More RFQs, Fewer Real Opportunities: A Sell-Side Qualification Layer for Agent-Generated Demand

Key takeaways

• Forrester expects at least one in five B2B sellers to be forced into agent-led quote negotiations in 2026 — buyer agents will deliver dynamically generated counteroffers, and a seller's systems must answer them with tenable prices and terms.
• Harvard Business Review documents B2B firms receiving more RFQs without more real opportunities — AI makes polished, low-quality procurement email nearly free to send, so inbound volume no longer signals pipeline health.
• A three-bucket triage — quote, clarify, decline — changed the pipeline for a UK precision moulder — in HBR's Meridian Mouldings case, the qualification questions opened real commercial conversations the original RFQ never contained.
• Pactum's production deployments prove the pattern at scale: Honeywell runs autonomous negotiation across $500M+ in spend, and SUEZ cut requisition cycle times 90% with agents inside Coupa — agents execute within governance guardrails while buyers keep exceptions and strategy.
• Gartner projects 90% of B2B buying will be AI-agent-intermediated by 2028, pushing over $15 trillion through agent exchanges — the sell side has a two-year runway to become machine-quotable before buyer agents do the shortlisting by default.

Forrester expects at least one in five B2B sellers to be forced into agent-led quote negotiations in 2026 — compelled to respond to AI-powered buyer agents with dynamically delivered counteroffers via seller-controlled agents. That prediction was published in Forrester's 2026 B2B marketing, sales, and product predictions. The inbound half of the shift has already landed: Harvard Business Review reports that B2B customers now use AI to generate supplier lists, draft procurement emails, and prepare RFQs, firing polished enquiries at a dozen firms in the time it once took to draft one. The consequence, in HBR's words: firms receive more RFQs without receiving more real opportunities, and sales teams spend more time quoting and less time closing.

For a mid-market manufacturer or distributor — a 200-person precision components shop, a regional industrial distributor running NetSuite and BigCommerce, a specialty fabrics supplier whose quote support is two engineers and a sales coordinator — this is not an abstract trend. It is a capacity problem measured in engineering hours: every AI-generated request that receives the full quoting treatment pulls time from the work that wins real orders.

This article maps the sell-side qualification pattern that fixes the measurement problem first and the volume problem second: score inbound RFQs for completeness and buying intent, sort them into quote / clarify / decline, and reserve human quoting time for qualified demand. It builds on From Email Chains to Agent Delegation: B2B RFQ Automation with A2A and Hermes Agent, which covered the buy-side quoting cycle; this is the other end of the RFQ — your inbox.

The noise problem: inbound volume stopped being a signal

When Meridian Mouldings — the small UK precision moulder HBR studied under a disguised name — reviewed its inquiry pipeline, three signals stood out. Many RFQs had clearly been sent to a long list of suppliers at once; some addressed the wrong company by name, or named several suppliers in the same message. The language was generic, making each job look simpler than it was. And the documents omitted the context a responsible quote requires: what the part is for, how it will be used, what stress tolerances matter, whether earlier versions failed, whether the buyer wants design advice or is collecting prices. The details are documented in the HBR reporting and in MarketScale's coverage of the same case.

None of this looks like noise from the outside. AI-drafted RFQs arrive professionally formatted, often with drawings and comparison tables attached. That is exactly what makes them expensive: the only reliable way to discover a request is hollow is to spend the engineering time a real request deserves. AI did not produce better demand. It made low-quality demand cheap to generate.

The same mechanism is reshaping the buy side's first screen. MarketScale's analysis of AI-mediated procurement puts it plainly: before a procurement team books a vendor call this year, an AI agent may have already culled the shortlist. Buyer-side agents optimize for coverage over fit, and vendors with well-structured, machine-readable data surface more reliably than equally capable competitors whose credentials sit in PDFs and gated portals. The sell side faces both problems at once — a noisier inbox today, and buyer agents that shortlist on machine-readable evidence tomorrow.

Manual triage versus agent-assisted qualification — what changes when a quoting team scores inbound RFQs before quoting them:

The qualification pattern: score before you quote

Meridian's response started with measurement, not software. The team sorted recent inquiries into three buckets — ready to quote, needs clarification, and quote fishing — and ran a short qualification screen before anyone opened a drawing. The screen asked whether the inquiry explained the application, specified operating conditions and material requirements, included a drawing or sample, and showed evidence of a genuine buying process rather than a broad market scan.

