“The AI we use today will be the dumbest we’ll ever have worked with.”
For Sören Petsch, Head of Procurement at Root, testing the limits in procurement via new digital tools should be encouraged.
“We’ve made measurable progress on the AI agents that we’re using in our software orchestration tool, and we’re able to prove a fundamental business case that now allows us to think about how we leverage AI agents very differently,” reveals Petsch. “Normal software that is not AI software is deterministic, meaning you always have the same outcome if you enter certain types of information, whereas AI is probabilistic. The difference is that you have different outcomes even if you share similar data, which is interesting, but it also creates a different set of challenges such as inconsistencies and hallucinations.
“The foundational element was that we were able to figure out where we got a lot of creative inconsistencies and error messages. We focused on reducing the error rate, and it is now approaching zero. That is a pretty significant shift.”
Unlocking AI potential
Root is disrupting the fintech industry today. The company is an auto insurance carrier driven to create powerful insurance products and technology platforms to redefine the industry.
Petsch is a procurement practitioner passionate about AI. He has hands-on automation and AI expertise, specialising in transforming indirect procurement processes for internal stakeholders across legal, finance, InfoSec, IT, and procurement teams. Petsch focuses on workflow technology and AI, and helps organisations streamline operations, enhance visibility, and drive efficiency across departments. With experience across software, retail, and banking industries, Petsch is keen to bring practical, results-oriented solutions to procurement operations. His team’s mission is to “unbreak” procurement to power Root’s mission with data, technology, and repeatable processes – fast.
Inside Root
Since joining Root in January 2025, Petsch has sourced and implemented procurement orchestration software, automated the intake-to-pay process, and deployed AI agents to further automate it. During this time, Petsch discovered Vendr’s AI agent, Ruth, and considers it the best AI tooling for procurement of 2025. Petsch and his team use it to obtain real-time pricing benchmarks and negotiation levers before entering negotiations, providing key context before any conversation. When asked about AI’s importance to modern-day procurement, Petsch explains that there is always a “trade-off” between a shiny new innovation and a real use case, but that this AI tool hits the bullseye.
“The way I think about AI or automation is that tech should enable me to reduce administrative tasks since most procurement people likely spend a minimum of 95% of their time on tasks such as re-communicating the same information, answering emails, or transferring data from different systems into one place, etc. By adding an orchestration system, I now have all the information and data in one place. Now data and decisions are entered only once from beginning to end and leveraged throughout the decision and approval process. This data orchestration reduces the number of emails, Slack messages, and repetitive meetings by 75%, by my estimate. With everything in one place, I eliminate the need to search for relevant information and reduce mental switching costs. So, orchestration is the foundation to make AI agents actually deliver what the technology promises.
“If I can now use my time more efficiently by having AI prepare thoroughly and quickly for relevant conversations, it enhances my effectiveness as a strategic procurement professional. For example, we recently had a software renewal, and we just avoided the auto-renewal triggering by two days because it came in very, very late. At this stage, we had only 30 days to renew. Over the course of a week, I did my research, had a couple of internal alignment calls with stakeholders to validate my findings, and another one with the vendor to build rapport. Ultimately, we reduced the cost of that software by 75%. This was because we were able to understand where the leverage points were and spend my time aligning with the internal stakeholders, instead of doing research online.”