AI in marketing: how to move from AI tools to AI systems

How do you bring AI into marketing?

You bring AI into marketing by building it into specific workflows, feeding it your brand and keeping human judgement in charge. Handing people tools and hoping is not enough. Adoption is no longer the hard part. According to McKinsey, 78 per cent of organisations use AI in at least one business function, and marketing and sales is where generative AI is used most (McKinsey, March 2025).

The hard part is results. Gartner found that only five per cent of marketing leaders who use generative AI solely as a tool report significant gains in business outcomes (Gartner, 2025).

At IW Agency, we think the gap sits between tools and systems. A tool writes a caption when someone asks for one. A system knows the brand, sits inside a workflow and answers to a person who decides. We break a system into three parts:

• Source: what the AI knows about your brand.

• Flow: where the AI sits in the work.

• Judgement: who decides what ships.

Conclusion: A tool saves minutes. A system changes outcomes.

What do you need before you start?

Before you start, you need three things: a brand clear enough to brief a machine, an owner for each workflow and permission to get the first version wrong. Most marketing teams already have the licences. Fewer have the rest.

1. A written brand platform.

   • Document the positioning, tone of voice and proof points so that an outsider could follow them.

   • Settle any internal disagreement about what the brand stands for. An AI system will simply average the argument.

   • If yours is outdated, start with what a brand platform is and when you need a new one.

2. A named owner.

   • Make one person accountable for each workflow, not a committee responsible for "AI".

3. Room to experiment.

   • In a Gartner survey of 402 CMOs, 80 per cent said staff fear and anxiety are a barrier to experimenting with AI (Gartner, 2026).

   • Say out loud that early versions will be imperfect.

How do you choose the first workflow?

Choose a workflow that repeats often, has a clear output and carries the brand. That way the gain is visible and the lesson is real. We would start where volume meets brand, and leave the brand idea itself with people.

1. List the repeat work.

   • Map what the team produces every week: briefs, adaptations, product copy, reports.

2. Score each task.

   • Rate each task on volume, on how measurable the output is and on how much brand it carries.

3. Pick one.

   • Five pilots teach you less than one done properly.

4. Record today's baseline.

   • Write down the time, the cost, the number of revision rounds and the brand owner's view of quality.

A typical first candidate is a food brand adapting one campaign across ten markets and three formats. The work is repetitive and on-brand, and it is easy to measure.

How do you give the AI system your brand?

You give an AI system your brand by turning the brand platform into material the system reads every time. That material covers the rules, the examples and the things you never say. This is the source. Without it, the system writes like your category's average, because the average is what it has seen most.

Consider rye bread. For a long time it was a category defined by muted tones and functional messaging, and Hatting set out to break those conventions. Ask a generic model to write about rye bread and you get muted tones and functional messaging back.

1. Translate the platform.

   • Turn positioning and tone into instructions: what we believe, how we sound, which words we use and which we avoid.

2. Add examples.

   • Include ten pieces of work you are proud of and three you would never publish, each with a line on why.

3. Name the feeling.

   • Machines handle logic well. The emotion the brand should leave behind has to be spelled out.

4. Version it.

   • Keep one owner, one current version and a date.

Conclusion: Feed it the average and you get the average.

How do you redesign the workflow around AI?

You redesign a marketing workflow by deciding, step by step, what the system does, what a person does and where the work changes hands. This is the flow. Adding AI to the old process mostly makes the old process faster. McKinsey describes a range of AI use. At one end, a few employees experiment. At the other, AI is built into business units whose processes have been entirely redesigned (McKinsey, March 2025).

Integration is where most teams stall. Salesforce, which sells AI agents itself, found that full integration is still a work in progress for 61 per cent of marketers. Its survey also found that only 26 per cent are completely satisfied with their ability to connect customer touchpoints (Salesforce, 2026).

1. Draw the current process.

   • Map every step, hand-off and approval.

2. Assign each step.

   • Label each step as machine drafts, human decides or human only.

