AI Transformation vs AI Implementation

There is a lot of conversation about AI implementation right now.

Agents. Automations. Copilots. AI-enabled CRMs. New models. New tools. New workflows.

And increasingly, we're seeing another conversation emerge alongside it.

Over the past few weeks, I've noticed more operators, GTM leaders and transformation specialists on LinkedIn questioning whether we're starting in the right place.

The question is shifting from “Where can we use AI?” to something much more interesting: “How should this business work differently now that AI exists?”

Those questions sound similar. They're not.

The difference is the gap between AI implementation and AI transformation.

We've seen this movie before

Jay Gaines recently made a point that stuck with me. Jay is a former CMO of Forrester and SiriusDecisions who now advises B2B CMOs, CEOs and PE operators. He's been speaking with marketing leaders about how organisations are rebuilding around AI.

His observation is that we've been through versions of this cycle before. Marketing automation promised greater efficiency. Then ABM. Then intent data.

Each brought powerful new capabilities. But organisations often layered the technology onto existing structures, processes and ways of working. The technology changed. The operating model didn't. And the expected commercial outcomes didn't always follow.

AI risks becoming the next, much larger, version of the same mistake. Because there is a fundamental difference between using new technology to make an existing process faster and asking whether that process should still exist in its current form at all.

Implementation starts with technology. Transformation starts with questions.

An AI implementation conversation often begins like this:

• What can we automate?
• Where could we deploy an agent?
• Could AI save our team time?
• Which AI tools should we be using?

Those aren't bad questions. They're just not necessarily the best first questions.

A transformation conversation starts somewhere else:

• Where is the business losing value?
• Where is growth being constrained?
• Where are margins being eroded?
• Why does this process take so long?
• Why aren't more opportunities converting?
• Where is valuable knowledge trapped in people's heads?
• What prevents this team serving twice as many customers?
• Could something we currently do internally become an entirely new product or service?

If we were designing this business today, knowing what AI can now do, would we design it this way?

That's a very different conversation. And sometimes the answer has surprisingly little to do with AI.

Automating a broken process gives you a faster broken process

Take a manual process involving six systems, three approvals, duplicated data entry and information moving backwards and forwards between departments. You could introduce AI and automate parts of it. You might save 20 hours a week. That's useful.

But you've also potentially taken a badly designed process and made it run faster. Transformation asks a harder question: why does the process work this way in the first place?

Maybe one approval can disappear. Maybe three systems can become one. Maybe information doesn't need to move between departments at all. Maybe a decision that previously required several people can now be made automatically within agreed commercial guardrails. Maybe an entirely different workflow is now possible. Only once we understand that should we decide where AI belongs.

Consider something as ordinary as creating a quote

We've recently been working through this problem with an industrial business. On the surface, the opportunity seems straightforward: use AI to create quotes faster. That would be an AI implementation project.

But once you start asking questions, the problem becomes much more interesting. How does the salesperson determine the correct configuration? Where does product knowledge live? How is pricing calculated? How are margins protected? Who can approve discounts? What happens when the customer's budget doesn't match the original specification? Can the salesperson intelligently reconfigure the solution while protecting profitability? And how much of this process depends on knowledge held by a handful of experienced employees?

Suddenly, we're not talking about automating quote generation. We're talking about redesigning a commercial system. The eventual solution might include custom software, automated product configuration, AI-assisted optimisation, margin guardrails, approval workflows and proposal generation.

AI is absolutely part of it. But AI isn't the strategy. The strategy is creating a faster, more intelligent and more commercially effective way to sell.

We're seeing the same debate emerge in GTM

Another conversation that caught my attention came from Daniel Zamudio, a four-time Head of Sales now working on AI-powered deal execution. Daniel pointed out that much of the current GTM engineering conversation is concentrated on getting opportunities into the pipeline: lists, enrichment, intent, personalisation, outbound automation, routing.

All useful. But considerably less attention has been given to engineering what happens after an opportunity exists: discovery, meeting preparation, business cases, stakeholder alignment, buyer consensus, deal strategy and commercial execution. His question was essentially: what good is more pipeline if win rates remain low?

If the actual commercial constraint is conversion, automating more prospecting may simply deliver more opportunities into the part of the system that's already broken. The technology works. The strategy doesn't.

Start with value. Then work backwards.

This leads to a relatively simple principle: don't start with the AI use case, start with the business outcome. What are we trying to achieve? Grow revenue? Improve margin? Increase conversion? Enter a new market? Reduce operating cost? Increase capacity? Improve customer experience? Launch a new product? Make better decisions?

Business objective → Opportunity → Strategy → Technology → Impact

The technology might be AI. It might be automation. It might be custom software. It might be an existing SaaS platform. It might be an integration between systems you already own. It might be a combination of all of them.

And occasionally, the answer might be: you don't need AI at all. That's not a failed AI project. That's good strategy.

The evidence is starting to support this

McKinsey found that companies attributing 5% or more of EBIT to AI were three times more likely to pursue broad operating-model redesign and twice as likely to redesign workflows before selecting AI tools.

Another 2026 study describes three stages of organisational AI adoption: enablement, automation and reinvention. The crucial distinction is that organisations at the automation stage are still starting with the current process and improving it. At the reinvention stage, companies begin redesigning work, roles and operating models from scratch around what is now possible.

AI implementation can improve what exists. AI transformation asks what should exist.

As implementation gets easier, choosing what to transform becomes more valuable

Implementing AI is getting easier extraordinarily quickly. Models are improving. Development cycles are shortening. Integration tools are becoming more accessible. Agents are becoming more capable. Small teams can now build systems that would recently have required substantially more engineering resource.

That's an extraordinary opportunity. But it also means technical implementation alone will become less differentiated. The harder questions become: what should we build, and what should we change?

That requires understanding the business, its customers, economics, processes, constraints and strategy before deciding where technology belongs. The competitive advantage won't come from having the largest collection of AI tools. It will come from knowing where to apply them.

The goal isn't more AI. It's a better business.

AI transformation shouldn't be measured by how many agents have been deployed. Or how many employees have Copilot. Or how many workflows have been automated. The real questions are much less fashionable.

• Did revenue increase?
• Did margins improve?
• Did the sales cycle shorten?
• Did capacity increase?
• Did customers get a better experience?
• Can the organisation now do something it couldn't do before?
• Did we create a new competitive advantage?

Because ultimately, the objective isn't to become an AI-powered business. It's to become a better business because AI exists. And that requires starting somewhere different.