When software starts thinking and acting for marketers
There is a shift happening in software that I think marketers should pay attention to.
The way we interact with software is changing.
For decades, software has essentially worked like this: we tell the software what to do , the software does it , we look at the result.
We click. We search. We filter. We configure. We approve.
Even SaaS, which transformed business software, largely kept this model intact. The software became easier to access and more powerful but we were still the ones operating it.
Agentic software changes that.
Instead of telling software how to do something, we tell it what we want to achieve. And the software figures out the steps.
That sounds like a small difference. It isn’t.
From tools to operators
I’ve been in the position of a brand marketer preparing for a new product launch. Along with my team I was using:
one tool to analyse my audience
another to research competitors
another to build a media plan
another to generate creative
another to manage the CRM
another to monitor the performance
We move between systems, interpret the information, make decisions and tell each system what to do next.
Now it looks more like this:
‘‘I’m launching this product in 3 markets. Identify the highest potential audiences, analyse what competitors are doing, propose a positioning, develop three campaign territories, build a launch plan and tell me what I should test first.”
An agentic system could potentially break that request into tasks, use different tools and sources, analyse what it finds, generate outputs, evaluate them and come back with a recommendation.
It can generate a lot of content and you have to use your experience, analysis and judgement skills.
But the important point here is that it can move through a problem. That is a very different from before…and a massive help.
The difference between AI and agentic software
We’ve become used to AI that helps us do things faster.
Ask ChatGPT to write a brief.
Ask an AI tool to summarise research.
Ask it to analyse a spreadsheet.
Ask it for 10 campaign ideas.
This is useful but I’m still the operator.
I decide what to ask and what to do next. I move the work from one tool to another. I make the final decisions.
Agentic software starts to break that sequence.
You give it an objective and the constraints. It can determine a plan, take actions, inspect the results and adjust.
Traditional software relies on decision logic written in advance, while agentic software can generate and adapt its decision logic at runtime. That’s insane.
In other words, the agent isn’t simply a feature inside the software. The agent becomes the software.
For us marketers, this distinction matters.
Because marketing is full of work that isn’t really a sequence of fixed rules.
It’s messy, it’s contextual, it involves a lot of human judgement. And there are usually many possible options to choose from.
Marketing is particularly suited to this shift
Consumer marketing has always involved connecting dots.
Consumer behaviour, culture, data, brand, technology, media, commerce, creative.
In my view, the interesting problems sit between those things.
That’s also where traditional software starts to struggle.
A dashboard can tell you that conversion is falling.
It can’t necessarily understand that a competitor has changed its positioning, a cultural conversation has shifted consumer sentiment, your creative is becoming repetitive and your distribution strategy is poorly executed and then decide what to investigate next.
An agent potentially can.
Not because it has magical marketing intuition. But because it can work across multiple sources of information and take a sequence of actions without requiring a human to manually orchestrate every step.
That is where I think the opportunity gets interesting.
The marketing stack may become less important than the marketing system
I’ve spent years signing off and using martech stacks.
CRM, CMS, analytics, media platforms, social tools, commerce platforms, research tools etc they all solve a particular problem
Then we become the integration layer and the API - we carry information from one system to another, reconcile different views of the customer, brief, report, coordinate, check, chase etc it’s so time consuming and complex.
Agentic software could change that if done well
We describe properly the outcome we want and let an agent orchestrate the underlying systems. So the interface becomes less important and the outcome becomes more important.
In fact, you don’t need to know exactly how the work gets done. You specify the outcome.
However as marketers we can’t hand everything over to agents. The more autonomous software becomes, the more important human judgement becomes.
An agent can optimise for a goal but we need to define the goal, decides what is acceptable, understands the brand, knows when a technically correct answer is strategically wrong, decides whether an idea feels culturally relevant rather than merely statistically likely to perform etc
And keep leadership, control and responsibility over the agent - very important.
The role of the marketer may therefore shift from designing the intent and operating tools to designing the intent only. So less workflows and more intelligence. Yay!
Perhaps we’ll see the end of a lot of marketing work that exists because our software is fragmented - endless reporting, manual research, repetitive optimisation, movement of information between platforms, task coordination.
The work that is necessary, but doesn’t really require so many human brains.
That could free us to spend more time on the things that actually create differentiation and excitement: understanding people, spotting cultural shifts, defining strategy, shaping brands, making choices and having taste.
There is a useful irony here.
As software becomes more autonomous, the distinctly human parts of marketing may become more valuable.
Because when everyone can generate 10 campaign ideas in a second, the advantage isn’t having more ideas.It’s knowing which one matters.
What I would be watching
For consumer brands, I think the interesting question is “Which parts of our marketing system should become agentic?” rather “How do we add AI to our marketing?”
Start with the workflows where:
there is a clear outcome
the process involves multiple steps
information sits across different systems
there is a lot of repetitive coordination
decisions can be evaluated
and human judgement can remain at the right control points
Then go and experiment with the ambition of redesigning how the organisation works.
The real promise of agentic software is that software may finally be able to do more of the work between the brief and the outcome. and fast.
And if that happens, we won’t just have better marketing tools. We will have to rethink what a marketing organisation actually is.