Every large company has a junk drawer. Inside are replacement parts, software renewals, repair jobs, catering orders and one-off services - purchases that matter to somebody, but rarely enough to earn a category manager's afternoon. Procurement people call this “tail spend.” The name makes it sound trivial. The transaction count says otherwise.
Fairmarkit built its business in that drawer. The Boston company connects to an enterprise's purchasing system, spots requests that fit its rules, recommends suppliers, runs a competitive RFQ or RFP, collects bids and helps evaluate the result. The promise is not that a machine suddenly becomes a master negotiator. It is that thousands of events which once died in email, spreadsheets or simple renewals can move through a repeatable process.
That distinction matters. Plenty of procurement software records decisions. Fairmarkit wants to execute the work around them. Its latest product language is “Total Agentic Sourcing”: KIT, an AI copilot, coordinates agents for intake, supplier discovery, RFx execution and evaluation. A person still supplies policy, context and approval. The agents handle the scavenger hunt.
The first idea was wrong in a useful way
Kevin Frechette's original irritation came from enterprise technology sales. He saw companies paying different prices for the same equipment and wondered why business purchasing lacked the price visibility that consumers had on Zillow or Glassdoor. He later teamed up with former colleague Tarek Alaruri and technologist Victor Kushch. Fairmarkit was founded in 2017.
Price transparency was the opening theory. Customer conversations exposed the larger problem: buyers were not merely missing a benchmark. They were drowning in the process required to create competition at all. The founders moved toward automating requests for quotation. It was a less glamorous observation with better economics - if no buyer has time to invite five suppliers, the perfect reference price is academic.
This is what failed first, and what changed their minds. The team did not abandon pricing; competitive bids still produce it. They changed the unit of value from an answer to a workflow. By 2019, Fairmarkit said dozens of large enterprises were using its machine-learning system and reporting more than 15 percent savings on the transactions then being optimized. Insight Partners led an $11 million Series A. A $30 million Series B followed in 2020, then a $35.6 million Series C led by OMERS Growth Equity in 2022.
Five agents, one unlovely job
Picture a plant that needs an unfamiliar component. The request arrives with a thin description. An intake agent can clean up requirements and route the event. Discovery searches approved vendors, the customer's history and Fairmarkit's supplier marketplace. An execution agent drafts the event, invites bidders and chases responses. Evaluation reads documents and compares price, delivery, experience and compliance. The award returns to the existing system.
The useful loop is not a chatbot window. It is structured work moving through the procurement stack, with policy and approvals attached.
For a tactical $500 request, much of that sequence can run automatically. For a complex contract, the same workspace assists a category manager rather than replacing one. Fairmarkit says its Bid Analysis Agent can structure and interrogate unstructured submissions in under 30 seconds. Its SOW Builder drafts statements of work. Performance and compliance tools retain the audit trail.
The less photogenic product is integration. Fairmarkit connects with systems such as Coupa, SAP Ariba and ServiceNow instead of asking customers to throw away the procurement stack they already fought to install. That makes Fairmarkit both a competitor and a neighbor to the big suites. Against specialist vendors such as Globality, Keelvar, Arkestro, ORO Labs and Zip, it argues for breadth of execution and the reach of its supplier data. The cheapest alternative remains email and Excel, which is free in the same way a traffic jam is free.
“We get our ROI every year within about six to eight weeks just on our savings.”Jim Haining, Purchasing Manager at Las Vegas Valley Water District
Customers buy savings. Then capacity wins the renewal
Fairmarkit does not publish list pricing. Contracts are negotiated enterprise subscriptions, typically paired with implementation and integration; suppliers use the bidding portal free. So the public evidence on cost is indirect. Las Vegas Valley Water District says savings repay its annual subscription in six to eight weeks. Atlantic Methanol reported implementation in one month and ROI in roughly six weeks. Arauco reported a four-times annualized return based on awarded savings.
The more interesting customer stories begin when savings stop being the only score. An unnamed Fortune 200 manufacturer started with a one-year tail-spend pilot. In eight months, a limited scope produced nearly $1 million in awarded savings. With the initial target looking safe, leadership changed the brief: use the platform to scale operations. Fairmarkit says the expanded program went on to generate about $15 million in awarded savings on roughly $285 million of baseline spend, while more than 50,000 events ran with no human intervention.
Snowflake offers a cleaner before-and-after. Its sourcing cycle fell from 25 days to 6.4. Nearly a third of events ran autonomously, and awarded savings improved by another five percent. BP says it has sourced $1.2 billion through Fairmarkit and cut cycle time by half. Boeing reports eliminating 115,000 hours of sourcing cycle time annually. These are company case studies, not controlled experiments, but they reveal what enterprise buyers are rewarding: more throughput per buyer and less waiting for internal stakeholders.
The rollout is more copyable than the algorithm
The practical lesson is not “buy an agent.” It is to choose a neglected, measurable workflow where automation can earn trust. Fairmarkit's strongest examples begin with a bounded category or geography, connect to the system employees already use, and define savings, cycle time, participation and compliance before launch.
Pick high-volume, repeatable spend with clear baselines. Tail-spend MRO worked because the old process was visible and disliked.
Route requests from the ERP or P2P system people already know. A second inbox is an adoption tax.
Start with review gates. One manufacturer tested auto-send in phases before widening the rollout.
Track creation-to-open, open-to-close and close-to-award, not just negotiated price. Capacity is often the durable gain.
Sonoco connected Fairmarkit to Coupa and went live in two weeks in Latin America. The manufacturer paired its SAP ECC integration with milestones, tailored training, supplier communication and executive sponsorship. Boston University went further, acting as an early collaborator on self-service sourcing, where employees outside procurement could run events. The pattern is unusually consistent for enterprise software: integration, rules, training, then autonomy.
This approach depends on enough addressable spend, comparable bids, reachable suppliers, usable purchasing data and leaders willing to enforce a workflow. It is a poor fit for a tiny buying operation, a sole-source category, chaotic requirements or an organization that cannot connect its systems. AI can accelerate a clean process; it can also accelerate a confused one.
A marketplace hiding inside the machine
Fairmarkit's longer wager is not just that agents can create an RFQ faster. It is that a marketplace grows more useful as buyers run more events and suppliers build histories. Better matching draws more bids; more bids improve price and choice; structured outcomes improve the next recommendation. That loop is why the supplier portal is free and why discovery is central to the product.
It also creates the hard questions. Supplier ranking must be explainable. Customer policies must survive automation. A cheap bid that fails on safety, delivery or diversity is not a win. Fairmarkit's pitch is “people-first” autonomy: agents execute, humans set strategy and decide. The credibility of that phrase will depend less on a demo than on exceptions - what the system refuses to automate, how it shows its reasoning and how quickly a buyer can take the wheel.
Nine years after the founding idea, Fairmarkit has moved from a tail-spend RFQ engine into strategic RFPs. The company said RFPs represented half of all spend run through its platform in 2025, while autonomous-sourcing adoption grew 50 percent year over year. That is the expansion in one statistic: deeper work, with more of the execution delegated.
The company still lives by its first useful correction. Buyers did not need one more dashboard telling them that purchasing was inefficient. They needed fewer tabs open, more suppliers responding and a result returned before the requester found a workaround. Fairmarkit began with the purchases no one wanted. It may have found its market by respecting exactly why no one wanted them.