The race to adopt AI has many organizations asking the wrong question. Lasting value doesn't come from implementing AI—it comes from solving real business problems, with AI only where it genuinely adds value.

Why "adopting AI" is the wrong starting point — and what to do instead

There is a question making its way through every boardroom, every leadership offsite, and every industry conference right now. It sounds urgent. It feels existential. And in the way most organizations are asking it, it is almost completely unanswerable.

The question is: how do we adopt AI?

The reason it's unanswerable is not because AI is too complex or too new. It's because the question skips over the only thing that would make the answer meaningful: a clear description of what the business is actually trying to achieve.

AI is not a strategy. It is not a destination. It is not a transformation. It is a tool — a powerful one, a genuinely consequential one — but a tool nonetheless. And no one has ever built anything lasting with a tool alone.

What Leaders Actually Mean When They Say "AI"

When a business leader tells us they want to adopt AI, the first thing we do is slow down and ask what they mean. Not to be pedantic, but because in our experience, most leaders who say they want AI are actually describing something else entirely.

Some are experiencing peer pressure. Their counterparts at other companies are talking about AI constantly. They feel the social cost of not having something to contribute to that conversation — of being perceived as behind, as old school, as not relevant to where the industry is heading. The desire to "adopt AI" is, in part, a desire to be part of a conversation they feel excluded from.

Some are genuinely worried about efficiency. They know intuitively that AI-powered tools can help people do more with less. They want that for their teams. But they're also afraid of what less looks like in a business where high-touch, personalized service is the core value proposition. They don't want to automate their way to a worse customer experience.

Some have conflated AI with digital transformation entirely. In their mental model, transformation means AI, and AI means transformation. The two terms have become interchangeable — which means neither one is doing the work it should be doing in a strategic conversation.

What almost none of them mean, when they say they want to adopt AI, is that they have a specific, measurable business outcome they are trying to achieve and they have identified AI as the most effective path to that outcome. That version of the conversation is rare. It is also the only version that leads anywhere useful.

The Scissors Problem

Telling a business to "adopt AI" without a clear outcome attached to it is like telling a child not to run with scissors. You are describing how to handle a tool safely. You are not describing what you are going to create with it.

And here is the thing about scissors: no one has ever created anything meaningful with scissors alone. Scissors are one instrument in a larger creative act. They cut things that need cutting, at the right moment, in service of something being built. The snowflake is not the outcome. The snowflake is what you get when scissors are the only tool you have.

A premature AI initiative looks exactly like this. It is a mandate that every person on the team uses a specific AI platform for specific functions. There might be an AI policy — guidelines for how to use tools responsibly, how to cite AI-generated content, what platforms are approved. These are not unimportant. But they are the equivalent of teaching scissor safety. They describe the rules for using the tool. They do not describe what the organization is building.

Meanwhile, the underlying business problems that transformation is supposed to address remain exactly where they were. Front-line employees may be somewhat more efficient. The business is not growing. The customer experience has not meaningfully improved. The competitive pressure that drove the AI conversation in the first place has not been answered.

What Actually Goes Wrong First

Every article in this series has pointed to the same root cause, and the AI conversation is no exception: businesses try to solve problems with tools before they understand what they are actually trying to solve.

The 80/20 problem we described in an earlier article applies here with particular force. AI performs well in processes that are standardized, well-defined, and repeatable. When a process has 20% variability — exceptions, edge cases, the situations that require human judgment — AI will fail in exactly those situations until the exceptions have been mapped, the human logic for handling them has been documented, and the system has been trained on that logic.

An organization that hasn't done the current-state work — that hasn't mapped its processes, audited its data, and identified where its exceptions live — is not ready to deploy AI in any meaningful way. It is ready to run a pilot that will underperform expectations and add another data point to the list of initiatives that didn't deliver.

Data is the other foundational dependency that almost always goes unexamined. AI's ability to recognize patterns, analyze trends, and generate insight depends entirely on the quality and structure of the data it works with. In most services businesses on day one of a transformation engagement, that data is fragmented, inconsistently structured, and partially trapped in legacy systems that don't talk to each other. Deploying AI on top of bad data does not produce good insights. It produces confident-sounding wrong answers — which can be more dangerous than no insight at all.

The house has to be in order before the technology can do what it promises.

Where AI Actually Delivers

None of this means AI isn't valuable. It is — in the right context, deployed with intention, in processes that have been prepared to receive it.

