simple three-point scale: can lead it, can do it with support, aware of it only. Then do the part most audits skip - verify against artefacts. Not courses completed, not self-assessment. Ask what they've configured, shipped, measured or challenged in the last quarter. Capability shows up in things that exist.
Then map the result against what the roles actually require. You're looking for two things: capabilities where nobody can lead, and capabilities where one person is the single point of failure. Those two lists are your hiring plan and your coaching plan respectively, and they're usually shorter and cheaper than the training catalogue you were about to buy.
The local picture
In the Australian market this matters more than it does in larger ones. Teams here are generally smaller, so single points of failure are the norm rather than the exception, and the marketing operations talent that would normally carry systems and data capability is scarce and expensive. For most teams, building that capability internally is more realistic than hiring it - provided someone senior has actually mapped what's missing first.
Frequently asked questions
What skills do marketers need in 2026? Six capabilities matter most: systems fluency, data and measurement literacy, funnel and lifecycle design, AI direction and judgement, commercial literacy, and craft. AI has reduced the value of routine execution and increased the value of connective work between tools, data and decisions.
Is the T-shaped marketer still relevant? Less so. The T-shape assumed breadth was expensive to acquire. AI has made breadth cheap, so the advantage has shifted to people with genuine depth in two or three areas - particularly in the joins between systems, data and commercial outcomes.
Can marketing capability be trained, or do you have to hire? Systems fluency, data literacy and funnel design are reliably coachable in eight to twelve weeks with live problems to practise on. Commercial instinct and creative taste are much harder to build quickly and are usually better hired or borrowed.
How do you audit a marketing team's capability? Score each person against the six capabilities on a three-point scale, then verify against artefacts they've actually produced in the last quarter rather than courses completed. Map the result to role requirements to find capabilities nobody can lead and capabilities with a single point of failure.
Why doesn't marketing training work? Usually because the operating model is wrong, the team has no capacity to practise, or the capability gap is at leadership level. Training individuals inside a structure that has no owner for connective work produces certificates rather than capability.
Sources: CXL Marketer AI Maturity Benchmark (2026); Marketing Week Career & Salary Survey (2026); Gartner CMO research (2026).
If you don't have the time, or the engagement levels just aren't there right now - I've got you. I built this Marketing Capability Assessment tool just for you. Shh, don't tell anyone.
Neil Collins is a marketing consultant based in Sydney, working with scale-ups and enterprise organisations on marketing capability, coaching and team design, alongside fractional CMO engagements and martech transformation. If your team is busy but the number isn't moving, get in touch.