RFTs in the AI Era: 5 Things Every Procurement Team Should Know
AI has changed what it costs to respond to a tender. What used to take a supplier 40 hours now takes 8 to 15. The barrier to bidding has dropped, submission volumes are rising, and procurement teams are dealing with more responses, faster, from a wider field.
That's not a bad thing. But it does raise the stakes for how tenders are designed, how responses are evaluated, and how organisations on both sides protect themselves from bad AI habits.
Earlier this year, Unimarket's Jarrod Stevens and Lewis Young sat down with Laurie Nicol, founder and CEO of Tendl, for a webinar exploring exactly this. Here are the five things that stood out.
AI is raising participation, not just speed
The headline efficiency gain is real: suppliers using AI-powered tools are seeing win rates improve by around 22%, and they're able to bid on more opportunities without cutting corners. But the bigger shift is structural. High-quality suppliers who previously had to pass on opportunities due to bandwidth are now in the running. That's better for market competition and better for buyers who want genuine choice.
The catch is that lower effort also lowers the bar for low-effort submissions. More bids doesn't automatically mean better bids.
Buyers need to design for quality, not just compliance
When volume goes up, evaluation workload follows. The solution isn't to add more filters after the fact. It's to build better tenders from the start.
That means publishing clear qualification thresholds upfront, stating non-negotiables explicitly, and weighting criteria in a way that helps unfit vendors self-select out before they submit. Locked PDFs, stitched-together question sets, and duplicated requirements all create unnecessary friction. They slow suppliers down and make evaluation harder. Clean, consistently structured tenders produce sharper responses and faster scoring.
As Laurie put it during the session: "What's good for AI is also good for humans."
A well-designed tender works better for everyone, and procurement teams who invite opportunities for suppliers to humanise their responses with videos, presentations and stories will have an easier time finding quality submissions.
Winning responses pair AI with genuine expertise
If every supplier can produce a clear, well-organised response, then polish is no longer a differentiator. What separates winning bids now is substance: specific understanding of the buyer's problem, evidence of past delivery, and the kind of informed nuance that only comes from someone who actually knows the space.
Generic AI output is easy to spot. Evaluators are looking for credibility signals: named resources, relevant case studies with real outcomes, risk registers with actual mitigations, and commercial assumptions that reflect a realistic read of the project. The suppliers doing well are the ones using AI to handle structure and consistency, while keeping humans in charge of the narrative.
Data security can't be an afterthought
Shadow AI is the real risk. Staff pasting tender content into free chatbots creates data exposure that enterprise platforms simply don't. Blanket bans don't work, either. Teams find workarounds.
The more effective approach is to provide sanctioned tools with clear guardrails and train people on what is and isn't shareable.
For any AI tool handling procurement data, the basics matter: no model training on customer data, clear data residency, and proper access controls. If the secure option isn't easy to use, people will use the insecure one.
AI supports evaluation. It doesn't replace it
Automated tools can check hard criteria, flag compliance gaps, extract data, and surface inconsistencies across vendor responses. That's genuinely useful. What they can't do is weigh trade-offs, validate real-world feasibility, or judge whether a proposed solution actually addresses the underlying business need.
The framing that came up in the webinar is a good one: treat AI like a fast, tireless intern. It's excellent with direction and focused tasks. The judgment calls still need a person.
Want to go deeper on any of these? The full session with Jarrod, Lewis, and Laurie covers tender design, AI-assisted evaluation, and practical approaches to secure adoption.