The use cases where we have seen AI create immediate, demonstrable value in services businesses share a common characteristic: the process is defined, the data is reasonably clean, and the stakes of an error are recoverable.

Customer-facing AI that replaces static chatbots is one of the highest-impact opportunities — but only when it is implemented well. The difference between an AI that feels like a better version of the robotic chatbot everyone has learned to hate and one that genuinely serves customers comes down to training. An AI customer service capability that has been trained as comprehensively as your best human representative — on your products, your policies, your most common exceptions, your brand voice, the nuance of your highest-value customer relationships — can deliver something close to the high-touch experience your customers expect at a fraction of the cost. An AI that has been pointed at a generic knowledge base and told to answer questions will erode trust faster than no AI at all.

Internal AI agents are often the better starting point. In any process that involves quality assurance — order entry review, compliance checking, manual review of human work — AI can replace repetitive human effort and increase throughput significantly. The risk is lower because the output is not immediately customer-facing. A human can review what the AI produces before it reaches the customer, which creates a natural feedback loop for improving the system while protecting the relationship. AI agents reviewing the work of other AI agents is a pattern we are seeing increasingly in well-structured operations — and it is one of the cleaner illustrations of what AI actually does well: high-volume, rule-based evaluation of structured inputs.

Pattern recognition and trend analysis are also areas of genuine AI strength — assuming, again, that the underlying data is reliable. A services business that has done the work of cleaning and structuring its data can use AI to surface insights that would take a human analyst weeks to produce. But this capability is downstream of the data work, not a replacement for it.

How We Actually Figure Out Where AI Belongs

When we work with a leadership team that wants to understand where AI fits in their transformation, we do not start with AI. We start exactly where we start with every other engagement: with a clear description of the business outcomes the organization is trying to achieve.

What does the business need to look like in twelve months? Where is the revenue pressure coming from? Where are the operational inefficiencies that are eating margin? Where are customers experiencing friction? What are the processes that are currently dependent on individual expertise that can't scale?

Once those questions have been answered — once there is a current-state map, a data audit, a process inventory, and a set of specific, measurable future-state outcomes — the places where AI belongs become identifiable rather than speculative. We look for processes that are standardized and repeatable. We look for data that is clean enough to work with. We look for exceptions that have been documented. We look for the human work that is high-volume and low-judgment — where AI can take on the repetition and free the humans for the work that actually requires them.

AI almost never appears first on the roadmap. It appears where it belongs, which is usually after the foundational work has created the conditions for it to succeed.

For the Leader Who Feels Behind

If you are looking at your competitors and feeling like you are falling behind on AI, we want to validate that concern — and then reframe it.

If it feels like you are behind, you probably are. But the gap is almost certainly not only about AI adoption. The businesses that are genuinely pulling ahead are not just using AI tools. They are operating with cleaner data, clearer processes, better cross-functional alignment, and a more disciplined approach to measuring what they invest in. The AI is the visible part. The infrastructure underneath it is what makes it work.

The answer to feeling behind is not to rush toward the tool that your competitors appear to be using. The answer is to build the foundation that makes any tool — including AI — actually deliver on its promise.

That starts with a plan. A specific, describable picture of where you want to be. A roadmap that sequences the work in the order that creates momentum rather than debt. And a clear understanding that AI, when it appears on that roadmap, will be positioned to do something measurable — not just to be there.

The leaders who will look back in three years and feel good about how they navigated this moment are not the ones who adopted AI first. They are the ones who knew what they were building before they picked up the tool.

Why We Build It This Way

Every article in this series has returned to the same conviction: technology is a vehicle, not a destination. The Plan phase exists to establish the destination before any vehicle is chosen. The Grow phase ensures that investments compound rather than scatter. The Scale phase ensures the roadmap sees the whole organization rather than just the piece that's most visible right now. And the Repeat phase ensures that what's working gets more investment and what isn't gets corrected before it becomes a failure.

AI belongs somewhere in that system. For most organizations, it belongs later in the sequence than the conversation suggests — after the data is ready, after the processes are mapped, after the exceptions are documented, after the team understands what they are building and why.

When AI arrives at the right moment, in the right process, with the right training, it does not feel like an initiative. It feels like the obvious next step in a business that has been building toward it all along.

That is the difference between AI as a mandate and AI as a multiplier.

This is part of a series of six articles exploring the most common challenges facing digital transformation leaders — and how a connected, cross-functional approach changes the outcome. Read the full series here.