Reindeer founder and CEO Yoav Naveh told FreightWaves at the Supply Chain AI Symposium in Chicago that the market for enterprise AI agents is no longer being won on model quality. Instead, he argued, the competitive battleground has shifted to how vendors handle the fragmented, ever-changing workflows that make up core business operations.
Three waves of enterprise AI adoption
Naveh described three adoption waves, each with a different set of expectations and results:
- Wave one — assistants: "Everybody got a copilot, Claude, ChatGPT," Naveh said. He noted strong success with assistants in software engineering—"coding agents work really well"—but weaker results in legal. "Generally," he said, "we haven't seen assistants make companies 10x or 100x more efficient."
- Wave two — call centers: Customer support was "a very big low-hanging fruit, really," Naveh said, because it is one big workflow with similar case types. "I definitely see some success there."
- Wave three — core operations: Enterprises are now chasing workflows that "represent 70, 80% of the work that everybody does," spanning finance, procurement, treasury, accounting and supply chain. Naveh estimated these workflows number in the dozens, if not hundreds, per enterprise, often with no single department owner.
| Wave | Focus | Naveh's assessment |
|---|---|---|
| Assistants | Copilots, Claude, ChatGPT | Mixed; strong in software engineering, weak in legal; no 10x/100x efficiency gains |
| Call center | Customer support | "Very big low-hanging fruit" with "some success" |
| Core operations | Finance, procurement, treasury, accounting, supply chain | 70–80% of work; fragmented and resistant to one-size-fits-all software |
The fragmentation problem
Naveh's clearest example of why core operations resist standardisation comes from a client he described as one of the largest consumer packaged goods companies in the world. "We work with one of the largest CPG companies in the world. Five different departments, five different accounts payable processes within the same company," he said. "They say, 'Hey, we're doing accounts payable in my company,' but your accounts payable is different from their accounts payable."
That fragmentation is why Naveh is skeptical of the forward-deployed engineer model, where vendors embed technical staff inside a client to hand-build an agent for one process at a time. "I'm going to throw three engineers in and sit with your team and build the agent for that particular process. Are they ever going to come back and be available to be sent to another department, or are they stuck maintaining whatever they built as processes change and evolve?" He concluded: "I think the math doesn't work because it doesn't scale."
I think the math doesn't work because it doesn't scale.
The model becomes a commodity
Reindeer's bet runs against the assumption that owning or fine-tuning a proprietary large language model is a durable edge. "Strong belief that LLMs are going to be a commodity," Naveh said. "We see that open-source models are maybe six months behind the frontier models. But what we're thinking is, I don't need the frontier model for everything anymore."
He cited a large European insurance company that decided to build its own LLM. "We're happy to plug in their own LLM," he said. "We're happy to restrict the list of allowed LLMs based on the company's data privacy policy, and figure out how to run their case with what's whitelisted."
Maintenance is the real moat
That puts pressure on competitors whose pitch rests on model superiority alone. In Reindeer's model, differentiation comes from the maintenance layer: detecting when a workflow drifts, reconciling the change and adjusting the agent without opening a new engineering ticket. "Even after you deploy an agent, the workflow keeps changing," Naveh said. "If you don't have a method to have agents detect these changes, reconcile, self-heal, self-learn, we're not really going to be able to rea[ch the scale enterprises need]." The truncated quote underscores an implicit point: without that maintenance layer, agents degrade as business processes evolve, and the forward-deployed engineer approach leaves vendors stuck maintaining hand-built integrations. For enterprise technology decision-makers, Naveh's argument suggests that procurement should weigh an AI vendor's ability to keep agents aligned with changing internal processes at least as heavily as model benchmarks, since the model layer is increasingly interchangeable and the workflow layer is where the cost and complexity actually live.
Sources:
