I sat down with Mattia Santin to talk about what marketers actually need to do differently now that AI is part of the everyday workflow. He came out of it with seven ideas and a take on each. Here they are, with mine alongside his.
Mattia is a friend and a former colleague, which means the conversation skipped the warm-up. Neither of us was interested in whether AI is going to change marketing. It already has. The interesting question is narrower and more awkward: what do you personally have to do differently on Monday?
What follows is his outtake of that conversation. Seven ideas, my view on each, and his take underneath. We agree more than we disagree, but the places where he sharpens my point are the places worth reading twice.
1. Hire for curiosity, not just technical depth
AI can bridge most technical gaps now. What it cannot do is keep asking better questions. So I hire for inquisitiveness and creativity — for people who finish a project and immediately ask what else can we do?
Mattia's take: Curiosity becomes disproportionately valuable. People who challenge assumptions and keep learning are hard to replace.
2. Start with a clear business objective
AI makes it easy to do too many things at once. Start with the outcome and work backwards. My current test is simple: does this generate sourced demos, or just vanity metrics?
Mattia's take: AI is not the goal. Growth is. Define the business target first, then use AI to reach it faster or in a more differentiated way.
3. Experiment constantly
My team can now build pages, test SEO ideas, spin up demos and produce content without waiting on multiple specialist teams. That should dramatically increase testing. If your output looks the same as it did last year but arrives sooner, you have bought a discount, not an advantage.
Mattia's take: AI should not only make the old workflow cheaper. It should increase the number of experiments, and therefore the learning a team generates.
4. Speed is a competitive advantage
Just do it. Ship, make mistakes, learn and keep moving. AI makes quality work much faster, so learning velocity matters more than perfection.
Mattia's take: An 80 percent solution live today teaches you more than a perfect solution weeks later. The loop that matters: ship, learn, adjust.
5. Enable everyone, then share what works
When someone on my team discovers a useful AI workflow, we teach everyone immediately. New capabilities should spread, not sit with one specialist.
Mattia's take: AI knowledge should not live with one designated AI person. The compounding happens when the whole team applies the same capability to different problems.
6. Connect AI to the data that makes you different
My biggest step change came from connecting AI to product data, customer questions, CRM records and real sales conversations. Proprietary knowledge creates better marketing. A model with no access to your business can only give you the median answer, and the median answer is what your competitors are getting too.
Mattia's take: Generic models have generic knowledge. Your own data is the advantage, but only if it is connected and trustworthy. Messy data creates messy answers faster.
7. Culture is the prerequisite for everything above
The first six only work when people have permission to experiment, connect tools, publish quickly and learn from mistakes. Autonomy turns AI into an advantage.
Mattia's take: This is the most important point. Without access, trust and permission to ship, AI mostly makes existing processes slightly faster rather than creating exceptional marketing.
Where we landed
AI does not automatically create better marketing. It gives good marketers more leverage. That distinction is the whole conversation, and it is the reason two teams with identical tool stacks can end up in completely different places.
The teams that win will combine:
- Curious people
- Clear business goals
- Fast, constant experimentation
- Proprietary data, connected and trusted
- A culture with permission to actually ship
Read that list again and notice how little of it is about AI. Five of the seven ideas are things a good marketing leader should have been doing in 2015. What has changed is the penalty for not doing them. Curiosity used to be a nice trait; now it is the only part of the job a model cannot cover for you. A vague objective used to waste a quarter; now it wastes a quarter at ten times the volume.
The seventh idea is the one I would defend hardest, and Mattia called it too. Every organisation I have seen get real value out of this did the boring governance work first — access, trust, permission to publish. Everyone still waiting for AI to pay off is usually waiting for someone to sign something.
Download the carousel
Mattia turned the conversation into a ten page carousel — the seven ideas, both takes, and where we landed. It is a good thing to put in front of a team and argue with.
Download the carousel (PDF, 10 pages)
Your homework
Pick the one of the seven your team is worst at, not the one you find most interesting. They are rarely the same. Then ask the more uncomfortable version of the question: if it is number seven, who exactly is withholding the permission, and have you actually asked them?
With thanks to Mattia Santin for the conversation and the takes.