Can AI make brainstorming less mind-numbing?
Thrilled to share this fantastic piece featuring my AI research with Kian Gohar, written by Andrew Hill in the Financial Times
We have all been there: a company-wide brainstorm, a high-level “ideation” session, a — lord help us — “strategy jam”. Tables of 10, a flock of flip-charts and a pile of Post-it notes on which an over-caffeinated moderator invites us to scribble our priorities for the coming year. A week later, a fat attachment circulates by email, only to sink rapidly into an inbox brimful of the day’s more urgent demands.
Mercer chief executive Martine Ferland organised just such a gathering for 150 leaders of the human resources consultancy recently. But instead of sticky notes, she asked them to tap their priorities into iPads. “By the time the facilitator went onstage, we had a summary created by AI,” Ferland told me recently. The 15 tables then refined the outcome with follow-up prompts relevant to the industry sector they served.
The work is one example of how companies are applying generative AI. As Stanford University’s Erik Brynjolfsson told delegates at the World Economic Forum summit in Davos last month, this is the year gen-AI “gets a body and starts to do things — transforming the world, transforming work, transforming productivity”.
First, though, companies will need to transform the workers themselves, persuading them, for example, to unlearn some old brainstorming ways and work out how best to collaborate with the help of this novel and powerful tool.
Gen-AI’s potential to revolutionise leadership away-days is not exactly world-changing. But it fits neatly into the box almost every executive I met at Davos claims they plan, for now, to confine gen-AI to: a job-improving “augmenter” rather than a job-destroying “disrupter”, a revenue enhancer rather than a cost reducer.
AI will replace some roles, Ferland believes. But it will extend others, helping to “bridge the talent shortage” and make better use of the time available.
Other chief executives I’ve talked to recently boast about how gen-AI is already assisting on tasks: speeding up due diligence for investments, searching and screening job candidates, onboarding and training new recruits, prompting customer service staff to find the right answers and pose the right questions.
Gen-AI’s instant accessibility and usability mark it out from previous top-down waves of technology such as digitisation. “I see less obstacle [to general use] with Gen-AI than all the other technology we’ve been dealing with for decades,” said Barbara Lavernos, deputy chief executive of cosmetics company L’Oreal.
The lack of friction also poses challenges, however. Creativity expert Jeremy Utley, also at Stanford, says the “text-box” interface of apps such as ChatGPT invites queries from users, much like a Google search. That “predisposes someone to treat AI like an oracle” rather than a creative collaborator, taking some of the steam out of brainstorming.
Utley and Kian Gohar have just published a working paper about how corporate innovation teams are using the new tool in problem-solving tasks. Most, they found, were failing to make the most of AI’s help and falling short of their own problem-solving potential.
Teams using AI tended to produce fewer really good or really bad answers than unassisted human peers, for instance. Those teams that did produce A-grade answers with AI’s help found it harder work than those who settled for “good enough” ideas. The underperforming teams tended to use the AI chatbot as a problem solver, rather than a conversation partner.
Gen-AI has been likened to a “summer MBA intern”, albeit one with far more power and knowledge at its disposal. But, Utley says, if a manager gave an intern one sentence of instruction and received a terrible report, “the problem wouldn’t be the intern, it would be the manager”.
Gohar and Utley suggest a few ways to improve. Ask humans to think through a specific problem alone before interacting with the AI. Engage a neutral facilitator to help navigate the final phase of organising ideas. Above all, as one executive told them, to reach the most creative outcomes, “place the emphasis on the chat, not the bot”.
This isn’t, says Brynjolfsson, “the time to sit back and wait.” Companies should prepare the ground and tread carefully, however. At logistics group DHL, management board member Nikola Hagleitner says she is having to encourage front-line delivery workers to embrace AI’s potential, while holding back managers tempted to move too fast: “We have to pace one group and take the fear away from the other group.”
In theory, the technology should increase the volume and variation of ideas generated by pairing humans and AI. But Utley says AI can simply amplify brainstormers’ underlying cognitive biases, such as accepting the chatbot’s first plausible answer. Unless the people in the loop change how they work, rather than stimulating innovative strategy, gen-AI will accelerate poor practice and bad outcomes — more storm than brain.
(This piece is written by Andrew Hill in the Financial Times)