When ChatGPT first came out, everyone Cheryl Strauss Einhorn worked with wanted to talk about the tools. She saw a bigger question: “what is our role as decision makers when the tools are in the room?”
Einhorn began her career as an investigative business journalist, built her own problem-solving system to check her assumptions, and has since written four books on decision-making, including The Human Edge. On October 14, 2026, we hosted her class “The human edge: smarter decisions in the age of AI” at the Nasdaq Entrepreneurial Center, with Allison Berardi, program coordinator for high growth impact entrepreneurs at the Center, as host. Here is how she says founders can use AI without handing over their judgment.
Key takeaways
AI asks for some of your thinking, and it answers faster than you can process. Einhorn warned that critical thinking is a muscle that weakens if you stop using it. So her first rule is simple. “AI knows nothing about you.”
Before you type anything, answer three questions:
With those answers in your prompt, the tool stops giving you general, popular answers and starts answering your actual question. You can also ask AI to act as a teacher: share your three answers and ask it to question you back, so you know what you are missing.
You stay responsible for the outcome. “When the hammer falls, it falls on you,” she said. “AI doesn’t care about the consequences.”
AI answers sound confident. “Almost everything it gives you is plausible. But that doesn’t mean it’s accurate,” Einhorn said. Her starting assumption is blunt: “you should always assume that something is wrong in the information.”
She learned this firsthand. Looking for a citation, she asked an AI tool for a source. It named a respected researcher she knew, with a plausible paper title in a plausible journal. When she clicked the link, “the journal didn’t exist. The researcher had never written on that topic either.”
To strength-test any answer, ask:
“There’s a difference between easy and meaningful,” Einhorn said. When a decision feels hard, that friction is useful. It is a signal to look inward for what you have not told the tool, and to use the tool to push back on your own ideas.
AI tools tend to flatter. They tell you your idea is good or that you are on the right track. Her rule is never to accept the first answer, and never to use AI “as a transaction.” Instead, “it’s always a conversation.”
Watch for bias on both sides. People often ask AI to confirm what they already believe, and AI tries to be helpful by agreeing. People also tend to favor information from experts. AI brings its own biases, from the data it was trained on and from leaning toward popular, frequently repeated answers. Ask it how its information might be outdated, wrong, or easy to disprove.
AI can bury you in information. Einhorn’s fix works like browsing for a show on a streaming service. You start broad, such as a crime show, notice an option you like, such as one in French, and narrow down. Then you narrow again, such as only 30-minute episodes, until you have what you need.
She calls a version of this progressive prompting. Say you are thinking about opening a store in Baltimore. Start by asking about the city’s real estate and demographics, then zero in on the neighborhoods that fit the customers you want to serve.
Along the way, ask yourself what you would actually do with each answer. Researching a new client’s background, for example, will not tell you what that client is likely to ask you on the call. If an answer would not change your next step, skip the question.
AI does not need encouragement, Einhorn said, but it does need boundaries. Tell it what is out of bounds, and share why you are asking. “Incentives drive behavior,” she said. If you are hiring, AI will not know whether you need a seasoned manager who can lead a team on day one or a newer employee you plan to train, unless you tell it.
Her most practical advice for founders already using AI:
For creativity, use AI to jog ideas, then improve on them. “AI doesn’t come up with things that are that original,” she said. You can also ask it to question you from the perspectives of different well-known leaders.
Einhorn shared the story of a bike shop owner who asked an AI tool where to open a second location. It suggested a space about an hour away, in a busy area that was not safe for cycling. After she walked him through the problem definition, motivation, and key factors, he asked again. This time he found options about 20 minutes away, at ground level, visible to people on bikes, with parking and within his budget. He knew all of that already. He just had not brought it forward.
The same thinking applies across all eight moments of complex problem solving she teaches: defining the problem, naming your motivation, setting the context, directing the research, making meaning from the data, challenging bias, including stakeholders, and coming to conviction. In each one, only you can decide what is true and what matters. She called this self-awareness, or “thinking about thinking,” a competitive advantage.
Before your next important AI prompt, take two minutes and write down three things: the problem, your motivation, and the key factors. Then, when you get the answer, ask the tool how it could be wrong.
Some decisions deserve a human sounding board, too. A free session on Mentor Makers, the Center’s mentor-matching platform, can help you pressure-test a big call on business development, fundraising, hiring, or partnerships with someone who has faced it. For more free sessions, browse the Center’s Community Classes.