How should AI tools be used in B2B digital marketing?

What you should never do with AI

We’re not anti-AI, but there are some things you should never use it for, period. Recently, hackers were able to break into several high-profile Instagram accounts by exploiting Meta’s AI support chatbot. All they had to do was ask it for a password reset.

“If AI can make changes in your system, someone will find a way to exploit it,” says Mark Whiting, VP of Technology at Motum B2B. “It’s a real security threat.”

You need to think like an engineer and give AI the guidance to work like an engineer. Mark Whiting, VP of Technology

The lesson: don’t give AI access to user accounts or sensitive customer data. It shouldn’t have the freedom to handle security-related functions without oversight.

As Ryan Iusi, Technical Program Manager at Motum B2B puts it, “Don’t give it the keys to the house.”

You should also avoid providing general-purpose AI platforms with sensitive information about your business such as asking it to audit the company financials, identifying information about customers, or anything else you’d typically consider confidential. Even if the platform has privacy settings or policies around sensitive information, it’s often unclear what can happen to that data once it has been shared with AI. Mark clarifies that there are ways to safely handle sensitive information through AI, but it should be done with a qualified expert behind the steering wheel.

To explain what AI can be used for, Ryan suggests visualizing your project’s infrastructure as a tree. The roots need to be firmly grounded in expertise up through the trunk and branches, which maintain various interconnected parts of your project.

“It’s risky to use AI to build the roots, but if you use it to create a leaf, there’s no risk of affecting anything further down the line,” he explains.

Create the preview, not the product

Now that we’ve (hopefully) deterred you from giving AI your treasured secrets, there’s plenty it can do to help ideate and iterate on B2B marketing ideas. We’re talking about the use cases that are helpful for people at varying levels of experience in the task they’re trying to accomplish — where extensive expertise is not necessary.

First off, AI can be great for prototyping, validating ideas, and solidifying your vision. Let’s say you have some ideas for an infographic, but you’re not sure about the visual design. AI can help you quickly explore different directions, compare, and gather ideas into a mood board or rough storyboard. You can bring these to your team for discussion and approval, then pass them on to the designers and copywriters who will use their expert eye to ensure your vision meshes with the brand identity.

“AI is an intermediary, not an initiator or a final reviewer,” says Mark, meaning it’s best used in the in-between steps to shape the next phase and visualize the end product.

When you look at a much larger project like a website build, there are plenty of those small in-between steps that could be sped up with an AI tool. For example, website development always requires wireframes, an early blueprint for the site’s layout and functionality, and later a set of high-fidelity mock-ups, which provide more detail around the visual design.

To non-developers, wireframes and even mock-ups can just look like a bunch of blocks on a screen. It’s hard to visualize how the site will look with your brand’s own copy and images. AI can generate convincing placeholder content for these steps, making it easier to preview the final product. The speed of AI generation helps cut down development time, as devs don’t need to get bogged down in the minutiae of content at this early state — they can keep building and focus on what they do best.

Notice that we didn’t say, “Use AI to create the infographic.” Or “Use AI to develop the website.” AI tools have plenty of useful applications, but the more complexity you add to the project, the more they require an experienced human in the driver’s seat.

AI with guardrails and human oversight

For more complex tasks, AI should still be used by people with expertise in the task they’re trying to accomplish.

Mark compares it to a power tool. Someone with little experience in construction can pick up a drill and install a shelf, and they probably won’t destroy their home in the process. If that same person wants to grab a jackhammer and retile the bathroom floor? Let’s hope they have an emergency contractor.

People call contractors for home repairs because they understand the cost of a poor DIY job. Although AI isn’t as well-understood yet, the same principle applies: if they try to reach beyond their expertise, a layperson can do real damage to their systems.

“If you don’t know anything about how to run a segmented campaign well, or what makes a good landing page, you shouldn’t be punting those types of decisions to an AI,” Mark says.

That’s also why you shouldn’t ask AI to develop your website from top to bottom if you’re not a dev yourself.

“The dream is very tempting: that a non-technical person can have access to these tools,” Mark continues. “If you use a generative tool to make a code base nobody understands, you’ll have to hire someone anyway to triage problems. You can’t work totally in AI, because it’s not built for a human to maintain long term.” Coders have noticed an issue called AI drift, which is when fully AI-coded projects gradually start to veer away from the original path and lose consistency. Working this way can easily lead to technical debt, which can drain resources and cause setbacks for your business.

What’s more, AI won’t ask you for clarification or challenge the logic of your prompt. AI platforms like ChatGPT and Claude are explicitly programmed to agree with the user except in very specific circumstances, a tendency called sycophancy.

“Sycophancy can lead to outcomes where bad decisions are boosted, codified, and broadcast by AI in an unquestioned way that always frames the initial idea as a great one,” Mark explains. “This is why the human user always needs to be able to critique the response and know what they're requesting.”

Beyond coding skills, web development requires strategy, critical thinking, and knowledge of the brand and how the business operates. You need these as a starting point to correctly prompt AI — and to predict the outcome of your prompt. A layperson may create a feature with AI without fully understanding the implications of that feature, which can cause problems down the line.

“The risk is that you don’t know what you don’t know,” says Ryan. “As developers, we understand how websites are built, and it’s important that someone in the driver’s seat is a domain expert that can chart the course and make shifts when necessary.”

“You need to think like an engineer and give AI the guidance to work like an engineer,” Mark adds.

Domain experts are tastemakers that can’t be replaced by AI. B2B brands often have very specific problems and goals, and humans are much more likely to come up with bespoke solutions and unique ideas that actually fit into your business objectives.

Working with a B2B marketing agency means having a team of these domain experts at your disposal. They can build tools for your business, AI and otherwise, with a clear vision and continuous monitoring to ensure it’s meeting your objectives.

“Discipline is a really important element that an agency can bring to bear,” Ryan explains, noting that humans can provide the framework and strong development practices these web projects need to succeed.