The pressure to use AI is there. The structure isn’t.
Ask almost any marketing leader and they’ll tell you the same thing: the pressure to bring AI into the team is real. What’s usually missing, as Nic explained to host David Wain-Heapy of Prodigi, is the organisational design to make it work.
Take one CMO at a well-known IT services company, a business that sells AI consultancy to its own clients. They were told to use Copilot and figure it out, with no support from the business and no headspace to learn anything complex.
It’s a pattern Nicola sees again and again. AI adoption happens person by person, rather than across the marketing function as a whole. So what does the alternative look like?
What AI in account-based marketing looks like done well
The best example we’ve come across was shared at one of our recent roundtables. A large global infrastructure company spent two years mapping every process involved in building an ABM strategy and go-to-market plan, then built an agent for each step.
Those agents live in Copilot Studio, looked after by a central ABM excellence team that checks them constantly for accuracy and hallucination. Field marketers around the world use them to get to market faster, with insight they know they can trust.
And nobody lost their job. The company is spending more, not less, but it’s moving much faster, and the people who used to do the work manually are now the experts building and managing the agents.
What struck us most was that the checking never stops. Even an RFP evaluator agent trained on more than 200,000 RFPs still has its output validated all the time.
Why individual AI habits don’t scale
That level of central oversight can sound like overkill until you’ve seen the alternative. We have, and Nicola is happy to put her hand up about it.
A team member who has since left built a Claude agent in their personal account to produce monthly reporting graphs. They handed over the prompt, but when a colleague ran it on the same data, the result came out different.
A shared project in our company Claude account would have solved it. Even well-defined prompts can drift, so if you want consistent results, the setup has to belong to the team, not the individual.
AI can spot patterns, but it doesn’t know your strategy
Consistency is one challenge. Context is another. AI is brilliant at finding patterns and building charts, but it doesn’t know your strategy, your objectives or what’s happening in your market, which is why Nic wouldn’t rely on it for analysis.
Say your cost per click rises. AI might tell you to change your bid, while an experienced marketer knows something else is going on, perhaps a heatwave affecting demand. Follow that recommendation blindly as a junior marketer and you can end up somewhere painful.
Where AI really earns its keep is in helping you ask better questions. Querying a dashboard to find out why something is happening surfaces issues faster than you’d spot them yourself. For next best action, especially on new year-long programmes with nothing to compare against, a little caution goes a long way.
A practical use: measuring ABM before pipeline appears
Used with that kind of judgement, AI can help with one of ABM’s oldest headaches. Pilots are often reviewed at six or nine months, yet with sales cycles of nine to 12 months, opportunities may not have appeared yet. That makes a £100k to £150k programme very hard to justify on quarterly numbers alone.
One roundtable attendee got around this by using AI to scan call notes in the CRM, pulling out positive conversations and intent signals across target accounts. It gave them a read on account momentum that engagement data alone couldn’t, and it’s one of the most practical uses of account-based AI we’ve seen.
If that sounds familiar, it’s because many of our clients already do something similar with Gong, analysing call transcripts for sentiment, churn risk and sales coaching.
The AI cost problem nobody can budget for
Of course, all of this has to be paid for, and that’s where things get tricky. Consumption pricing makes AI very hard to plan for.
Nicola found this out first-hand when she built a simple landing page in Lovable. Every prompt cost a different amount, and her credits disappeared quickly. If you can’t say what a landing page will cost, how do you budget for one?
It’s the same story at enterprise scale. Salesforce Data 360 is sold in tokens, with different queries consuming different amounts and no clear monetary value attached. One client moved data unnecessarily while working in a sandbox and burned through 15 million credits.
As platforms push further into agentic features, more of what marketers do will be consumption-led, and some teams will simply choose not to do things. That’s why any plan for account-based marketing AI use needs a view of consumption costs, not just licence fees.
Vibe coding is changing build versus buy
Cost isn’t the only thing reshaping decisions. AI is also making it easier for businesses to build their own tools, sometimes faster than anyone expected.
We saw this with a global professional services firm of around 5,000 people. It was 14 months into a CRM implementation, with a systems integrator, one of our consultants embedded five days a week and more than 100 stakeholders involved in design.
Then, over a bank holiday weekend, two senior directors vibe coded much of the solution design themselves. The following week, leadership paused the project to explore building its own system, which now sits with IT to manage security, compliance and data.
How far businesses go down this route will come down to data governance. The infrastructure company from our roundtable, by contrast, has ring-fenced all contact and email data, keeping it well away from AI.
Clients want to own production, with support on strategy
Put all of this together and you can see why the relationship between brands and their partners is changing. Cost pressure and growing maturity mean many clients now bring production in-house, often by their third year with a partner, and aim to scale it with AI.
What they want from us is different: strategy upfront, oversight and support while they learn, help choosing the right tools, and visibility at the end. It’s why we’ve spent the last three years moving towards strategic consultancy, across go-to-market, ICP and the structure and infrastructure behind marketing.
Clients come to us when they’re doing something for the first time. With AI moving this quickly, there will always be something new.
Stay relevant by staying curious
Which brings us to Nic’s closing advice, for businesses and people alike: stay relevant, keep learning, stay curious and make space for yourself.
For AI in account-based marketing, that means building shared, governed systems, keeping human judgement in the loop and knowing what it will cost before you scale.
You can hear the full conversation on the Aperture podcast. And if joining up your data is next on your list, Adrianne Haynes Samuel’s post on the B2B buyer journey is a good place to start.Â
