I’ll be honest, even for someone who nerds out about this stuff for a living, the pace of AI can be exhausting. Just as we’re all getting comfortable with AI that writes or makes pretty pictures, the next wave has arrived: AI that does things.
This is the dawn of the ‘agentic’ or ‘machine-to-machine’ era. It’s the digital equivalent of ‘have your people call my people’, except ‘your people’ are now highly intelligent, autonomous bots. For brands, this means your next customer interaction might not be with a person, but with their AI agent.
And while it sounds massive and complicated, the kind of thing that makes you want to just close the laptop and go for a walk, one of my current obsessions is figuring out how little we actually need to do to get ready. Less is more, right? It turns out, preparing for this shift isn’t about some massive, bank-breaking project. It’s about getting a few simple things right.
You can start preparing now by becoming a bouncer for your digital front door, finally tidying up that data mess you’ve been avoiding, and tinkering with accessible tools to build your first simple automations.
Let’s do this.
1. Become the bouncer your website deserves
Right now, your website has a digital front door with no one watching it. All sorts of bots are wandering in and out: Google’s crawlers, partner APIs, social media scrapers, and a few shady characters from the darker corners of the web.
An “Agent Traffic Audit”, as my very clever colleague Brian Yamada calls it in VML’s Win the Future of Machine-To-Machine Marketing With Agentic AI report, is basically getting a clear picture of all the non-human traffic hitting your digital properties.
A good bouncer doesn’t just block people; they manage the flow. They know who’s who.
They check IDs: In the digital world, this means looking at a bot’s signature (user-agent) to know if it’s Google, a partner or something unknown.
They have a VIP list: The bouncer knows to let verified partners and valuable bots straight in, maybe even guide them to a special entrance (a specific API). Who doesn’t love a good VIP list?
They do some bouncing: They identify and block malicious bots trying to scrape your data or find security holes, which is especially necessary in the age of vibe coded applications.
These have always been good practice, but it’s now critical. A human can navigate a messy, crowded website if they really want to. An AI agent can’t. It needs clear rules. By simply identifying and managing the bots already on your site, you’re building the foundational logic your future ‘brand agent’ will need to interact with customer agents safely and effectively. It won’t win you the shiny awards, but it’s the work that matters.
2. Tidy your house: the data-readiness reckoning
For the last decade, ‘data cleanup’ has been the corporate equivalent of ‘I’ll start my diet on Monday’. It’s the thing everyone knows they should do, but it keeps getting pushed. Well, agentic AI is the catalyst that finally forces our hand.
Imagine you have a personal AI agent you’ve tasked with ordering your groceries. It knows you prefer a specific brand of oat milk, but you’re flexible on the brand of cheese as long as it’s cheddar. Now, if that agent goes to a supermarket’s website and the data is a complete mess, products are mislabelled, stock levels are wrong and the APIs are a tangle of spaghetti code, the agent will fail. And what’s worse, you won’t get your cheese.
An AI agent acting for a customer is only as good as the data it can access. If that data is a mess, the worst-case scenario isn’t it failing, it’s that it will go to your direct competitor. With AI, the concept of loyalty will be less relevant.
Getting ‘data-ready’ isn’t some futuristic project. It’s about paying off years of technical debt. In practical terms, it means:
Documented APIs: Clean, secure, and clearly explained digital doorways for your data.
Structured content: Your products and info are labelled in a way a machine can understand without guess work.
Reliable information: Your stock levels are accurate and your prices are correct.
This is the unglamorous, un-sexy work that will separate the winners from the losers. The brands that do this now won’t just be ready for AI agents, they’ll also just have a more efficient and reliable business.
3. Take the time to tinker
Here’s my favourite part because reading about this stuff is one thing, but getting your hands dirty is where you really learn. You don’t need a massive budget to start building your own simple agentic systems.
The goal is to start off by automating a simple, nagging, internal process. Find a repetitive task and build a workflow to handle it. A classic example is a bot that automatically pulls competitor news from various platforms into a single Google Sheet each morning. That’s an agent!
A great place to start is workflow automation platforms. Tools like N8N and Google Opal are like digital LEGO sets for APIs. You can visually connect different apps (like Google Drive, Slack, and your CRM) to create automated workflows, often without having to write a single line of code.
This is the essence of an agentic system: one thing triggers another, which triggers another. This is the stuff we’re getting our hands dirty with at VML, moving from theory to actual practice, and it’s where the real learning happens.
You don’t need to use it tomorrow, but understanding the concepts shows you where the industry is heading.
By building one or two small, internal bots, you demystify the entire concept. You learn about APIs, authentication, and what happens when things break, all in a low-stakes environment. This is infinitely more valuable than just reading another think piece. Except this one. This one is deeply valuable of course.
The end goal is still human
In all this talk of machines talking to machines, it’s easy to forget the whole point. This isn’t about removing people (I hope), it’s about removing friction. It’s about delegating the grunt work to machines so we can focus on creativity, strategy, and human connection.
Your brand’s voice, its values, and the trust you’ve built will be the ultimate differentiators that convince a person to send their AI agent to you and not someone else. The future isn’t about building Skynet (although I can’t speak for all of us; some AI founders make me nervous), it’s about building smarter doorways.
And the best time to start checking the locks was yesterday. The second-best time is now.
Tilesh Bhaga is creative director at VML South Africa.
Download VML’s full report Win the Future of Machine-To-Machine Marketing With Agentic AI here.
