Exploration v. Exploitation

In 1991, a Stanford professor, James Marsh, wrote a paper called “Exploration and Exploitation in Organizational Learning”. The paper highlighted the fundamental tension between what he called exploration - “search, variation, risk-taking, experimentation, play, flexibility, discovery, innovation" - and exploitation - “refinement, choice, production, efficiency, selection, implementation, execution”. 

He suggested that organizations have a choice: they can exploit the knowledge and resources they already have or they can explore in search of an uncertain outcome that might be better. 

Sound familiar?

The Tradeoffs 

You have limited resources, so there’s always going to be a tradeoff. There’s no getting around that, but understanding exploration and exploitation helps maximize where those resources might go. 

Exploration pays off in the long-term. Whether you’re exploring new service offerings, technology, engagement models, or office locations - you’re more likely to find the top 10% outcome when you search and experiment. But it certainly comes with a cost. Exploration is risky and often inefficient. You might never find that top 10% outcome, but still spend all your money and time in the search. 

Exploitation pays off now. You’ve found the key solution or niche that you’re really good at and pays well. So double, triple down and milk it for all it’s worth. Exploitation is predictable but can become obsolete quickly.  

The tradeoff sequence that’s most common for firms is a success trap. Success from exploitation makes us do more of it, and exploration disappears over time. But the answer can’t be all or nothing. Exploration without exploitation is a hobby. Exploitation without exploration is a melting ice cube. 

What’s Happening Now

This dynamic is playing out in accelerated fashion right now as firms scramble to identify how AI should be woven into their business models, and I’m guessing there’s a mix of perspectives on explore v. exploit within your own firm.

For most of you, “business as usual” has never really been the job. You’ve always had to evolve your offerings, adjust to clients, and respond to change. That instinct is useful here.

What’s interesting is that I’m not seeing many firms double down on what already works. The uncertainty around AI makes that difficult. Instead, the more common reaction is to explore (sometimes too broadly).

There’s an underlying assumption that AI resets the game, and that the right response is to build entirely new offerings around it. So you see marketing firms experimenting with products, design agencies moving into AI advisory, and industry-focused firms loosening their focus to chase new categories.

My Suggestion

Constraints are what make exploration useful. They create stable learning and allow insights to compound into something meaningful. 

If you’re an SEO firm, don’t get distracted by a vibe-coded product idea. Focus on how your discipline is actually evolving. The same goes for industry expertise. Don’t walk away from years of accumulated understanding because “AI makes it easier.” It doesn’t. It just increases the amount of information. And without focus, more information is more likely to derail you than help you.

The goal is to find new opportunities for exploitation while continuing to deliver value to your clients. That only happens when exploration is targeted.

Here’s how I propose approaching it within your firm: 

1. Define your “non-negotiables” for exploration

Not everything deserves to be explored. Tie exploration directly to your positioning, your audience, and the problems you already understand. If it doesn’t make you better at solving those problems, it’s a distraction.

2. Mine your existing work before chasing new ideas

Your best signals aren’t online, they’re inside your current client base. What are clients asking? What’s changing in their world? What are your practitioners already seeing? Start there before you go looking elsewhere.

3. Put a ceiling on exploration (time, budget, and scope)

Exploration expands to fill whatever space you give it. So don’t. Cap it. 10–20% of team time, a fixed budget, a defined window. Constraints force better thinking and prevent exploration from quietly replacing the work that actually pays you.

4. Turn exploration into something you can sell (or kill it)

Exploration isn’t valuable because it’s interesting. It’s valuable when it becomes usable, something a client will pay for or that materially improves your current offering. If it doesn’t get there, move on.