At SNAK we use an AI notetaker. It is sync’d with our calendars and automatically joins meetings. Sometimes, the founder or investor we are meeting with also has an AI notetaker. These AI notetakers are impeccably punctual, so sometimes I’ll join a call and see both notetakers hanging out before any human participants have joined. My partner Adam & I joke that in the future these notetakers will do the meeting without the humans and report back.
AI agents, also known as agentic AI, are autonomous software entities that execute tasks with minimal human oversight. We’ve previously written about our belief that AI has potential to disrupt the procurement function and enable founders to build demand-led marketplaces using agents. Taking that one step further, if every vendor on the supply side similarly has agents representing their interests, could these agents be trusted to fully, or mostly, automate transactions?
We see the biggest opportunity in B2B transactions that involve fragmented supplier options, complicated contract negotiations, and complex delivery logistics. Here AI agents can autonomously source and compare suppliers, covering a broader base of options and getting hundreds of bids vs. a handful, which should better optimize across price / delivery speed / sustainability goals. The next step would be to have an AI agent negotiate contract terms and legal documents. This is not as far out as one might think– AI negotiation startup Pactum is being used by Walmart, and this HBR article details the learnings for crossing the pilot to production threshold– e.g., starting with indirect spend categories and pre-approved vendors to lower the risk of adoption. Lastly, because B2B purchases tend to be larger (in $s and physical volume), the fulfillment logistics are nontrivial and involve identifying options, negotiating terms, monitoring delivery and handling dispute resolution. These agents could also more efficiently monitor global supplier performance, simulate multiple sourcing scenarios in real time, and predict market disruptions before they happen by looking at macro data across a marketplace. As with most AI opportunities, the quality of the data in the system will directly correlate with the value a startup can create — and thankfully supply chain technology has come a long way over the past decade to include robust ERP platforms and digital twin platforms to access real-time data and historical data so startups need not start from scratch.
Here are 7 areas where AI can be used across the procurement cycle:
We are likely a ways away from fully autonomous marketplaces as there is a high trust hurdle and the importance and dollars associated with these transactions are too high to risk making a mistake. In the near term, companies working on fully autonomous transactions will need to consider AI governance. There are three kinds of human-machine collaboration models when it comes to AI:
“Human in the loop” - whatever the machine does there’s always a human to see the result and approve, disapprove, and so forth. This is a good approach if those decisions are supercritical or if there’s a rather small number of high value decisions to be made.
“Human on the loop” - the machine does work on its own, but there’s a human supervising the process, which works well with routine tasks.
“Human out-of-loop” - machines run without humans to intervene, e.g. high-frequency trading. In this case, speed is of the essence, which requires the machine to make decisions autonomously because humans are too slow to intervene.
While a number of the players in this space are horizontal (e.g., Pactum, Sirion CLM, Ironclad) and are focused on a single point in the transaction value chain (e.g., supplier negotiation) we believe there’s an interesting opportunity to start focusing on a narrow vertical, build across the value chain above, and potentially provide AI seller solutions as well to create an end-to-end marketplace. If you’re taking this approach, or even starting at one point in the value chain but have the larger ambition to vertically integrate, we’d love to connect – as always, you can reach us at contact@snak.vc.
Weekly Deals
What We’re Reading
How to Scale Autonomous Negotiations - video link
How Walmart Automated Supplier Negotiations - HBR - link
Meet with us
We’re on the move, and love to meet with founders and investors in person. Reach out to contact@snak.vc:
April 29 - May 1 | Palm Springs - investor meetings
May 22-23 | NYC - we will be at VSC’s Dirty Jobs Summit and would love to meet with founders and investors before and after
June 25 - 27 | Indiana - Great Lakes Venture Summit
Loved this piece, we have been building Globixo (AI Powered Global B2B Sourcing - think Perplexity with Agentic Capabilities) around the same thesis.
Link to our platform: https://www.globixo.com/
We have started with food & beverage category, listed suppliers from leading global trade shows, launched our web chat agent in a few closed communities and can already see some interesting use cases emerge and firsthand validation of a lot of things you mentioned here.