P&G/Harvard Business School AI Study [new research]
Last week new research was published by Harvard Business School following a study carried out last year with professionals from Procter and Gamble.
If you are not familiar with P&G they are one of the world’s largest consumer packaged goods companies behind brands like Olay, Oral-B and Gillette.
The study involved 776 P&G employees from both the US and Europe, with them coming together for in person workshops to solve real world tasks such as developing new product ideas, packaging, and retail strategies.
Some of the employees worked independently, some in teams. Some were provided with ChatGPT 4 or 4o, and some were not.
The results were striking and should impact your organisational decision making and strategy.
Prefer to watch instead of read?
Individuals with AI outperform teams without
In this first chart the base line (0) shows the quality of an individual’s work without AI.
The blue bar shows what we’ve always known - that bringing a team with different perspectives together improves the quality of the output (two heads are better than one).
But look at the yellow bar - an individual with AI dramatically outstrips the performance of the team without AI, and a team with AI is only marginally better than the individual with AI.
Consider what this means in terms of costs. For the cost of a $35-$65 per month licence each individual with AI has a better quality output than the cost of surrounding everyone with a team.
For a company of any size, adding AI to each employee allows you to act more effectively than a company ten times your size without AI.
Teams with AI create the best quality solutions
Before you jump on the first chart and start looking at who you can let go, take a look at this second chart.
During the workshops the teams had to submit their ideas and solutions to the senior leaders present who assessed them for quality.
When looking just at the top 10% of solutions in terms of quality we see a different picture.
Now an individual with AI is less capable than the team without - likely because of the lack of diverse perspectives.
But look at the team with AI - they are disproportionately represented in the highest quality solutions.
Giving an existing team access to AI dramatically improves their performance compared to a team without AI.
AI removes knowledge siloes
As humans we only know what we know. As an individual when posed with a problem, our own lived experience heavily influences the solution we put forward.
In this chart, on the left you can see that when commercial teams and R&D teams were assigned the same problem, the nature of their solution was heavily weighted on their own background.
Commercial teams more often came up with market-based solutions - branding, packaging, pricing.
R&D teams came up with more technical solutions - compounds and molecules
But look on the right, when an individual was provided with AI, these differences disappeared and the solutions coalesced around an optimal solution.
This is a great demonstration of AI removing the knowledge barriers. Legal teams can incorporate financial knowledge, marketing can incorporate sales, product development can incorporate customer support.
AI-enabled employees are happier
The final chart I’ll share looks at the experience of the employees using AI.
A pessimist might think that employees fear AI’s arrival, but the study refutes that.
On the left you see the increase in positive emotions - with individuals and teams using AI outstripping those without.
On the right you see the reduction in negative emotions - again with individuals and teams using AI outstripping those without.
Consider what creates negative emotions at work - typically this is having to wait on someone else which slows you down, or your lack of knowledge outside of your domain making you feel helpless.
AI removes this.
Consider this report when planning your corporate strategy
For me the outcome of this report is simple.
Give every one of your employees access to a Team or Enterprise account of Claude, ChatGPT or Gemini right now.
It will have an outsized impact that will dwarf the $30-$60 per user you will spend on it.
For more information, I encourage you to read Ethan Mollick’s blog post on the report, which also links to the full report itself.
Click the report to view
Access Ethan’s blog post and link to the full report
Get Started
Whenever you are ready there are three ways I can be helpful:
Model 101 Playlist: 20 minute guides to using ChatGPT, Claude, NotebookLM for work
AI Inspiration Briefing: Show your people the path forward in this 60 minute live session
Kowalah: Buying platform to help you pick the right AI tools for your business