An agent layer operationalizes that screen at the volume AI-generated demand arrives. The division of labor matters more than the technology:

• Completeness scoring. Does the request specify application, operating conditions, tolerances, volumes, and timing — the fields Meridian found missing? Missing fields become structured clarification questions, not silent guesses that surface as change orders later.
• Intent evidence. Is this a production-ready buyer or a price scan? Signals include a named buying process, drawings, supplier-specific language — versus a generic template that, as HBR noted, sometimes addresses several suppliers in one message.
• Three-bucket triage. Requests that pass go straight to quoting. Promising but incomplete ones get a short technical clarification note; in the Meridian case, those questions often opened a real commercial conversation the original RFQ never contained. Requests that go quiet after clarification are a result, not a loss — buyers unwilling to answer basic application questions rarely valued the firm's specialist expertise in the first place.
• Human review stays on qualified quotes. The agent does the reading and sorting; the engineers price the work. That is the same division procurement automation is converging on at enterprise scale: Pactum describes its deployment model as agents executing routine work within governance and commercial guardrails while buyers keep strategy, exceptions, and relationships.

The honest caveat: no scoring layer replaces engineering judgment about what a request is worth. What it replaces is the reflexive default — quoting every polished inbound request on equal terms — which is the race HBR says a specialist least wants to win.

The outcome

Enterprise deployments put production numbers behind the pattern. Pactum's published customer results: Honeywell runs autonomous negotiation across more than $500 million in spend, engaging its entire supplier base on payment terms, rebates, and discounts; Ahold Delhaize USA moved supplier negotiation into autonomous execution within defined governance guardrails and saw ROI within weeks; SUEZ embedded agents in Coupa to review every purchase requisition and cut requisition cycle times by 90%. In each case the human role narrowed to exceptions — at SUEZ, buyers step in only when a requisition needs human review. The details are in Pactum's customer summary.

At mid-market scale, the same pattern buys back the scarcest resource: qualified quoting hours. The measurable shifts are conversion-oriented rather than volume-oriented — quote hours per qualified opportunity, clarification-response rate, and the share of engineering time spent on requests that can actually become orders. Meridian's monthly sales review stopped leading with raw RFQ volume and started leading with those questions. When AI makes inbound volume cheap to inflate, inquiry count is the wrong KPI; qualified-opportunity count is the one that survives.

There is a second half to the runway. Gartner projects that by 2028, 90% of B2B buying will be AI-agent-intermediated, pushing over $15 trillion of B2B spend through agent exchanges — and Gartner's accompanying analysis notes that verifiable operational data becomes a prerequisite for participation in that ecosystem. Qualification handles today's noise. Machine-quotable catalogs, structured pricing, and agent-addressable systems handle tomorrow's buyer agents. Suppliers who invest in both are optimizing for the intermediary that now sits between buyer and seller.

Representative build vignette

A mid-market industrial distributor's quoting team was spending its mornings on AI-generated RFQs that would never convert: generic line items, no application context, salutations naming three other suppliers. An intake agent now scores every inbound request against the qualification screen — completeness fields, intent evidence, duplicate-blast detection — and routes it: qualified requests flow into the existing quote workflow with an availability check against NetSuite; incomplete ones get a clarification note naming the specific missing fields; template blasts get a polite decline. The team reviews qualified quotes only. Quote hours per real opportunity fell by more than half in the first quarter, and the monthly review now tracks qualified-opportunity conversion instead of raw RFQ count.

Request a scoped build. One-week discovery. You get a system inventory, workflow map, and fixed scope — whether or not you build with us.

Related reading

• From Email Chains to Agent Delegation: B2B RFQ Automation with A2A and Hermes Agent — the parent article: the buy-side quoting cycle, from 3-day manual workflows to agent delegation with MCP tools and A2A
• Commerce Protocols for AI Agents: UCP, ACP, AP2, and MCP — How the Stack Fits Together — the machine-quotable half of the runway: making your catalog and pricing legible to the buyer agents that shortlist suppliers
• When the Agent Places the Order: How Agentic Payments Close the B2B Procurement Loop — what happens after a quote is accepted: agent-initiated ordering, payment, and settlement audit
• AI Agents as Advertising Surfaces: What Sponsored Placement Means for B2B Visibility — the visibility layer above the funnel: OpenAI's Sponsored Agents, citation volatility, and the two-layer strategy for being found when agents do the buying