Systems thinking in marketing

01

From personal productivity to company-level benefits

One reason is the gap between the personal productivity gains we are getting from AI and the company-level benefits.

McKinsey published some interesting numbers on this recently. In its 2026 global AI survey, 80% of respondents said AI had improved their individual productivity, while only 37% attributed any EBIT impact to their organisation’s use of AI.

Let’s say you are now producing content five times faster than before. The workflow around that content might still look something like this:

02

Look beyond the task you are optimising

If the approval process still takes three days, producing the first draft in 15 minutes instead of two hours may not change very much. You may simply have created a bigger queue somewhere else.

For me, systems thinking in marketing means looking beyond the task you are optimising and paying attention to what happens around it.

In marketing, we have spent quite a lot of time optimising individual parts. How can we create content faster? Improve the conversion of a landing page? Improve a newsletter’s open rate? Generate more leads?

All these are valid questions.

But increasing the number of leads doesn’t help much if sales cannot follow them up. Producing ten times more content is not necessarily useful if you don’t have a sensible way to distribute it, measure what works, and use that information when planning the next content.

Sometimes improving one part of the process can simply move the problem somewhere else.

03

Why AI agents make systems thinking more relevant

A lot of AI use so far has been task-based via a chat interface: write this, summarise this, analyse this, draft this, or create an image, video or animation.

I still use AI for these kinds of tasks every day.

With agents, the scope is getting larger. AI can use tools, gather information, and carry out several steps of a process, and we are starting to delegate longer pieces of work instead of individual tasks.

Jim Lecinski from Northwestern University’s Kellogg School of Management wrote about this in a Think with Google article called Agentic AI means marketing leaders must now be “systems thinkers”.

He suggests identifying a small number of important marketing workflows and mapping them properly: where the workflow starts, what information it needs, what decisions are made along the way, where the bottlenecks are and where people need to be involved.

I think that is a useful way to approach agentic AI as well.

If we take the way we work today and add AI to individual steps, we will probably get efficiency gains. There may be much bigger opportunities in looking at the workflow itself.

Agents also become part of the system themselves. When an agent takes over a step and passes the result on to a colleague, that handover needs as much thought as one between two people.

04

The lines between roles and functions are blurring too

I’ve noticed this happening in marketing over the past few years.

With generative AI, you can describe what you want and the AI produces it. It started with text, so people who were not writers got tools that helped them produce reasonably good content. Then the same became possible with images, video, and audio.

Someone working in marketing can now analyse data, create content and images, make videos, do extensive research and build simple applications. The same is happening outside marketing. People in sales, product development and other functions can do things that previously belonged quite clearly to another team or profession.

The Cybernetic Teammate research at Procter & Gamble is interesting from this perspective.

Commercial and R&D professionals worked on real product innovation challenges, either individually or in teams, and with or without AI.

Without AI, the R&D professionals tended to produce more technically oriented solutions and the commercial people more commercially oriented ones. With AI, people from both backgrounds produced more balanced solutions.

In these product innovation tasks, individuals working with AI also produced solutions of roughly the same quality as two-person teams working without AI.

The experiment was done in 2024 using GPT-4 or GPT-4o, which are old models and capabilities from today’s perspective.

I don’t take this to mean that specialist expertise is disappearing or that everyone can suddenly do everyone else’s job. It does suggest that the boundaries around our jobs may become less rigid. And when we increasingly work across those boundaries, understanding the whole workflow becomes more useful.

05

How to get started with systems thinking in marketing

Pick one marketing workflow that matters and map what actually happens today.

If a podcast is an important part of your marketing, for example, your current workflow might look something like this:

Then start looking at the workflow more closely.

1. 01

   Where does the information for each step come from?

2. 02

   Where do people wait for somebody else?

3. 03

   What is still copied manually from one place to another?

4. 04

   Which decisions are made along the way?

5. 05

   Where do you need human judgement?

6. 06

   Which parts could AI help with?

7. 07

   Which parts could potentially run with very little human involvement?

8. 08

   And what happens with the information at the end?

If certain podcast topics consistently attract the right audience, does that affect what you create next? If customers repeatedly ask the same questions in sales calls or customer service, does that information find its way into your content? If sales knows that a certain type of lead is consistently poor, does marketing know it too?

06

Understand the work before building an agent

Once you map a workflow properly, you may also notice that AI is not always the answer.

Maybe you can simply remove one approval. Maybe information should be stored differently. Maybe two people are doing overlapping work. Maybe a handover has never really had an owner.

At ViaNueva, we run a number of agents in our own work and keep building new ones. Before building an agent, it is worth understanding the work the agent is supposed to become part of.