Latest / The Buyer's Side with Fexingo: Procurement, Sourcing, and B2B Purchasing Decisions / How a Buyer Used AI Sourcing Tools to Cut RFQ Time 70 Percent
Transcript
- Lucas: So there's this question I keep hearing from procurement folks, especially in mid-size manufacturing: can AI actually save me time, or is it just another dashboard I have to maintain? Luna: And I bet the answer depends a lot on where you start. Lucas: Exactly. So today I want to walk through a specific case. A mid-size industrial component maker, about $400 million in revenue, 14 categories they sourced. Their RFQ cycle — from initial spec to three qualified bids — was running six weeks on average. Luna: Six weeks? That's brutal. Especially if you're in a volatile raw materials market. Lucas: And they were. Steel and specialty alloys. So their procurement director, a woman named Priya Patel, decided to pilot an AI sourcing tool. Not one of the giant suite players — a smaller platform called BidMind, specifically built for industrial buyers. Luna: BidMind — I've heard of them. They focus on unstructured data, right? Like pulling specs from PDFs and emails? Lucas: That's the core of it. The biggest bottleneck in their RFQ process wasn't finding suppliers. It was translating internal engineering requirements into a format suppliers could actually respond to. Engineers would send over a PDF with hand-drawn annotations, or a marked-up CAD file, and the procurement team would spend days cleaning that up into a standard RFQ document. Luna: So the AI basically ingests that raw engineering output and structures it automatically? Lucas: Yes. BidMind uses a large language model trained on manufacturing specs. Priya's team uploaded 47 past RFQs and their corresponding winning bids. The model learned how their engineers talked about tolerances, materials, surface finishes. After that, they could drop a new engineering PDF into the system and get a draft RFQ in under 30 minutes. Luna: Thirty minutes versus what — a couple days manually? Lucas: Typically two to three days for a senior buyer to rewrite and verify. So that alone was a huge gain. But the real step change came in supplier matching. The tool also scraped supplier websites and public quality certifications to create a shortlist ranked by capability fit. Luna: So instead of the buyer going to their rolodex and picking the same three suppliers they always use, the AI was suggesting suppliers they might not have considered. Lucas: Right. And Priya made a deliberate choice there. She configured the tool to weight supplier diversity as a factor. So if a minority-owned or woman-owned shop had equivalent capability, it would rank higher. That was important to her company's ESG targets. Luna: Did that add time or reduce it? Lucas: Initially it added a filter step, but because the AI could screen 200 suppliers in seconds, the overall process still shrank. After three months, their average RFQ cycle dropped from six weeks to nine days. That's a 70 percent reduction. Luna: Nine days. So what did the suppliers think? I imagine some of them were skeptical about bidding against ai generated specs. Lucas: There was pushback. A few long-time suppliers complained that the spec sheets were too rigid — they'd lost the ability to call up the buyer and ask, 'Hey, can we relax this tolerance a bit?' The AI didn't capture those informal conversations. Luna: So Priya had to build in a human review step. One day per RFQ for the buyer to review the AI's output and add notes about flexibility. Lucas: Exactly. She called it the 'common sense pass.' And that actually improved the final cycle time because it reduced rework. Suppliers submitted fewer non-conforming bids. Luna: What about cost savings? Faster sourcing is great, but did they actually pay less? Lucas: They measured that too. In the first year, across the 14 categories, they achieved an average 6.8 percent cost reduction on sourced materials. That's partly from competition — they were getting more bids per RFQ, 5.2 on average versus 2.8 before — and partly from the AI identifying substitute materials that met spec at lower cost. Luna: Six point eight percent on a $240 million annual spend is over $16 million. That's not pocket change. Lucas: Not at all. And the tool cost them about $180,000 for the first year, including training and integration. So the ROI was pretty dramatic. But I want to be fair here — not every category performed equally. In highly engineered components with complex certifications, the time savings were smaller. The AI struggled with regulatory nuance. Luna: So the lesson might be: start with your simplest categories first. Prove the concept on something like standard fasteners or commodity chemicals, then move to the hard stuff. Lucas: That's exactly what Priya recommends. She told me their first pilot was on hydraulic fittings — a relatively standard category with clear specs. Once the team saw it work there, they had the credibility to roll it out to electrical components and then to custom machined parts. Luna: And how did the procurement team itself react? I imagine some buyers worried the AI would replace them. Lucas: That was a real concern. Priya addressed it head-on in a town hall. She said, 'This tool is going to take over the data entry and the document formatting. That frees you up to do the strategic work — supplier relationships, negotiations, risk management.' And she backed it up by reassigning two buyers who had been doing nothing but RFQ paperwork to a new supplier development role. Luna: So they actually saw it as a promotion path. That's smart change management. Lucas: It was. And within six months, the buyers who had been most skeptical were the ones advocating for the tool. They realized they'd rather spend their time negotiating than formatting spreadsheets. Luna: I think that's a really important point for any procurement leader listening. The technology itself is powerful, but the cultural adoption is what makes or breaks it. Lucas: Absolutely. And that's actually a good place to mention something. If these conversations have moved your work forward in some small way, if you've ever used a tactic from this show to improve a negotiation or cut a cycle time, we'd love your support. This podcast stays ad-free because of listeners like you. If today was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. Luna: Yeah, it really does help us keep going. And we genuinely appreciate every single contribution. Lucas: So back to Priya's case. One more thing I found striking: they also tracked supplier satisfaction. At the end of year one, they surveyed their top 50 suppliers and asked, 'Has the new RFQ process improved your experience?' 68 percent said yes, mostly because the specs were clearer and they got fewer follow-up questions. Luna: That's a win-win. Faster for the buyer, clearer for the supplier. The AI basically acted as a translation layer between engineering and procurement and the supply base. Lucas: Exactly. And I think that's the real takeaway. AI in procurement isn't about replacing judgment. It's about removing friction so judgment can happen faster. Luna: So for a buyer listening right now who's curious but not sure where to start, what's the one thing you'd recommend? Lucas: Pick one category. One product family. Ideally something with a history of past RFQs you can use for training. Run a three-month pilot. Measure cycle time and cost before and after. Don't try to boil the ocean. Priya started with hydraulic fittings. You can start with whatever your version of hydraulic fittings is. Luna: And be ready for the cultural conversation. Talk to your team early. Lucas: Right. And if you want to see the full case study, we'll link to Priya's presentation from the recent ISM conference in the show notes. It's got the exact numbers on defect rates, cost savings, and supplier satisfaction scores. Luna: I'll be reading that. Thanks, Lucas. Lucas: Thanks, Luna. That's all for today.