The U.S. Office of Personnel Management (OPM) gave every employee access to six commercial AI tools, according to a MeriTalk article. OPM Chief Information Officer Adam Starr announced this on September 16 at the GovExec Government & AI Summit in Washington, D.C. 

Starr said the multi-model strategy was to give employees several AI platforms to choose from so they can shift to whichever tool works best. He said OPM used the General Services Administration’s (GSA) governmentwide OneGov agreements to get the tools at $1 and roll them out fast.

The six tools are Google’s Gemini, Microsoft Copilot, xAI’s Grok, OpenAI’s ChatGPT, Anthropic’s Claude, and Perplexity AI. More than 70% of OPM’s workforce now actively uses at least one of the approved tools, based on the agency’s telemetry data.

Behind OPM’s Multi-Tool Approach

Starr said the strategy is deliberate. He wants employees to have several options, since AI vendors keep leapfrogging each other with new capabilities. That way, staff can move to whichever tool works best as the technology changes. It also puts pressure on vendors to keep competing on price and performance for OPM’s business.

The OneGov agreements matter most for smaller agencies like OPM, which lack the purchasing power of larger departments, Starr said. GSA struck the original OneGov deals to let agencies access leading AI platforms at little to no cost.

GSA is moving to the next phase of its OneGov arrangement with OpenAI. Starting October 1, the new 27-month agreement drops platform-access fees entirely and gives agencies a 50% discount on token-based usage for covered ChatGPT models. 

It replaces the initial $1-per-year pilot pricing, which expires September 30. Starr said the updated pricing will help OPM keep giving employees access to advanced tools going forward.

Is a Multi-Model Strategy Better Than a Single Model?

OPM’s approach to multi-vendor AI procurement raises certain questions for chief data officers (CDOs). When an organization hands every employee six different AI tools at once, oversight becomes more difficult. Each tool typically comes with its own data-handling practices, output quality, and risk profile.

This is where AI governance becomes essential. AI governance means having clear policies for which tools employees can use, what data they can put into each one, and how outputs get checked before they’re relied on. 

A multi-tool strategy can deliver real benefits, such as letting teams pick the best tool for the job. But those benefits depend on governance keeping pace with adoption.