The Best AI Strategy Isn’t About AI

There are AI task forces, internal pilots, copilots, approved tools, AI labs and an increasingly long list of agentic experiments. Boards are asking about AI, CEOs are talking about it, and leadership teams are trying to figure out how quickly they need to move.

All of that makes sense, although I think we're starting with the wrong question.

The question isn't really "What's our AI strategy?"

The more important question is: How should our company operate differently now that intelligence is becoming cheap, abundant and increasingly programmable?

We've seen versions of this before. Companies didn't ultimately need an internet strategy. They needed to understand how the internet changed distribution, customer behaviour and business models.

The same was true with cloud. The technology mattered enormously, but the real advantage came from redesigning how companies built products, deployed software and operated at scale.

I think AI will follow a similar path.

If increasingly capable AI is available to everyone, simply having access to the technology isn't going to create much of a competitive advantage. The advantage will come from what companies are willing to change because of it.

That includes how work gets done, how teams are structured, where people spend their time and ultimately where capital gets allocated.

Start With the Work, Not the AI

The first wave of enterprise AI is understandably focused on productivity.

Companies are using AI to write faster, analyze information, generate software, automate customer service, create marketing assets and eliminate repetitive administrative work. There is enormous value here, and some of the productivity gains we're going to see over the next few years will be significant.

The risk is that companies simply use AI to make existing organizations slightly more efficient without questioning whether those organizations should continue operating the same way.

If AI makes a ten step process 20% faster, that's useful.

The bigger opportunity may be realizing that the process shouldn't have ten steps anymore.

This is why I think companies should start by looking at the work rather than starting with the technology.

Where does the organization spend the most time and money? Where are decisions unnecessarily slow? Where are people manually moving information between systems? Where does institutional knowledge live inside a handful of people's heads?

More importantly, where are customers frustrated because something takes days that should take minutes, or where are margins constrained because serving a customer requires too much human labour?

Those questions are much more interesting to me than asking where we can deploy another AI tool.

Once you understand the work, you can start redesigning the workflow around what AI now makes possible. That's a very different exercise from simply inserting AI into the way the organization already operates.

This is also why I think many AI pilots will ultimately disappoint. The technology may work perfectly well, but the organization around it doesn't change.

An AI agent generates an answer instantly, but three people still need to approve it. An employee saves five hours a week, but nobody has decided what should happen with those five hours. A company automates one part of a process while leaving everything surrounding it untouched.

At some point, AI stops being primarily a technology implementation problem and becomes an organizational design problem.

That's where things get much harder and also much more interesting.

One question I've started asking is relatively simple:

If we were designing this company today, knowing what AI can already do and where it's heading, would we build the organization the same way?

For most companies, I suspect the answer is no.

You might have smaller teams, fewer layers of management and dramatically different workflows. You might automate more work before adding headcount and organize people around outcomes rather than traditional functional boundaries.

You might also be able to serve customers who historically weren't economical to serve or build products that would have required dozens of people and millions of dollars only a few years ago.

That starts to look very different from adding AI tools to an existing organization.

It starts to look like an AI native operating model.

I don't think an AI native company means removing humans from everything. Human judgment, creativity, relationships, trust and accountability may actually become more valuable as intelligence and execution become increasingly automated.

It does mean becoming much more deliberate about where humans create differentiated value and where machines can increasingly take responsibility for the rest.

If an AI system can do 80% of something, the answer probably isn't to have a person continue doing 100% of it because that's how the organization has always worked. The more interesting question is what the human should own, what AI should own and how the workflow should be rebuilt around that combination.

This has implications well beyond technology.

It affects hiring, organizational structure, incentives and management. It also affects how companies think about internal capabilities, external partners and the amount of capital required to operate and grow the business.

That's why I increasingly believe the best AI strategies won't come from the AI team.

They'll come from leadership teams willing to rethink how the company actually works.

This is where AI becomes particularly interesting to me.

Every dollar a company puts into AI is ultimately competing with another use of capital.

A company can invest in making the existing business more efficient, redesign the core business around new capabilities, acquire technology or talent, partner with startups, build new products or create entirely new businesses outside the core.

Those are very different investments with very different risk and return profiles.

I think about the opportunity in three broad buckets: 

Optimize the core, transform the core and build beyond the core.

Most companies will naturally start with optimization because it's the easiest investment to justify. Automating a workflow, reducing costs or making employees materially more productive can generate relatively straightforward ROI.

The more interesting question is what happens after that.

What happens when AI doesn't just reduce the cost of running the existing business, but materially reduces the cost of creating something new?

Historically, building a new business required significant upfront investment in research, product, design, engineering, marketing and operations before you could learn very much. Today, much smaller teams can research markets, analyze competitors, build prototypes, write software and test customer demand at a fraction of the historical cost.

That changes the economics of experimentation.

It should also change how companies think about allocating capital toward innovation and new growth.

When the cost of experimentation falls dramatically, companies should be willing to run more experiments. I'm not talking about more innovation theatre or endless pilots that never go anywhere.

I'm talking about more capital efficient attempts to create new enterprise value.

I think this could become one of the biggest strategic advantages available to large companies over the next decade. They already have many of the assets that startups spend years trying to build, including customers, distribution, proprietary data, trusted brands, regulatory knowledge, industry expertise and access to capital.

Combining those assets with dramatically cheaper venture creation changes the economics of building beyond the core.

I don't think most companies have fully appreciated that yet.

The conversation can't stop at productivity, copilots, agents or having an AI strategy that can be presented to the board.

The more interesting question for CEOs, boards and capital allocators is:

What should we be doing differently because AI exists?

What should we automate or eliminate? What should we rebuild? Where should we stop adding headcount? What can we now offer customers that wasn't previously economical?

What new businesses have suddenly become viable? Where should we be experimenting because the cost of being wrong has fallen? Where should we allocate capital as a result?

That's why I think the best AI strategy isn't really about AI.

It's about redesigning the company and reallocating capital for a world where intelligence is increasingly abundant.

That's a much bigger opportunity.