AI will not fix your marketing team. It will expose it.

At the recent Accelerating AI conference, I spoke about what it now means to build a marketing team with AI.

Not a marketing team that has a few ChatGPT licences. Not a team that occasionally asks an AI tool to rewrite a social post. Not a team that has added a shiny new platform to an already crowded stack.

I mean a marketing team that is genuinely capable of using AI as part of how it thinks, builds, decides, publishes, measures and improves.

That distinction matters, because the buying journey is already changing faster than most businesses have noticed.

Google has said AI Mode has now passed one billion monthly active users globally. For many people, the old search journey is no longer the default route. They are asking longer questions, expecting interpreted answers, comparing options inside AI-generated summaries and, increasingly, letting assistants take action for them.

At the same time, products such as Gemini Spark and Microsoft Copilot are going to push AI away from passive response and towards always-on assistance, background work and delegated tasks. That changes the practical role of marketing. It changes how people discover brands, compare suppliers, assess trust and make decisions.

The uncomfortable bit is this: most marketing teams are still asking old questions.

What should we post?

Which keywords should we target?

What tools should we use?

How do we report on performance?

Those questions have not disappeared, but they are no longer enough.

The better questions are now:

What should we learn, create, automate, publish and improve today?

How do we appear in AI-generated answers and recommendations without chasing every new acronym?

Which parts of our marketing workflow should now work differently?

How do we spot issues and opportunities continuously, when the market is moving this quickly?

That is the real AI marketing challenge.

GEO is not a strategy

One of the topics I touched on in the talk was the current obsession with GEO, or Generative Engine Optimisation.

The idea is simple enough. If SEO is about appearing in traditional search results, GEO is about appearing in AI-generated answers, summaries and recommendations.

It matters. Of course it does.

If AI assistants are helping customers decide what to buy, which supplier to trust or which brand to speak to, you want to be visible in those answers.

But I am wary of the panic around it.

For years, SEO has mainly meant trying to influence Google. Now we have Google AI Mode, ChatGPT, Gemini, Claude, Perplexity and thousands of other models and applications interpreting information in different ways. Trying to reverse-engineer and manipulate all of them is not a serious strategy for most businesses.

The more useful answer is less fashionable, but far more durable.

Run a good business.

Give customers a reason to talk about you.

Publish useful, specific content.

Make your expertise visible.

Look after your reviews.

Build proof that exists beyond what you say about yourself.

AI systems are not only looking at your website copy. They are trying to interpret reputation, usefulness, consistency, authority and customer sentiment.

In plain English, what other people say about you matters.

So yes, think about AI visibility. But do not let GEO become another distraction from the basics you should already be doing properly.

An AI-capable marketing team is not a team that uses ChatGPT

This is probably the most important point from my talk.

An AI-capable marketing team is not defined by tool adoption. It is defined by how well people, data, tools, workflows, governance and measurement work together to create growth.

That means the team understands how AI can help, but also where it can cause damage. They know enough about how the technology works to spot opportunities and risks. They have enough permission to experiment, but enough structure to stop that experimentation becoming chaos.

Leadership matters here.

If you are running the business, your job is not to become the person who tries every single tool.

Your job is to make AI useful inside the business.

That means setting clear commercial goals, giving the team good data and context, identifying repeatable workflows, keeping human approval in the right places and building feedback loops so the work improves over time.

The organisations that get this right will not necessarily be the biggest. In many cases, smaller businesses have the advantage because they can move faster. They have fewer layers, fewer legacy politics and less internal theatre.

But only if they are disciplined.

Start with the problem, not the tool

At Qoob, one of the best examples of this is WatchTower.

We host and manage more than 80 websites. Every one of those websites has hundreds of moving parts: title tags, meta descriptions, alt text, schema, plugins, pages, content, performance signals and technical details that change all the time.

That creates a real operational problem. Human beings cannot reliably monitor every important data point across every site, every day. Not at the level needed to spot issues early and act quickly.

So we started with the problem.

What would it take to collect hundreds of website data points from every site, report them into one command centre, identify what needs attention and use AI to help prioritise the work?

