Just a few years ago, businesses wrestled with artificial intelligence mainly in the abstract — a “future of work” problem they’d have to contend with down the line. Now? More than half the companies around the world are actively adopting AI. Although investments are particularly high in industries such as health care, data management and processing, cloud computing, and fintech, all types of organizations and functions have incorporated AI technology into their operations. And generative AI tools such as ChatGPT are forcing leaders to ask where and how AI can help their businesses.
Even so, most companies still view AI rather narrowly, as a tool that alleviates the costs and inefficiencies of repetitive human labor by automating mundane physical tasks (like moving merchandise in warehouses) and increasing organizations’ capacity to produce, process, and analyze piles and piles of data. But the technology can do much more than that.
Paired with “soft” inquiry-related skills such as critical thinking, innovation, active learning, complex problem solving, creativity, originality, and initiative, this technology can further our understanding of an increasingly complex world, allowing us to engage in more abstract questioning and shifting our focus from identification to ideation. In our research and workshops with executives, we’re finding that companies have much to gain by treating AI as a knowledge-work collaborator in diverse areas such as product design, process efficiency, and prompt engineering. Partnering with the technology in this way can help people ask smarter questions, which in turn makes them better problem solvers and breakthrough innovators. We are also seeing the initial impacts of more context-aware AI systems (like ChatGPT), and as they continue to improve, the skill of asking questions (or creating prompts) will only become more valuable in the discovery process.
Although experts have recognized the need for software engineers to ask smart questions upstream, when developing automated tools (to bake in fewer biases and assumptions), little has been said about the flipside of the relationship between AI and inquiry: the technology’s potential to help people become more inquisitive, creative problem-solvers on the job. We aimed to correct this oversight through design-thinking sessions and extensive follow-up conversations with tech-driven business leaders from a diverse array of countries and industries. We also surveyed roughly 200 leaders, from more than 30 countries who participated in our executive education programs at MIT —to learn how artificial intelligence has affected questioning patterns and innovation behaviors and outcomes in their organizations. (For this research, we’ve defined “artificial intelligence” broadly to include machine learning, deep learning, robotics, and the recent explosion of generative AI.)
We have found two distinct, yet related, paths that leaders follow to strengthen their (and their teams’) inquiry muscles as they tap the power of AI in their question-asking work.
On the first path, they can use the technology to change the cadence and patterns of their questions: AI increases question velocity, question variety, and question novelty. Results from our ongoing research show that AI can significantly increase all three.
On the second path, AI can help transform the conditions and settings where people work so that questions that spark change — what we call “catalytic” questions — can emerge. This pushes leaders out of their comfort zones and into the position of being intellectually wrong, emotionally uncomfortable, and behaviorally quiet and more reflective, all of which, it turns out, promotes innovative thinking and action.
Let’s look at how each path can lead to breakthrough ideas.
Increasing velocity, variety, and novelty.
Partnering with AI to ramp up the velocity, variety, and novelty of questions requires companies to train algorithms to answer the basic, easy (yes/no) questions independently and to reveal deeply buried patterns in the data. When this foundation is laid, humans can start exploring the power of more context-dependent and nuanced questions that AI technologies are not yet capable of answering alone.
Question velocity
Algorithms can provide immediate answers to questions that leaders pose, in turn allowing them to ask more — and more frequent — questions. In our research, we found that 79% of respondents asked more questions, 18% asked the same amount, and 3% asked fewer.
At the cybersecurity firm Cybereason, researchers rely on AI and machine learning to immediately answer the basic questions about what happened in an apparent breach so the team can more quickly turn its attention to formulating deeper questions about why it happened. In the past, CEO Lior Div said, findings were more black-and-white: “It’s an attack. It’s not an attack. It’s good or it’s bad.” But the speed with which AI filled in those blanks opened up a whole new line of questions around intent — and what hackers are really after in a given situation.
Of course, there are risks to using AI to generate rapid-fire questions. For one, people may keep asking more and more questions without working their way toward an actionable path, making it important to recognize when the process stops being productive. For another, more questions don’t necessarily amount to better questions, which means you’ll still need to exercise human judgment in deciding how to proceed.
Question variety
AI helps uncover patterns and correlations in large volumes of data — connections that humans can easily miss without the technology. Knowing they have this tool at their disposal frees up leaders to ask farther-ranging questions and explore new ideas that they may not have otherwise considered. In our research, we found that engagement with AI led respondents to ask different questions than they otherwise would have 94% of the time.
Consider this example: Kli Pappas, the director of predictive analytics at Colgate-Palmolive, told us that his team tapped AI to understand how charcoal became a wildly popular ingredient in consumer products so they could “find the next charcoal.” Their algorithm generated and answered thousands of questions based on their initial search for data, sketching out a decades-long trajectory from charcoal scrubs in South Korea 20 years ago to charcoal appearing in face washes in the U.S. and then in all kinds of products around the world. The AI-generated data led the team to ask hundreds of less-obvious questions to spark creative thinking about future trends that may be lurking in unexpected places. “We look backwards across categories and try to see how do trends move between categories from hair care, to skincare, to oral care,” Pappas said. “Just doing that puts you a decade or more ahead of the curve.”
