No company will immediately trust an AI with a $100K negotiation. The path to autonomy starts with a human-in-the-loop approach. This builds user trust gradually while simultaneously feeding the AI agent critical feedback and learnings on the company's specific operational nuances.
It's easy for an internal team to build an AI prototype with 70-80% performance. However, this doesn't deliver 70% automation; it means a human must still review 100% of the work, creating more overhead. Startups win by solving the complex "last 20%" of exceptions needed for true autonomy.
A single AI agent cannot solve a complex enterprise task like procurement. True automation requires a multi-agent system where specialized agents (e.g., for contracts, inventory, negotiation) coordinate and share information, mirroring how human departments collaborate to get a job done.
In industrial sectors like automotive, large procurement departments dictate the technology their suppliers must use. An AI platform that first serves the buyer can use this leverage to push its agents onto the supplier side, creating a powerful GTM wedge to own both sides of the transaction.
Incumbents bolting AI onto their software are confined to their data silos. AI-native startups gain an advantage by building agents that operate across the entire job-to-be-done, connecting disparate systems (legal, finance), documents, and conversations to resolve issues completely.
Systems like ERPs only capture the final outcome (e.g., a price of $8K), missing the vast preceding context of emails, meetings, and spreadsheets. AI agents create value by automating this "dark matter" of enterprise work that happens outside formal systems.
Enterprises often don't negotiate invoices below a certain threshold (e.g., $50K) because they lack the human capacity. AI agents can autonomously handle these negotiations at scale, capturing savings that were previously left on the table. The risk is low because the alternative was zero negotiation.
An AI agent that resolves an issue end-to-end is in direct conflict with a product providing a workflow for a human agent. These products often belong to different VPs with competing P&Ls, creating organizational friction that stalls true AI innovation.
While price is a zero-sum outcome, it follows thousands of preceding tasks like information sharing and compliance checks. In these areas, both buyer and seller agents share the same incentive for speed and reduced friction. This shared ground is key to productive agent-to-agent transactions.