The goal was not another report. Nobody needs more reports for the sake of it. The goal was actionable website intelligence that helps us act at scale for our clients.

That became WatchTower.

The important lesson is not that we built a tool. The important lesson is that AI helped us take a painful, valuable and repeatable operational problem and turn it into a working system.

That is where businesses should be looking.

Not, “What can we do with this new AI tool?”

Instead, “Which annoying, expensive or unreliable process could work better now?”

Agents need job descriptions

We have also been building internal agents at Qoob.

Alfred helps with internal marketing. Penny helps with finance. Other agents support other specific workflows. They have names and avatars, partly because it makes them easier for the team to understand and work with, but the naming is not the point.

The point is that they have specific jobs.

Alfred does not just sit there waiting for someone to say, “Write me a blog.” He researches, spots trends, plans ideas, writes briefs, drafts, designs supporting assets, reviews his own work and then asks for human approval. He carries context, memory, tone of voice, brand rules and previous feedback into the next task.

That is much more useful than a blank chatbot.

But it only works because the role is bounded. The context is managed. The process is defined. The human approval points are deliberate.

The temptation is to build one giant AI assistant that does everything. That sounds efficient, but in practice it usually becomes vague, unreliable and hard to control.

Specialist agents work better. Give them clear responsibilities. Give them the right context. Give them the right permissions. Make sure they ask before taking important action.

That last point is not optional.

If an agent can post to your LinkedIn page, access your Facebook account, change your website or read your financial data, then vague instructions become risky instructions. Security and permissions are not boring technical details. They are the difference between useful automation and avoidable damage.

Do not automate chaos

This is the bit a lot of businesses will be tempted to skip.

AI will not fix a broken workflow. It will accelerate it.

If your process is unclear, your data is poor, your responsibilities are vague and your approval steps are inconsistent, adding AI will not magically create order. It will just help the mess move faster.

That is why boring work is often the best place to start.

Choose something repeatable. Choose something with a clear commercial benefit. Choose something where success can be judged. Map the process. Decide where AI genuinely helps. Add the guardrails. Measure the outcome.

Then improve it.

Then repeat.

That approach is less exciting than chasing every new product announcement, but it is how you build something that lasts.

Keep the human in the loop

AI-generated work can be fast, impressive and occasionally brilliant. It can also be wrong, bland, overconfident or strangely off-brand.

That is especially dangerous in marketing.

If people can see that your content looks like AI or sounds like AI, the damage is immediate. The connection drops. Your brand starts to feel lazy, generic or automated in the worst possible sense.

Use AI to help you research, structure, draft, analyse and produce. But keep human judgement involved where brand, money, customers or compliance matter.

In practice, that means humans still decide what is worth saying. Humans still protect the voice of the business. Humans still judge whether the work is useful. Humans still stop the machine from publishing something that should never have left the draft folder.

The goal is not to remove people from marketing.

The goal is to give good marketers more leverage.

The real question

The wrong takeaway from an AI conference is, “I need to try that tool.”

You probably do need to try some tools. But that is not the strategic question.

The better question is:

Which part of our marketing should now work differently?

Marketing itself has not changed. The fundamentals are still the fundamentals. You still need to understand the market, know the customer, build trust, create demand, communicate clearly and turn attention into commercial action.

What has changed is how that work gets done.

The best marketers of the next few years will not just be campaign planners or content producers. They will be thinkers, builders, creators and architects. They will understand customers, but they will also understand systems. They will be able to design workflows, connect data, brief agents, judge outputs and improve the machine over time.

That is not a future problem. It is already here.

So start with one process. Not sixteen ideas. Not a giant AI transformation project. One specific, high-friction workflow that matters to the business.

Map it. Improve it. Add AI where it belongs. Keep the human judgement. Measure whether it works.

Then do the next one.

That is how AI becomes more than theatre. That is how it becomes capability.

If you want to look at where AI could make your marketing team faster, sharper and more commercially useful, book a consultation with Qoob.

Regards, 
Matthew