Updated May 28, 2026
Why P&G Got 39% Better Outcomes From AI (And Most Teams Get 5%)
P&G's field study found AI-paired teams produced 39% better work on the same tools you already pay for. The difference isn't the software. It's onboarding.
The AI divide isn’t coming. It’s already here, and it’s wider than most leaders think.
While a lot of executives are still debating whether to embrace AI, P&G ran a field experiment with their own teams and found something that should end the debate: people working with AI produced roughly 39% better outcomes than people working without it. One person with AI matched what a traditional two-person team used to deliver.
That’s not a rounding error. That’s a team-size advantage hiding inside the same headcount.
Yet in my conversations with CEOs and marketing leaders, I keep hearing the same thing:
“We bought everyone licenses for ChatGPT, Claude, and Gemini. We’re still not seeing the return.”
Sound familiar?
Here’s the uncomfortable part. Your competitors aren’t just using AI. The ones pulling ahead have figured out how to onboard it. Some teams scrape out a 5% bump and call it a day. P&G got close to a 40% lift on the same tools you already pay for. The gap between those two outcomes isn’t the software. It’s whether AI got treated like a tool you log into, or a teammate you actually trained.
What P&G actually found
The standout findings, in plain language:
- One person with AI matched a two-person team. Same deliverable, half the heads. That’s the closest thing to a free hire most marketing departments will ever see.
- Roughly 3x more likely to produce top-tier work. AI-paired teams landed in the top slice of quality far more often than the control group.
- Faster, with richer output. Work moved noticeably quicker, and most of the AI-assisted material survived into the final product instead of getting thrown away.
- The experience gap shrank. Junior staff working with AI performed at a level you’d normally expect from seasoned pros.
- Less anxiety, not more. Contrary to the tech-dread narrative, the people using AI reported higher enthusiasm and lower stress.
I keep coming back to that fourth one. Less-experienced employees, paired with AI, performed like veterans. If you’ve ever struggled to level up junior marketers fast enough, this is the closest thing to a shortcut that exists.
(P&G reported these numbers from an internal study. I’m citing the findings as the company shared them, not a public dataset, so treat the exact percentages as directional rather than gospel. The direction is what matters here, and it’s not subtle.)
What this means for your marketing team
AI is a new team member, not just software. You wouldn’t hand a new hire your logins on day one and expect senior-level output by lunch. AI is no different. The teams getting 40% are the ones onboarding it: brand voice, context, examples, guardrails. The teams getting 5% never got past “write me a cold email.”
Your talent hierarchy is flattening. A junior marketer with AI and a good process can produce work at the level of your senior team. That changes who you hire, how you train, and where the bottlenecks actually sit.
The work gets better and people get happier. The fear was that AI would stress people out or hollow out the job. The data points the other way: more enthusiasm, less anxiety. When the boring parts get handled, people spend more time on the parts they actually like.
This is the part of AI marketing ROI almost nobody measures. The return isn’t in the license. It’s in the onboarding.
Three ways to apply this this week
1. Create AI pairs
Put a junior and a senior marketer on a real project with shared AI access. Have them document where AI carried the load, where it fell short, and where the human still had to step in. You’re not just shipping the project. You’re building the playbook for everyone who comes after them.
2. Define what AI should and shouldn’t touch
Most teams stall because nobody drew the line. Write down what’s AI-assisted versus human-led for the work you do most: campaign briefs, first-draft copy, reporting, research. Then back each one with examples. Show the model two or three pieces that hit your bar and a couple that missed, and let it learn the pattern instead of guessing at your adjectives.
3. Start a shared prompt library
Pick one document everyone can reach and fill it with prompts that already work. The ones that carry your brand voice, your campaign frameworks, and the institutional knowledge that usually lives in one person’s head. This is how good output stops being a lucky accident and starts being the default.
Quick win: write its job description
The single fastest upgrade is to stop talking to AI like a search box and start briefing it like a teammate. Here’s a starter you can paste in and fill out:
You’re a senior AI marketing teammate for [BRAND] in [INDUSTRY]. Brand voice: [conversational / professional / playful]. Audience: [persona]. Current goal: [specific metric]. Key differentiators: [2-3].
For [campaign / content / analysis], generate [X options] that align with our positioning as [position], speak to the customer pain of [problem], and end with a clear call to action driving [desired action]. Present them in a table: Option, Key Benefit, Target Segment.
Save the filled-in version. Reuse it. That one paragraph, treated as a permanent brief instead of a throwaway prompt, is most of the distance between a 5% team and a 40% team.
The question almost nobody answers
How are you measuring AI’s impact on your team’s performance?
Most companies are stuck in the “it feels faster” zone instead of tracking anything concrete. That’s exactly why the return stays invisible and the budget conversation keeps stalling. The P&G results suggest that real measurement is worth every minute: output quality, cycle time, how much AI-assisted work survives to the final product. You can’t manage the 39% if you never put a number on it.
If you want the bigger picture on where AI actually pays off, the whole AI marketing practice comes back to the same idea. The money is in how you run the tools, not which ones you bought.
Get the onboarding playbook
I send one email every Friday for marketers who’d rather train their AI than just pay for it. Real teardowns of the onboarding docs, the tool-versus-teammate briefs, and the shared prompt libraries I run with clients, plus what actually moved the numbers. If “onboard it like a hire” landed for you, this is the rest of the build.