What buyers actually changed — and what they didn't
Over the past two years I've had a consistent conversation with OEM buyers in photonics and precision instrumentation. They've adopted AI-assisted procurement tools faster than most vendors expected. RFQ analysis, supplier benchmarking, compliance screening — these are being automated at the buying-committee level in ways that weren't possible three years ago.
The reaction from most sales teams has been anxiety. The assumption is that if procurement can automate analysis, the human sales relationship becomes less important. This gets it exactly backwards.
AI in procurement doesn't reduce the value of the sales relationship — it changes what the relationship needs to deliver.
What AI actually does inside an OEM buying committee
In practice, AI tools in enterprise procurement do three things well. They aggregate public data — pricing benchmarks, technical specifications, supplier financial health. They surface compliance flags faster than a human analyst. And they generate comparison matrices that make the early evaluation stage more efficient for the buyer.
What they don't do is navigate the internal politics of a 10-person buying committee. They don't understand why the Head of Manufacturing is resistant to a new supplier even when the technical specs are better. They don't know that procurement and R&D are operating from different assumptions about what the evaluation is actually for.
Those dynamics — which are the core of what makes OEM deals hard — are invisible to a procurement AI. They're barely visible to a vendor who isn't paying close attention.
The shift in what the sales conversation needs to do
If the buyer's AI can already pull your specifications, pricing and company financials before the first call, arriving to that call with a product pitch is a waste of both parties' time. The data layer has already been covered.
What the AI can't do is help the buying committee understand what the risks of the decision actually are — not the technical risks, which are in the datasheet, but the organisational risks. What happens if this supplier has a lead time problem during a product ramp? Who owns that problem internally? How does procurement get ahead of a quality issue without losing two years of qualification work?
The sales conversation that adds value in this environment is the one that helps the buyer think through those questions — before they become problems. That's a different conversation than feature-benefit selling, and it requires a different level of market and customer knowledge to have it credibly.
What this means practically
The first implication is that discovery has to go deeper. The standard qualification questions — budget, timeline, decision process — are necessary but no longer sufficient. You need to understand the internal dynamics: who loses if this decision goes wrong, who benefits if it goes right, and where the risk is being held inside the buying organisation.
The second implication is that the value of domain expertise increases. A vendor who understands the photonics supply chain well enough to have an informed opinion about risk management in that supply chain is not replaceable by a procurement AI. A vendor who shows up with a generic pitch and a polished deck is being commoditised by one.
The third implication is about timing. If the AI-assisted screening happens before the first formal contact, the window to influence evaluation criteria has narrowed. Getting into the account at the right time — before the formal RFQ, when the requirements are still being shaped — matters more than it did five years ago.
The longer view
AI will continue to change how large organisations evaluate and select suppliers. The direction of that change is not towards fewer human relationships — it's towards higher-quality human relationships at fewer, more consequential moments in the buying process.
The salespeople who will struggle are those who have been providing value primarily through information transfer — pitching features, explaining specs, sending comparison tables. That part of the job is being automated, and it's not coming back.
The salespeople who will do well are those who have been providing value through judgment — helping buyers navigate ambiguous decisions, building the internal case for a vendor selection, managing risk through a multi-year relationship. That part of the job is not automatable, and AI is making it more valuable, not less.