3. Connect the inputs.

   • Give the system the brief, the brand source and the data it needs, in one place.

4. Remove dead steps.

   • Cut the steps that only existed to move files between people.

In the campaign adaptation example, the system drafts every version from the master brief. A market lead reviews the drafts, and the brand owner signs off on a sample rather than on every file.

How do you keep human judgement in charge?

You keep human judgement in charge by deciding in advance which calls a person always makes: the idea, the feeling and anything published in the brand's name. This is judgement, and it is where we would spend most of the senior time that AI frees up. Gartner reports that marketing leaders expect AI-driven automation of marketing work to more than double, from 16 per cent in 2026 to 36 per cent by 2028 (Gartner, 2026). The more the machine does, the more each remaining human choice matters.

We believe feeling beats logic, and machines are strongest at logic. When Google wanted a presence at Almedalen as AI increasingly shaped public debate, the aim was a more human perspective on technology. The same principle applies to your output.

1. Set red lines.

   • Decide what never ships without a person: claims, prices, sensitive topics and the core idea.

2. Name the approver.

   • Put one name on each red line.

3. Review for brand, not only for errors.

   • Ask "does this feel like us?" as well as "is this correct?"

Conclusion: Automate the volume. Keep the taste.

How do you know it's working?

Your AI system is working when the baseline moves and the brand gets sharper, not just cheaper. McKinsey found that 80 per cent of respondents say their companies set efficiency as a goal for AI. The companies getting the most value often set growth or innovation as additional goals. Only 39 per cent report an impact on EBIT at the enterprise level (McKinsey, November 2025).

Signs that it works:

1. Time and cost fall against the baseline you recorded.

2. Revision rounds fall because first drafts already sound like you.

3. The brand owner stops correcting tone and starts debating ideas.

4. Freed time goes to growth work such as new concepts, tests and markets.

The warning sign is more content with the same results.

What does this mean for your brand?

• Write the brand down before you automate it. A vague platform produces generic output at scale.

• Start with one workflow and a baseline. You cannot prove a gain you never measured.

• Set growth goals, not only savings. Efficiency alone rarely shows up in the results.

• Protect the feeling. Keep people in charge of the idea and the emotion.

• Make it a leadership job. As Gartner's Kristina LaRocca-Cerrone puts it: "The CMOs who win won't just manage AI. They will integrate it into how marketing leads the enterprise."

Frequently asked questions

What is the difference between AI tools and AI systems in marketing?

An AI tool performs one task when someone asks for it. An AI system is built into a workflow, draws on the brand's own material and has defined points where a person decides.

Which marketing tasks suit AI first?

Repeated tasks with a clear output suit AI first. Examples include campaign adaptation across markets and formats, product and packaging copy variations, and reporting. The brand idea should stay with people.

Will AI make brands sound the same?

AI will make brands sound the same if every brand gives it the same generic input. A system that runs on a specific brand platform, reviewed by people who know the brand, can hold a distinct voice across far more output.

Do we need AI agents?

You do not need AI agents to start. Salesforce found that 13 per cent of marketers currently use agentic AI (Salesforce, 2026). Agents make most sense once source, flow and judgement are clear, because an agent acts on whatever brand knowledge it is given.

From tools to systems

The tools are already on every desk. What separates the teams that see results is the system around those tools: a clear brand, a redesigned workflow and people who decide what ships. If you want to talk about where to start, write to us at hello@iwagency.com.

Sources

• The state of AI in 2025: Agents, innovation, and transformation — McKinsey & Company, 2025

• The state of AI: How organizations are rewiring to capture value — McKinsey & Company, 2025

• Gartner Survey Finds 65% of CMOs Say Advances in AI Will Dramatically Change Their Role in the Next Two Years — Gartner, 2025

• Gartner Survey Reveals Marketing Leaders Expect AI Automation of Marketing Work to Double by 2028 — Gartner, 2026

• State of Marketing, 10th Edition — Salesforce, 2026