Bristol Myers Squibb has rebuilt its procurement process around an agentic AI platform from Globality, cutting the average time to run a request for proposal from six to nine months to under a month. The company says it now runs about 10 times more sourcing events with roughly half the resources, and processed more than $1 billion in spend through the platform in its first year against an initial target of $100 million. The figures come from BMS and the vendor.
How it works: Under the old process, a buyer spent weeks gathering requirements, drafting the RFP by hand, contacting suppliers, and standardizing proposals that came back in different formats. Globality’s AI agent, called Glo, now handles most of that work across five stages:
Scope: The buyer describes the need in plain language. The agent runs a structured interview and drafts the RFP, instead of the buyer writing it.
Select: The agent returns a ranked list of suppliers from BMS's approved panel and Globality's vetted database, and the buyer chooses whom to invite.
Analyze: Suppliers respond through a standardized template, so the agent can compare bids on total cost and flag risks without a manual spreadsheet.
Negotiate: The agent runs multi-round electronic negotiations, with the buyer setting the strategy and approving each round.
Buy: The agent drafts the statement of work from the winning bid and guides the buyer through compliance checks.
In effect, the buyer stops writing documents and instead supervises an agent that produces them. BMS credits that shift for shortening the cycle and increasing the volume its team can manage.
How BMS rolled it out: The company made two deliberate choices. It started with sourcing rather than an intake or workflow layer. Rhonda Spraker Griscti, executive director of agile sourcing, said adding intake on top of an old sourcing process “simply funnels demand into the same outdated processes.”
BMS also started before its procurement data was fully organized. “You don't need your data to be perfect to start. You just need to know where it is… and you can clean as you go,” Griscti told CFO Dive, addressing one of the most common reasons such projects stall.
BMS is among the first companies to hand an AI agent the full sourcing cycle rather than a single step. Others are automating parts of it. SAP has embedded a bid analysis agent in its Ariba platform, while Walmart uses Pactum to run automated supplier negotiations.
Dig deeper:
Agentic AI and Procurement (Part 3): What Agentic AI Can Actually Do — The Hackett Group — what separates genuine agentic systems from standard automation across 11 procurement stages, with a benchmark: only 4% of teams that piloted Gen AI reached scaled deployment.
Emergence of multiagentic AI in sourcing and procurement — Deloitte — how the orchestrator-agent architecture scales from a single sourcing agent to an enterprise-wide system.
