Feature Focus: Auto Negotiate

For the past year we’ve been hard at work laying the foundation for LLMs in our primary product, the Scientist.com Marketplace. The result of that work is a platform that we call Benchmate. You’ll probably hear a lot more about Benchmate in the coming months, but for now we’ll just say that it allows us to build different LLM personas that can either utilize locally running Language Models or reach out to commercially available models such as OpenAI and Anthropic, making it a multipurpose platform.

Once Benchmate was up and running to our internal operational standards, it was time to incorporate it into the marketplace, where over $300M worth of transactions between buyers and sellers of custom research services and products occur annually. Obviously, the platform has many benefits, meeting legal and compliance standards, saving time and promoting new collaborations among them, but let’s be honest, most enterprise software solutions stay in business by saving their customers money. So, we began by testing Auto Negotiate, one of our first customer facing features to leverage LLMs.

Simply put, Auto Negotiate is a feature that helps reduce the cost of sourcing for our buyside customers. We all know that we should negotiate incoming proposals, but unless you’re a procurement professional, and it’s part of your job, you probably don’t love negotiating, which is why we built Auto Negotiate.

The Auto Negotiate feature resides in the product right at the point of purchase, which is where a researcher can accept a proposal. We provide users with a simple feature where upon choosing the proposal they are interested in, they can pick an aspect of the proposal they would like to negotiate, such as price or turnaround time, and Benchmate will generate a response.

The automated response is accomplished by packaging up the proposal they are interested in along with the other proposals they have received and sending that to a LLM along with a prompt describing which aspect they would like to focus the negotiation on. The results, which have been shockingly good, are presented to the user who can then proceed with the message as is or modify the results themselves. Even if the user does modify the prompt, Auto Negotiate solves “the blank page” problem, where a user might not even know where to start the negotiation process. The results have been more proposals negotiated and meaningful savings for our customers.

Auto Negotiate is just the beginning of what we’re planning to do with Benchmate — we can’t wait to roll out new features.