The world’s hardest problems need human wisdom, not automation. But a smarter division of labor with AI can free your team for the thinking that moves the needle.

We hear a lot of hype these days on how AI will solve all our problems, from fixing climate change to solving education to curing diseases. Humans have struggled for decades, if not centuries, with these issues. The idea that AI will magically solve the world’s wicked problems seems like a hearty dose of magical thinking. 

However, it can definitely free up some mental bandwidth for people and organizations working on those problems. 

Tapping into Centaur Mode — thoughtful collaboration between humans and AI, rather than simply handing it all off to the machine — can help mission-driven leaders focus their precious time and energy on devising and implementing solutions. 

Achieving results will likely come down to AI’s unprecedented capacities in “kind” environments rather than tackling wicked problems head-on.

Kind and wicked worlds

A “kind” environment involves having clear rules, predictable patterns, and replicable solutions. There are objectively right or wrong answers. Feedback usually comes quickly — you know if there’s a winner or a mistake almost immediately. Games like chess or tennis and activities like computer programming or data analysis operate in kind environments. You can practice specific techniques and study patterns to increase your chances of winning or achieving your goals.

In contrast, issues like climate change and poverty involve a complex mix of interconnected factors — a “wicked” environment. Solutions or policies that worked yesterday might not work tomorrow. Many people, communities, organizations, or countries might be involved in various ways, with different, competing, or overlapping interests. It’s not always clear what “winning” looks like in a wicked environment — and there’s often a long delay before you know if your actions had any benefit.

AI excels at kind tasks. Give it a specific job to do with the right context, it will deliver gold in minutes. Synthesize hundreds of pages of dense text? No problem. Create a table comparing results from various kinds of interventions with pros, cons, and outcomes? Easy peasy. 

Wicked tasks, however, are a very different story as the following story illustrates.

Learning from the comic failure of Claudius

For all its impressive capabilities, current AI tools can’t handle messy activities that humans can navigate reasonably easily. Anthropic’s vending machine experiment provides a good example of how well an autonomous AI system would perform on its own in a reasonably complex environment. Their AI agent, Claudius, presided over what can only be described as a comic failure to run a simple business.

Especially compared to things like tackling climate change, selling drinks and snacks out of a machine doesn’t seem all that complicated. However, Claudius ran into countless problems in its primary goal (making a profit), due to things like:

• an inability to assess the market (turning down highly profitable offers like $100 for a $15 six-pack of Scottish cola)
• stocking — let’s call them — abnormal items (like one-inch cubes of tungsten and bags of russet potatoes)
• offering discounts to all Anthropic employees (ignoring the fact that ALL of its customers were Anthropic employees)
• hallucinating Venmo payment accounts, management conversations, and supposedly in-person visits to suppliers (at the home of the iconic cartoon family The Simpsons, no less).

While Claudius drove this business into the ground, the AI did succeed on a number of fronts. It classified customer requests, researched suppliers, and devised tools to keep track of profits and cash flow. And as one major bright spot in its decision-making, it turned down some requests for dubious items like crystal meth and medieval broadswords.

Each of its successes, except the last one, could be described as “kind” tasks — things that take a lot of time for a human, but aren’t always the best use of brainpower.

With that in mind, here are some ways we can capitalize on AI’s skills at executing in a kind environment.

Centaur communications strategy

Let’s look at a scenario: designing a communications strategy for a new conservation program in a conflict-affected region.

In a program like this, we’re looking at not one issue, but various intertwining wicked problems — each of which is hard enough to tackle in isolation.

We have to understand multiple complex factors: environmental and natural resource issues, context about conflict, economic and political drivers, and — most important of all to our comms efforts — the people involved and what matters to them.

For our comms strategy, we’d have a range of activities in front of us:

• Conducting interviews with stakeholders
• Compiling research on the context of conservation and conflict in the region
• Synthesizing insights from those interviews and research
• Exploring and identifying high-value communications tactics and messaging
• Designing an implementation plan
• Defining and measuring success indicators

Each of these has kind and wicked elements. By relying on AI for the kind tasks, we can save our time and brainpower for untangling the wicked aspects: understanding the context, assessing options, making decisions, and taking action.

Here’s an example of what some of the tasks might look like in practice:

And so forth. In each case, we handle the bigger picture thinking and decision-making. We exercise our human capacity for discernment to identify the best course of action given the circumstances. AI handles the tasks that would otherwise take many hours of tedium.

One of the challenges of wicked problems is that solutions aren’t easily replicable. Which means we have to do this kind of diagnostic work again and again. We need to understand the context we’re working in. Humans do this very well, with experience and given the time to do the work. AI in its current form can’t understand complex circumstances — but it can certainly save us a lot of time.

Conclusion

Given the scale and impacts of the world’s wicked problems, we can’t afford to waste our resources. Learning when to use AI and when to stick with human thinking will be a crucial skill in the coming years. This distinction between kind and wicked environments can give us a helpful framework to make those decisions.

Working with AI on kind tasks can help us focus our time, attention, and resources on designing and implementing strategies that make a difference. For organizations with limited budgets, that means getting more strategic value from the team you already have.

There’s no magic solution to the world’s wicked problems, and perhaps there never will be. But at least we can make better progress by using the best tools at our disposal.