- Ask CMO
- Posts
- 🤖 Why Your B2B Marketing Team Needs a Custom AI Model
🤖 Why Your B2B Marketing Team Needs a Custom AI Model
Test campaigns on an AI clone of your ideal customer
Imagine testing your next campaign on a digital clone of your ideal customer.
Or bouncing ideas off a chatbot trained on your team’s expertise and performance data.
Industry “experts” have been hyping this stuff for a while, but it’s not just talk anymore. B2B teams are actually using these tools and seeing results...
We talked to John Lahr, Growth Product Manager at WebAI, who’s building custom AI models like it’s a full-time job (because, well, it is).
Why Your B2B Marketing Team Needs a Custom AI Model
John broke down how companies are using custom AI models right now to save money, avoid guesswork, and finally make marketing decisions they can trust...
What can you do with your own custom AI model?
Ask ChatGPT and Claude for advice on your latest campaign and they’ll spit out some generic advice faster than a vending machine drops a bag of stale chips.
Why? These models are trained on everything... the good, the bad, and the downright irrelevant. No matter how brilliant your prompts are, they’re not built to understand your business.
Now, picture this: an AI that’s actually tuned into you. A custom model trained on your customers, your strategies, and your goals.
John explains how B2B marketing teams are doing exactly that:
1. Synthetic personas
With a custom AI model, you can upload customer insights and performance data to create a synthetic persona that thinks and acts just like your ideal customer.
You can fine-tune a chatbot that reflects a particular customer persona, run ideas by it, and ask, ‘What do you think of this messaging?’ or ‘How does this creative feel?’
You can ask your synthetic person: “Hey, does this message resonate?” And it’ll tell you without the guesswork or blowing your budget on real-world testing.
Pretty smart, right? Well, there's another way companies are using this technology that's also turning heads...
2. Digital twins
Every team has that one person who knows everything. The go-to for strategies, fixes, and creative sparks.
But what if they leave? Digital twins let you clone their brilliance, turning their expertise into a permanent asset.
You have people that have a ton of expert knowledge that might be retiring... can we actually capture that in AI so that when they leave or retire, they can still be referenced? And then how do we grow that area of expertise across multiple different points of view, that more junior employees or anyone can interact with?
This isn't about replacing humans with robots (John was clear about that).
It’s more about bottling the wisdom of your valuable human assets and turning it into a 24/7 AI consultant.
3. Internal experimentation
AI isn’t just for the tech team. According to John, smart companies are putting these tools directly in their marketers' hands.
Whether it’s automating reporting, modeling strategies, or improving collaboration, just a little bit of AI literacy can make these projects possible.
The ability for different people on the marketing and creative teams to get literacy around AI tools has been very helpful. Even having a junior marketer use chatbots to write Excel formulas for growth modeling can make a huge difference.
Give your team the tools, and watch them find clever ways to solve problems, test ideas, and boost productivity.
And frankly, this is barely scratching the surface of what AI can do to help you and your team...
Where do you even begin? How to pick the right ideas to test first.
AI can do so much, and that’s kind of the problem.
With so many possibilities, it’s hard to know where to start... and that’s often what stops people from diving in at all.
John knows how overwhelming it can be, so he’s come up with a few questions to help you figure out which ideas are worth testing first.
What areas of business are truly core? Identify the parts of your business that are absolutely core. Your key products, processes, or customer touchpoints. These are prime candidates for AI enhancement.
The goal isn’t to get an answer on everything right now. The goal is to start thinking about what pieces of your business are actually truly core. Is it your product? Is that your IP? Is it actually your marketing? Do you sell a commodity product and your marketing is actually your core differentiator?
Who, and where, are your customers? Where do they hang out, and how are they finding you? Use AI to enhance their journey without losing the personal touch.
What is your current distribution? SEO, ads, referrals. Where does your traffic come from? AI can help you strengthen those channels or explore new ones.
What is your team’s current level of AI competency? Are your employees AI-savvy, or do they need a crash course? Start with tools and ideas they can easily adopt.
Where are the areas that you would want to upskill your employees to have a base level of understanding of what these tools do and where you can be more efficient with them?
Where can you be more efficient? Repetitive tasks or tricky processes that slow you down? Start your AI journey there for some quick wins.
You’ve probably already got a few ideas. Let’s break down what it actually takes to bring these ideas to life...
Building a custom AI bot for dummies
Building a custom AI bot?
It sounds like rocket science, but John somehow manages to make it sound more like assembling IKEA furniture.
Sadly, this breakdown won’t turn you into an engineer, but it will give you a very simple roadmap for taking your first steps:
Choose a base model: Different models excel at different tasks. Open-source models are cheap and cheerful. Big models need big hardware investments, so pick what fits your setup.
Fine-tuning: Fine-tune the model with your secret sauce: key documents, tone of voice, core concepts, and inside jokes.
You want to think about what is the actual information that you want to imprint into the model. What are the concepts you want it to have? What is the tone of voice and kind of perspective that you want it to have?
Retrieval: Teach it to fetch like a good little doggo using RAG. RAG is an AI framework that allows the AI to pull in specific information from specialized documents (like a manufacturing manual), ensuring it provides accurate, context-specific answers.
Continued learning: Keep improving with human feedback to keep the AI sharp and relevant.
That’s where you can vote things up, vote things down... adding more information to the corpus of training data and retraining.
And don’t sweat the timeline. John says for most production-grade systems, you’re looking at a timeline of weeks to a couple of months.
Final tips from John
John mentioned how Andrew Chen's "Law of Shitty Clickthroughs" nails a brutal truth about tech...
Early movers grab the best returns. The rest fight for scraps. And he sees the same pattern unfolding with AI.
AI is going to be a huge catalyst for change... the question is: do you want to control your destiny or not? Because you can either rely on others to define the future - which means premium costs, lack of control, and explainability - or you can own it yourself.
The good news is you don’t have to figure everything out overnight.
Learn and experiment your way from fuzzy to clear. What types of systems do you want to build and integrate? Build yourself? Buy from others? Maybe use some open-source tools off the shelf and customize them. There’s so much going on in the open-source community right now where, with a pretty low amount of effort, you can build and fine-tune your own stuff.
Start small. Pick one project. Learn from it. Scale what works.
Your competitors face the same challenges. Some will wait for perfect solutions, others will start building their AI muscle today.
The question is: Which side will you be on?
P.S. Did someone send this to you? What a legend. There’s more where this came from, subscribe here.