When most large companies discovered employees were secretly using ChatGPT, they said "don't use it." Colgate-Palmolive went the opposite direction: "We'll build you a safe environment to use it."
What is this?
Colgate-Palmolive, founded in 1806, sells products in over 200 countries. From Colgate toothpaste to Palmolive soap — it's one of the most ubiquitous consumer brands on the planet.
When ChatGPT launched in late 2022, many companies either watched from the sidelines or banned its use. Colgate-Palmolive was different. By mid-2023, they appointed Kli Pappas as Global Head of AI and started building an enterprise-wide AI strategy for their 34,000 employees.
The result? The company built an internal AI Hub, turned thousands of employees into builders, and ran a company-wide education program — becoming a model for enterprise AI adoption. Wharton professor Ethan Mollick called it "a really good example of the benefits of having a dedicated AI Lab inside of large companies run by a company veteran who actually gets what these sorts of tools can do".
What changes?
There's a common problem most companies face with AI adoption. According to Microsoft's 2024 Work Trends Index, 75% of global knowledge workers were already using AI, and most were using their own tools because their companies didn't provide any.
Kli Pappas confronted this reality head-on.
"The technology is available to everyone, it's free, and it's super powerful. If you aren't putting it in the hands of your employees, they could use it anyway and create risk for your organization."
The contrast between Colgate-Palmolive and traditional enterprise AI approaches is stark.
| Traditional approach | Colgate-Palmolive approach | |
|---|---|---|
| AI policy | Ban or restrict usage | Design guardrails, then open access company-wide |
| Builder scope | IT and developers only | All employees, including frontline experts |
| Training | Optional online courses | Mandatory training + local ambassador network |
| Governance | Centralized control | Platform-level guardrails + field autonomy |
| Risk framing | "Adopting AI" is the risk | "Not adopting AI" is the risk |
The most critical shift is redefining risk. While most companies see AI adoption itself as risky, Colgate-Palmolive views incrementalism — slowly moving and hoping you'll get there — as the real danger.
"Companies are letting the frog get boiled. I think a lot of companies are stuck in this incrementalism phase where they just feel like things will move and 'we'll get there.' And I don't think that's how AI is going to play out at all."
The playbook: how to build an internal AI Lab
Here's the five-step approach Colgate-Palmolive actually executed. It's a practical playbook any company can reference.
Step 1. Build an internal AI Hub
Colgate-Palmolive built a dedicated company AI Hub. This isn't just a ChatGPT license — it's an integrated platform where employees can build, test, and deploy their own AI assistants.
They used OpenAI's Assistant API to enable custom GPT creation, with a staged approval process from personal use to small-group sharing to company-wide deployment.
The result? 3,000–5,000 custom assistants were built in just 18 months. Most were for individual or small-group use, but about 10% were deployed to entire business lines.
Step 2. Design guardrails (not bans)
Core principle: "Systems need to be configured in a way that the happy path is the easiest path."
Colgate-Palmolive's guardrail framework includes:
- Role-based access control (RBAC) — platform-level control over who can build and deploy what
- Secrets management — centralized handling of API keys and sensitive data
- Use-case risk classification — clear distinction between "asking ChatGPT for PowerPoint tips" (low risk) and "using AI to screen resumes" (high risk)
- Governance workflows — staged approval from personal experimentation to company-wide deployment
This works because new builders don't need to become security or policy experts. The platform handles the guardrails so they can focus on solving problems.
Step 3. Design company-wide training
Colgate-Palmolive's training has three layers:
The messaging is "Super You" — AI isn't replacing you, it's making you stronger.
Why the insistence on universal training?
"If half the team knows what technology can do for it and the other half doesn't, that's a non-functional team. Imagine half the people know what Google Sheets is and half don't. Your sequencing, role assignment, time estimates — everything falls apart."
Step 4. Turn domain experts into builders
Colgate-Palmolive's most distinctive strategy: instead of IT building solutions and pushing them out, they let the people closest to the problem build the solution themselves.
A compelling real-world example: a manufacturing facility manager in Germany used an LLM to instantly translate technical manuals from German to Greek and answer operational questions in real time. This would have taken weeks through IT.
Charter reported that Colgate-Palmolive also built a feedback loop for employee-created GPTs. When colleagues use a GPT enough times, they get a survey asking "how much time did this save?" If it saved 200 hours in a month, it's worth investing in. If nobody uses it, it's deprioritized.
Step 5. Focus on strategic domains
Alongside company-wide adoption, Colgate-Palmolive concentrated AI investment on three strategic pillars:
- Marketing — generative AI for concept and content creation
- Innovation — new product development. Using ML to analyze consumer search data, identify unmet needs, and feed them into R&D
- Operations — supply chain efficiency, demand forecasting, preventive maintenance. AI simulates scenarios and flags risks, but humans make final decisions
The future roadmap? Multi-agent architecture. Individual agents that work effectively on their own while communicating agent-to-agent. Process owners build their own agents that collaborate to handle increasingly complex tasks.
Going deeper
Professor Ethan Mollick proposed "Leadership, Lab, Crowd" as the formula for successful AI transformation. Colgate-Palmolive's case maps perfectly to this framework:
| Element | Mollick's framework | Colgate-Palmolive in practice |
|---|---|---|
| Leadership | Articulate why AI is urgent | Appointed Kli Pappas, framed inaction as the real risk |
| Lab | Dedicated team explores + builds + benchmarks | Internal AI Hub, focused on 3 strategic pillars |
| Crowd | Employees apply AI to their own work | 3,000–5,000 custom GPTs, ambassador network |
What's particularly noteworthy is Mollick's core insight: individual AI productivity gains don't automatically translate into organizational gains. Even if individuals become 3x faster with AI, without changes to processes, incentives, and work structures, companies see only "moderate improvements." That's precisely why Colgate-Palmolive didn't stop at "giving people AI tools" — they built governance, education, and feedback loops around them.