Question novelty
AI also facilitates deeper insights by helping users arrive at novel, “category jumping” questions — the gold standard of innovative inquiry — that apply understanding from one area to a completely different space. Our research shows that AI led respondents to ask unique questions that changed the direction of their team, organization, or industry 75% of the time.
When you know a technology can sift through much more data, and connect more dots, than you could ever do alone, it gives you license to ask wilder questions — things you would never ask if you had to answer them on your own, because they are intractable for the human brain or somehow go against entrenched cognitive biases.
While category-jumping questions will not arise in every encounter with AI systems, being open to the possibilities and allowing for freedom of inquiry can pave the way for more instances. Here’s how Mir Imran, a medical innovator and founder of InCube Labs, described the upside when we spoke: “AI can take really obscure variables and make novel connections. When these hidden connections come together, it causes you to reframe your question and deliver disruptive innovations.” In other words, AI’s novel connections can spark your novel questions, which in turn can lead you to investigate solutions others haven’t dreamed of yet — like the robotic pills that Imran’s team recently created to replace external injections with internal ones.
Creating conditions for better questions.
AI can take leaders out of their usual mode of operation and force them to cede control over where their questions will take them. That’s a good thing. Increased question velocity, variety, and especially novelty give facilitate recognizing where you’re intellectually wrong, and becoming emotionally uncomfortable and behaviorally quiet — the very conditions that, we’ve found, tend to produce game-changing lines of inquiry. Jeff Wilke — former CEO of Amazon Consumer Worldwide, now a cofounder of Re:Build Manufacturing — has embraced these conditions not only in his day-to-day work as a tech executive but also throughout his career, continually revising his mental models while moving from role to role. When we spoke, he had this to say: “If you seek out things that you don’t know, and you have the courage to be wrong, to be ignorant, to have to ask more questions and maybe be embarrassed socially, then I think you build a more complete model, and that model serves you well over the course of your life.”
But there’s a hitch to teaming up with AI: Research suggests that it can be challenging for people to do so congenially because AI’s superhuman capabilities and unpredictable moves may prevent them from fully trusting and engaging with the technology. That tracks with what we’ve observed in organizations and learned from our conversations with leaders.
Distrust of the technology is hardly conducive to creative inquiry. So, look for ways to offset that, and don’t just leave it to AI to produce the conditions for breakthrough thinking and problem-solving. Consider how else you might create them. Where is there room in your problem-solving processes for synthesizing things that don’t seem related? How might you use those opportunities to throw people off balance so they’ll generate questions that reach beyond what they intellectually know to be right, what makes them emotionally comfortable, and what they are accustomed to saying and doing? At the same time, how can you create psychological safety for people in your organization to ask far-ranging questions and to use AI more effectively to learn from them, ultimately leading to asking better questions? When psychological safety is present, people can say, without repercussion, “I am wrong,” “I am uncomfortable,” and “I am still thinking”?
Rather than neatly resolve all those tensions, leaders and teams must learn to sit with the uncertainty that comes from asking questions that take them into new territory. While the process isn’t easy, the results are exciting, which is perhaps the most important benefit of collaborating with an AI system. Excitement provides momentum and motivation to push through a tough process, fueling further creativity.
Mitigating AI’s Weaknesses with Human Strengths
Artificial intelligence may be superhuman in some ways, but it also has considerable weaknesses. For starters, the technology is fundamentally backward-looking, trained on yesterday’s data — and the future might not look anything like the past. What’s more, inaccurate or otherwise flawed training data (for instance, data skewed by inherent biases) produces poor outcomes.
Leaders and their teams must manage such limitations if they are going to treat AI as a creative-thinking partner. How? By focusing on areas where the human brain and machines complement one another. Whereas AI increases the volume of data we can process and the degree of complexity we can manage, our brains work in a reductive manner; we generate ideas and then explain them to other people. Whereas machines lack imagination and moral judgment, we can tap those critical skills as AI helps us increase the velocity, variety, and novelty of the questions we’re asking to solve problems in our organizations. Such differences are the stuff of fruitful collaboration — and optimizing them can reduce the threat of AI to human labor.
With humans and AI working to their respective strengths, they can transform unknown unknowns into known unknowns, opening the door to breakthrough thinking: logical and conceptual leaps that neither could make without the other. Harnessing this potential will require leaders to look at artificial intelligence in a new light — one that is less about cost savings, efficiency, and automation and more about inspiration, imagination, and innovation. It will also require building a culture that supports, incentivizes, and rewards asking big questions — and not necessarily knowing the answers.
