This FAQ covers the most common questions procurement leaders ask about adopting AI when their data isn't perfect, answered by experts from Bain, Hershey's, and Sievo.
Our recent webinar on AI-ready procurement opened with a live poll: how automated and integrated are your procurement data flows? Almost nobody picked "fully automated." Most picked "ad hoc."
That gap is the problem. AI is moving faster than most procurement data flows. Teams know their classifications are messy, ERPs don't fully talk to each other, and the same supplier shows up under three different names. The longer they wait for clean data, the further behind they fall.
The takeaway across all three views: AI in procurement is only as good as the data feeding it. General-purpose AI guesses. AI trained on procurement scale, like Sievo IQ which learns from more than 2% of global GDP in spend, understands the domain.
The webinar brought together three views: Brian Murphy (Partner at Bain & Company) on what leading clients are doing differently, Syed Naqvi (Sr. Manager of Procurement Data & Technology at Hershey's) on building a procurement Center of Excellence, and Johan-Peter Teppala (Chief Strategy & Transformation Officer at Sievo) on the data side. The audience pushed back with the questions you'd expect. What if our data is bad? Where do we start? How do I sell this internally? Below are the seven that came up most.
No. The leaders aren't waiting, they're working with what they have.
Brian put it bluntly: "I've never walked into a client of mine that had perfect data." Procurement data is structurally messy because of multiple ERPs, mergers and acquisitions, manual data entry, and inconsistent supplier names. Your data will always have rough corners.
Syed set the threshold: "Data will never be 100%. If it's 97 or 95%, it's good enough for our buyers to go in and direct category strategies, risk analysis, and project pipelines."
External benchmarks also de-risk the imperfect data problem. Sievo's Community Data covers more than $2 trillion in cross-customer spend and 11 billion data rows per year, so your decisions are grounded in market reality, not just your own messy data.
The right question isn't whether your data is perfect. It's whether you're getting enough signal from it to act.
First, AI on imperfect data still surfaces useful insights, especially when humans validate before acting. Brian: "Today we're in a heavy dose of procurement in the loop. The team still needs to apply their judgment." Trust but verify.
Second, AI improves your data quality over time. When embedded in your procure-to-pay flows, it catches classification errors, duplicate suppliers, inactive vendors, and master data inconsistencies. Sievo IQ, for example, delivers 98% classification coverage and 94% accuracy at the deepest taxonomy level. That's proof AI doesn't just consume data, it can help improve it. Syed: "It's like cleaning a swimming pool. It'll change very quickly, trust me."
Build the data foundation first, then layer AI on top of internal data plus external market context.
The real shift isn't faster internal analysis. It's combining your internal spend data with peer benchmarks, price indexes, and external market signals, so your team can answer "are we paying competitively?" and "where should we act next?" not just "what did we spend?"
Practically:
The point is to get your team interacting with data through AI, not waiting for a perfect dashboard.
Frame the conversation around signal versus noise. There's always going to be noise in your data. Your team's job is to act on signal, not chase clean data.
First, set up an AI Council. Brian recommends a regular cross-functional touchpoint where leaders share specific examples of how their teams are using AI. The use cases that work go into a shared library and become standard practice. Adoption builds through visible wins, not abstract mandates.
Second, reframe the expectation. Imperfect data hasn't stopped your team from getting work done today. They're already applying judgment to messy inputs. AI doesn't remove uncertainty, it just compresses the time between asking a question and getting an informed answer.
Don't push AI on your team. Pull it in.
Syed: "You wouldn't have to push it. There is a lot of pull in all organizations to use and embrace Sievo data in copilots." Most procurement professionals are already curious and using AI informally. Your job is to channel that curiosity into structured adoption.
Practical mechanics that work:
Tactical work automates first. Strategic and interpretive work gets more valuable.
Brian's projections:
What gets more valuable: prompt engineering, data science fluency, judgment, influencing, change management, AI governance.
What gets less valuable: the tactical execution work that AI absorbs.
If you're hiring or building development plans, optimize for people who can interpret insights and build alignment internally. The pure analyst role is shrinking. The strategic interpreter role is expanding.
Both efficiency and savings count, but savings is the bigger prize.
Sievo's Hackett-validated benchmark: $20 million in savings per $1 billion in spend, about 2% of total spend, at a 63x return on investment. For most large enterprises, that math eclipses anything an AI efficiency play alone can deliver.
Mature procurement teams hit a ceiling with internal data alone. To unlock the next stage of savings, they need market-aware insights: peer benchmarks, payment-term comparisons, and renegotiation signals that only show up when your data sits alongside everyone else's. Sievo's Community Data is what makes that step change possible.
To illustrate how that math plays out, JP shared a hypothetical webinar scenario: a $10 billion company with $7 billion in addressable spend and a $20 million procurement team.
These are illustrative figures from a webinar scenario, not universal benchmarks. Actual ROI depends on your spend mix, current maturity, and how deeply AI is integrated into procurement workflows.
Efficiency matters. Speed matters. But the bigger case for procurement AI is the savings it unlocks on external spend. That's what moves procurement from cost center to value driver, and that's the conversation you want to be having with your CFO.
Across all seven questions, the panel kept landing in the same place: don't wait. Procurement teams that learn to work with imperfect data are pulling ahead of teams that hold out for clean data, and the gap is widening, not closing.
The first move is knowing where you stand today. Sievo's procurement maturity assessment shows you where your organization sits on the data and analytics maturity curve and returns a tailored action plan for what to build next. It takes a few minutes.