When Bristol Myers Squibb decided to rebuild its procurement function with artificial intelligence, it made a deliberate choice: move forward without waiting for perfect data.

"We fix the data as we go," said Rhonda Griscti, the company's executive director of digital strategy and global process lead, in an interview.

The strategy has paid off. According to Griscti, the supplier sourcing and evaluation process has accelerated significantly, with improvements in other areas as well.

"Our RFP (request for proposal) process has gone from an average of six to nine months to less than 30 days," she said.

Bristol Myers Squibb is rapidly advancing its AI-driven procurement transformation, even as data challenges add complexity to technology deployment on a broader scale.

The data readiness debate

A report released in December 2025 by global professional services firm RGP shows that only 10% of CFOs fully trust their data quality, and more than a third of CFOs cite data trust as the top barrier to realizing ROI from AI investments.

On the issue of data readiness, Scott Rottmann, president of RGP's consulting business, takes a more cautious stance than Griscti.

"If the data is clean and of sufficient quality to support good decisions, I think that's fine," Rottmann said in an interview. "There's an 80-20 rule here. You can't be perfect at everything. So, you get to 80% and then move forward."

Data preparation: how much is enough?

The divergence between the two viewpoints highlights a broader debate in enterprise AI adoption: under pressure to deliver AI results quickly, how much data preparation is sufficient?

Skeptics of rapid deployment warn that rushing ahead with inaccurate, incomplete, or inconsistent datasets to train or drive large-scale automated decision-making systems could introduce operational and governance risks. Proponents argue that early deployment accelerates time-to-value and delivers measurable operational improvements faster.

In an April article in Forbes, AI strategist John Sviokla argued that companies often overinvest in data cleaning, delaying value creation in pursuit of perfect data quality, and thereby missing opportunities to extract insights from imperfect data.

"You shouldn't put cleaning before processing," wrote Sviokla, co-founder of AI consulting firm GAI Insights and an executive fellow at Harvard Business School. "Instead, you should build processing capability and let it reveal what's worth cleaning."

The data lake: key to getting started

Griscti said a key step was building a data lake—a centralized repository for storing and processing vast amounts of raw data. "You don't need perfect data to start," she said. "You just need to know where the data is. Put it in the data lake, make it accessible, and then clean it as you use it."

Griscti joined Bristol Myers Squibb in November 2021 as senior director of R&D procurement and was promoted to her current role four years later. When she joined, the procurement function was already under pressure to speed up. Momentum accelerated further when a new chief procurement officer came in with a modernization mandate.

The company launched its procurement transformation project in June 2024, focusing on process standardization, capability building, and eliminating fragmented email-driven workflows.

Rapid AI deployment

The next phase was AI applications. Griscti and her team conducted a "soft launch" in November 2024 with Palo Alto-based Globality, a platform that uses AI to automate key parts of supplier sourcing and selection. The technology was then rolled out organization-wide in February 2025.

The result was a significant increase in procurement volume, with more than $1 billion in spending through the platform in the first year, far exceeding initial expectations. Griscti said the transformation also dramatically expanded the scope of procurement activities, with roughly 10 times the number of RFPs, while bringing previously outsourced work back in-house with about 50% fewer resources.

Insourcing freed up funds that had been paid to external service providers, which Griscti reinvested in technology.

Griscti's advice to other executives leading similar transformations: don't get bogged down by data issues.

"People will argue with me about this, and I've discussed it at many conferences," she said. "But if you wait for perfect data, you'll never start."