8 ways to use generative AI to improve decision-making

Bill Gates recently wrote about the inevitable transformation of software from the individual, narrowly focused apps we use today into something much broader in scope and more powerful.

For example, today we go to specific apps to execute specific tasks, but soon “you’ll simply tell your device, in everyday language, what you want to do,” writes Gates. “And depending on how much information you choose to share with it, the software will be able to respond personally because it will have a rich understanding of your life.” 

What Gates is describing is a Gen AI-driven agent.  You might think of a Gen AI agent as a super-intelligent and uber-productive assistant that can be a powerful aide in decision making. 

For example, a Gen AI agent might make recommendations or point to areas of potential concern – like helping medical professionals analyze images to make a diagnosis, or helping companies quickly flag critical information in the thousands of pages of due diligence documents during an acquisition (see Hebbia.ai).  It can also help close the data-insight gap in time-critical, high-pressure situations by using predictive analytics across large and complex data-fields to flag potential risks much faster than humans can (HBR, 10/26/23).

Gates is also optimistic about how these agents will fundamentally transform how we make decisions related to our health. “The real shift will come when agents can help [us] do basic triage, get advice about how to deal with health problems, and decide whether [we] need to seek treatment."

When making a tough decision, one bit of advice that’s never gone out of style is to talk it over with a friend, coach or mentor–someone with the experience, wisdom and good judgment to provide a fresh perspective and sound advice. Having the right person available in a timely manner isn’t always feasible, but AI doesn’t have a life, a family, a job or other distractions. In a recent Harvard Business Review article (HBR, 8/24/23), the authors describe how ChatGPT can be an “increasingly intelligent conversation and sparring partner” during the process of making a decision.

For example, it can be helpful in coming up with additional options that we may not think of or easily – which we argue is a key skill in good decision-making (see our prior blog post, "Are you considering too few options?").

When asking it a question about strategic options in a particular decision, AI may be able to quickly offer up many of the strategies documented on the internet that you might not have considered, with information about the success or failure of each. Or when evaluating decision-making alternatives, one can ask Gen AI, “What mistakes do managers of medium-sized x-type companies typically make when they decide to expand into new markets?”

Articulating your decision question to Gen AI can also have a similar beneficial effect as the “rubber ducky technique” software engineers are taught to use to think through a coding problem.* Plus, you also get back ideas you might not have thought of.

Each week there are hundreds of activities and small decisions that Gen AI can manage much better than humans. 

By doing some of the less-value added tasks like helping lawyers draft initial documents, helping real estate agents write new listing for homes, helping all of us schedule appointments, enter data, conduct research, etc., time is freed up for reflection, critical thinking and the other uniquely human aspects of decision-making that lead to true innovation.

Making a quality decision is one thing, but a good decision is only as good as its follow-through.  AI can help decision execution by, for example, transcribing recordings of a meeting and generating a summary of the main points discussed, identify action items, clarify decisions made, and send out a summary to everyone, essentially locking in the decision.  (see for example, Fireflies.ai, Gong.io, Chorus.ai, Otter.ai)

Imagine having a Gen AI-powered archive of all the meetings in your company over the last 5 years (see #4 above).  Then, when facing a big decision like entering a new market or shifting your distribution strategy, you can query the Gen AI with questions such as:  Have we encountered this problem or decision before, and what did we do about it? What discussions did we have about it?  What did various people say at the time, e.g., the VP of Product, CEO, VP of Sales, etc.? 

We can’t be in two places at once … oh wait, maybe we can.  Sam Altman, the venture capitalist and CEO of OpenAI – creator of ChatGPT, frequently holds “office hours” for the founders of startups in which he is an investor.  For years he has dreamed of creating a “bot” that would replicate his advice and thought process to free him from his office hour duties. 

Recently, Altman stood in front of an audience at an OpenAI developer conference and created, on-the-fly, a CustomGPT to stand in for himself at these meetings.  He instructed the GPT to “help startup founders think through their business ideas and get advice, and then, after a startup founder has gotten some advice, grill the founder on why they are not growing faster.” 

To train the Gen AI he uploaded transcripts of lectures he had given to startups.  The entire process took less than two minutes. In short, when it comes to routine meetings and transactional conversations, sending an AI to represent us, instead of the real thing, can save time, conserve energy, and free up time to focus on conversations and dialogue where our presence matters most.

Have you ever wanted to run a product decision by Steve Jobs, or even a political decision by Abraham Lincoln? OpenAI recently released a new feature, “custom GPT” that empowers users to build a custom GPT based on the writings or transcripts of an individual person. For example, suppose you are an investor and you would like to ask Warren Buffett how he values a company.  By building a custom GPT based on all the writings of Buffett, we can create an AI that essentially models Buffett’s philosophy on investing. 

Before making a big decision, it’s always a good idea to ask, “What could possibly go wrong?” To combat the ever-present Overconfidence Decision Trap (one of the traps we address in Decision Mojo) we can ask Gen AI to generate opposing arguments, uncover disconfirming evidence, and list reasons why our decision may not work out as planned.   

Gen AI can play the role of devil’s advocate – or what in Decision Mojo we like to call the “loyal opposer” – by surfacing uncomfortable counterfactuals and scenarios that our overly optimistic brains don’t want to consider and may have subconsciously avoided. 

Since Gen AI is not influenced by the power of group think, nor burdened with the psychological need to be accepted by the team, it is uniquely qualified to be the objective participant that every good decision process needs.