From Chatbot to Performance Review Engine: How General Mills Deployed AI to 20,000 Employees
From Chatbot to Performance Review Engine: How General Mills Deployed AI to 20,000 Employees
When General Mills first launched MillsChat to a pilot group of 900 employees, it was a writing and summarization tool — a helpful assistant, but nothing transformative. Two years later, that same tool processes performance reviews for 20,000 employees across four countries. The journey from chatbot to HR infrastructure offers a playbook every people leader should study.
Phase 1: The Writing Assistant (Pilot)
MillsChat began as an internal generative AI assistant built on Google's PaLM 2, deployed within Microsoft Teams. The initial use case was straightforward: help employees draft documents, summarize meeting notes, and speed up routine writing tasks. The pilot reached roughly 900 users — enough to prove adoption was real, but far from enterprise scale (CIO Dive).
Jaime Montemayor, General Mills' Chief Digital and Technology Officer, led the initiative alongside a cross-functional sub-group of senior leaders spanning HR, legal, finance, and supply chain. As Montemayor put it: "It takes a village to be successful with AI in any corporation" (CIO Dive).
That cross-functional governance structure turned out to be the enabling decision. It meant HR was at the table from day one — not bolted on after engineering built the tool.
Phase 2: Self-Service Knowledge Hub
As adoption grew, MillsChat evolved beyond writing assistance into a self-service knowledge platform. Employees began using it to answer operational questions, access internal policies, and bridge knowledge gaps during onboarding — reducing their reliance on email chains and meetings for routine information (Consumer Goods Technology).
For HR, this phase delivered immediate, tangible value. New hires could onboard faster by querying MillsChat instead of waiting for responses from overwhelmed people teams. Internal email volume dropped. Meeting time decreased. These are directional outcomes reported by the company — not vanity metrics, but signals that the tool was absorbing real workload (Food Dive).
The platform also transitioned its underlying model infrastructure, integrating Microsoft Azure OpenAI services alongside the original PaLM 2 foundation — a natural evolution as the company deepened its Microsoft ecosystem integration (CIO Dive).
Phase 3: Performance Reviews and Development Planning
This is where the story gets interesting for HR leaders. MillsChat scaled to 20,000 employees across the United States, Canada, the United Kingdom, and India — covering all non-plant locations. Roughly 30% of the workforce now uses the tool consistently (CIO Dive).
HR emerged as one of the top functions leveraging MillsChat. The tool's role expanded into company-wide performance reviews and development planning — activities that traditionally consume weeks of manager time each cycle. Instead of starting from scratch, managers and employees use MillsChat to draft review inputs, synthesize feedback, and structure development goals (DigitalDefynd).
The implication is significant: General Mills didn't buy a separate "AI for HR" point solution. It grew one organically from a general-purpose assistant. The same tool employees already trusted for daily tasks became the engine for their most consequential HR processes.
Phase 4: The Agentic Horizon
General Mills isn't stopping at chat-based assistance. The company's digital roadmap now includes agentic AI architectures designed to redesign entire workflows — not just answer questions within them. The vision moves from reactive assistance to proactive, multi-step task execution across business functions (Consumer Goods Technology).
For HR, this means the next iteration of MillsChat could autonomously prepare review packets, flag development gaps based on skills data, or coordinate succession planning inputs across departments. The shift from "tool you ask" to "system that acts" represents the frontier General Mills is building toward.
What HR Leaders Should Take Away
General Mills' MillsChat trajectory reveals a pattern that any HR team can replicate:
Start with a general-purpose tool, not an HR-specific one. Adoption comes from daily utility. Performance reviews come later — but only if the tool already has trust and traction.
Put HR at the governance table from day one. Montemayor's cross-functional leadership group ensured HR use cases were designed in, not afterthoughts.
Let the tool earn its way into high-stakes processes. MillsChat didn't launch as a performance review engine. It became one because employees were already comfortable using it.
Plan for the agentic next step. Chat-based assistance is the starting line. Workflow redesign — where AI executes multi-step HR processes — is where the real efficiency gains live.
The companies that will lead in AI-powered HR aren't necessarily the ones buying the most sophisticated point solutions. They're the ones, like General Mills, that plant a general-purpose seed and let it grow into their most critical processes.