How to overcome the 3 barriers to AI adoption

What’s really holding companies back from adopting AI?

This issue comes up often in my conversations with clients. Although each company is different, I can see some commonalities.

I recently read a commentary written by a respected thought leader about AI and why companies hesitate to adopt it. Although I didn’t agree with all her conclusions, it helped me clarify what I see as the greatest obstacles to AI adoption.

When we understand what’s holding companies back, we can develop solutions that make AI adoption more widespread and avoid costly errors.

Is there really a hesitancy gap?

Yes, when you look at companies of different sizes. Smaller companies are moving faster because they can. They have fewer layers of bureaucracy to contend with, fewer people who need to sign off on a plan, and fewer fingers in the decision-making pie.

However, the author says companies fear public embarrassment from making the wrong choices about AI, which leads to the gap. She’s not altogether wrong, but I believe three other factors have a stronger impact on AI adoption.

3 factors that hold companies back on AI

These factors have the greatest impact on how companies approach AI adoption.

1. Different ideas on AI’s use and value

A client asked me to review the business as a whole to help it solve its growth problems. I spoke with a cross-section of employees, and my deliverable was a report with guidance on internal and external ways to accelerate growth.

Many of those employees, including C-suite executives, used AI in specific aspects of their work. Some of their examples were spot-on in terms of the scope, breadth, and depth of what AI can assist with.

But we discovered that people had haphazard ways of learning about AI. When asked how they learned to use the models, the answers pointed to a fractured learning system: “Trial and error.” “I read some articles.” “I just jumped in.”

Not surprisingly, everyone also used AI for their own purposes. This chaotic approach can skew your results, introduce bias, and send you down the wrong path.

This company needs governance that ensures everyone understands AI in the same way. And it must install guardrails to make the use of AI transparent both inside and outside the company. We recommended that the company begin with a unified training system.

There are many ways to dive into AI, whether it’s Google’s Gemini, Anthropic’s Claude, OpenAI’s ChatGPT, or a SaaS provider’s AI instance. Each one is different. Organizations need a common framework that accounts for those differences and their limitations.

Your company’s needs, goals, and structure will help you decide which tool, or tools, to adopt. But once you adopt a tool, you must train everyone on it so you’re operating from a common understanding.

It’s not enough to say, “Everybody will use AI.” You need the right tool and must ensure you don’t introduce bias to the process.

My main recommendation is to establish governance that defines when to use AI and when not to, and ensures all senior executives deciding your company’s future align on the best way to use AI.

A comment from one employee clarified this issue. She was on an email thread where the opening email clearly came from someone who used AI to write it. One respondent on the executive team sent an AI-generated response.

Eventually, she said the thread became worthless because it was just a bunch of AI agents talking at each other.

But this is where hesitancy can trip up a company’s plans. You might have to develop rules for a company with 10,000 or 20,000 employees and dozens of different purposes, then determine how to control AI use and ensure it benefits the company. That accounts for a major part of the hesitancy gap, as I see it.

2. Concern about introducing bias

Employees who use AI without strong governance, solid training, and a common understanding of how to use it can inadvertently introduce bias. If their training is incomplete and ad hoc, they might not know how to structure prompts or understand the importance of adding context and nuance.

It’s way too easy to skew AI. All you have to do is ask your LLM, “What’s wrong with this?” Now the AI will assume something’s wrong and find out what it is. Maybe there’s nothing wrong. But that’s not what you told your AI model to find.

With a human, you can use tone to indicate whether a situation is wrong. With AI, you need context, nuance, and lots of information.

Think of AI as another team member, like a new employee, a trainee, or an intern. As with humans, your interpersonal connection makes the relationship work and helps your team member understand what you want.

Prompts aren’t orders. They’re longer-form conversations with descriptions, context, and clarity that tell your AI what to attend to and what to ignore.

You can push back on your AI-generated results and tell it when it gets something wrong.

I recommended to my client that when they publish something that AI has informed, they should disclose that, along with the opening prompt.

This tells people what you’ve done, why you did it, and how you arrived at your decision. As you did in school, you have to show your work. This transparency makes your AI products more accountable and helps keep people on the same page in their AI use.

Yes, this means you have to put in a lot of work up front to produce unbiased, transparent, and useful AI results. If it sounds familiar, it’s because it reflects the same process you follow to set up effective email automations. But once you do this prep work, you can go on to the more meaningful work of AI.

This contributes to the hesitancy gap more than the fear of public embarrassment does – it’s the upstream problem of creating rules and definitions.

3. Putting tactics before strategy

You have to start with a strategy every time. Stop me if you’ve heard this before. If you did, you probably heard it from me.

As with marketing in general, and email marketing in particular, you need a strategic roadmap before you start playing around with tactics like prompts and models.

Remember: AI isn’t about making decisions for you. It’s about guiding you on the journey so that you can find them. Kinda like therapy, but cheaper.

That’s why developing your AI strategy comes first. You’re clarifying what you’re looking for. You’re assessing whether you gave your AI enough information and defined the problem or challenge.

Most companies are missing this strategic step because they use AI ad hoc, without governance, guardrails, and consistent training.

What happens when you skip strategy?

Lest you think I’m pointing fingers at everyone else, let me follow my own advice on transparency and confess that I’ve done this, too.

I did it just the other day when I jumped into tactics before figuring out my strategy. I forgot my own rule that the AI engine I want to use isn’t always good at what I want it to do.

I started writing copy in ChatGPT for a big project. I forgot that ChatGPT is great at the strategic layer, not that great at writing copy. I remembered it after a frustrating hour of working and reworking – all because I didn’t think through the strategic step.

Eventually, I switched to Claude. I even confessed to Claude that I was “ready to choke ChatGPT.” But it wasn’t ChatGPT’s fault. I went down the rabbit hole with the wrong tool.

That’s why you (and I) have to remember that the AI strategic brief must be part of any governance.

The hesitancy gap is really a clarity gap

Clarity comes when you see the disconnect between jumping into the middle of the process with tactics and starting strategically at the beginning.

Companies are scared to define the rules, governance, and guardrails. That lack of clarity makes companies hesitant to go all in on AI right away.

How do you get that clarity? Not necessarily from your internal staff, but from external people who can see what has worked for other companies and assess your company’s dynamics.

Viewing AI as the latest example of incremental innovation helps you see it as a scalable operation, not something you launch in its final form. You roll it out one phase at a time, assess the performance, and improve it before rolling out the next phase.

I’ve written before here in MarTech that you can turn your company’s hesitation about implementing AI to your advantage. This clarity gap is one way you can demonstrate to your executives that you can map out a sure-footed process.

It also gives you more patience to understand that your company’s reluctance to leap into AI isn’t just corporate foot-dragging. A plan to help the company avoid pitfalls, including public embarrassment, may be the most effective way to get things moving.

The post How to overcome the 3 barriers to AI adoption appeared first on MarTech.