Jamie Cuffe and the price of paperwork

Jamie Cuffe grew up moving between London, New York and Bermuda. The geography came with an education in insurance: his father worked in the business, including operations for a Lloyd’s insurer and a reinsurer. Years later, Cuffe would return to that world carrying an idea developed elsewhere. The price of insurance, he believed, had something to do with the price of getting through the paperwork.

There is a wonderfully unromantic quality to this proposition. A founder can talk about artificial intelligence changing civilization. Cuffe can also talk about collecting a missing document, entering information into an old application, or getting a request out of an inbox. Insurance contains plenty of the second sort of problem. His company, Pace, is built to finish those jobs.

The larger ambition follows from the smaller chore. If handling an account costs too much, some business becomes impractical to serve. Lower that cost, Cuffe argues, and insurers can consider customers and products that previously failed the arithmetic. He is trying to change what happens before someone decides an opportunity is too much trouble.

An insurance education, acquired sideways

Cuffe’s route back to the industry passed through a data-integration startup called Cheer. He co-founded it in 2019; Retool acquired it in 2020. At Retool, whose software helps developers build internal applications, he eventually led self-serve and new products. The work put him close to companies whose daily operations depended on moving information between systems.

His published Retool work gives that chapter some texture. He wrote about mobile applications and workflow automation. He also covered how Avo made access to customer data more secure, and how Pizza to the Polls managed its operations during the 2020 election. The range ran from engineering infrastructure to pizza logistics. Useful software has never been especially respectful of subject boundaries.

By 2023, he was discussing AI applications on the Building With AI podcast alongside Retool colleague David Dworsky. The conversation covered choosing models, finding valuable uses, and measuring whether applications actually completed their tasks. That last concern would become particularly relevant at Pace. A system can produce a plausible answer while leaving somebody else to finish the job.

Before the operating procedures, the music

At Princeton, where he studied computer science and graduated summa cum laude, Cuffe had already spent time thinking about how people work together. He was a co-director of IgniteSTEM and the entrepreneurship club’s Silicon Valley TigerTrek program. In a 2018 interview, his advice to students put considerable weight on choosing people they would enjoy working with and learn from.

IgniteSTEM supplied a practical experiment. Team members spoke directly with teachers and students attending its conferences, seeing the effects of their work rather than measuring the event solely by whether it happened. There were also group work sessions with music turned up. The early version of his management philosophy had room for both purpose and a decent soundtrack.

These details make an instructive companion to his later product career. The student organizer wanted builders to meet the people using what they built. The software executive would spend years near customer operations. At Pace, the question of whether a product works is inseparable from understanding the procedure, the person responsible for it, and the awkward case that arrives late in the day.

“What if insurance operations could become weightless?”

Jamie Cuffe · January 2026

The unfinished work between applications

In November 2022, Cuffe introduced Retool Workflows. Customers wanted a way to automate business logic as well as build interfaces. Some had tried scripting Retool applications to do it. The new product addressed scheduled jobs, alerts and moving data between tools. The underlying frustration was easy to recognize: an organization could have plenty of software and still lack a straightforward way to connect the work.

That experience helps explain the shape of Pace, founded in 2024. Insurance processing may involve a document, a phone conversation, an email chain and an application built for a very different computing era. Automating one step leaves the handoffs intact. Cuffe’s bet is that agents can carry a complete piece of work across those boundaries.

Pace’s June 2026 product announcement described a managing agent coordinating specialists for documents, calls and computer use, with context following the case. It also described recorded actions, checks against procedures and tests built from known examples. The attraction is continuity. Information learned during a call needs to survive until the record is updated. Forgetfulness is expensive even when it comes with a modern interface.

A customer is waiting at the end of the queue

Consider collecting loss runs, the records of an account’s claims history. Pace’s August description of that workflow involves logging into carrier portals, requesting reports over email, chasing replies and assembling the results for renewal. A missing report can leave everything else waiting. The glamorous part of automation is, on this occasion, remembering to send another email.

Certificates of insurance offer a different kind of difficulty. A request may appear routine, but its wording has to match the underlying coverage. Pace describes agents checking policy terms and issuance rules, filing the result and escalating unusual requirements before a certificate goes out. Speed matters because somebody needs the document; restraint matters because the document makes a representation about insurance.

The deployments give Cuffe something more concrete to discuss than a hypothetical future. Pace’s May 2026 account of its work with Palomar reported that agents resolved more than 90 percent of cases across three high-volume service flows. These involved policy documents, simple cancellations and non-premium endorsements. The figure belongs to those workflows; it is no measure of every decision an insurer makes.

Prudential’s individual life insurance business offers another example. Its agreement with Pace moved from a pilot into production, with customer acquisition operations the focus. The stated benefit was more time for the team to spend on customer relationships. That is the human outcome Cuffe’s argument depends on: less effort administering a service leaves more capacity to deliver it.

In claims review, Pace reported that RYZE’s initial quality-assurance workflow shortened review cycles by 30 percent. Those were internal analyses under specified guidelines, and the workflow did not make coverage decisions or determine damage. The distinction matters. Cuffe’s opportunity consists of particular jobs with particular success criteria, even when the ambition behind them is much broader.

The people who know the exceptions

Cuffe’s public case for automation leaves an important role for people who understand insurance. In February 2026, he published an essay defending brokers’ value as AI entered the industry. His argument centered on advice, expertise and relationships, while treating operational overhead as work that could increasingly be automated. The ambition is to free capacity around the expert’s contribution.

His deployment approach also involves engineers working close to customers. Sequoia’s account of Pace emphasizes the less theatrical requirements: learning operating procedures, integrating with existing systems, auditing decisions and meeting reliability commitments. It is a list that sounds more like a demanding customer meeting than a technology launch. That is where the product has to earn its keep.

Tempo, introduced in July 2026, extends that approach to the people writing the procedures. Operators upload instructions and sample cases to build an agent, then test and improve it. Pace described a quick initial build while acknowledging the difficult final stretch of exceptions and unwritten rules. A veteran operator’s apparently fussy qualification may be the detail that prevents the next mistake.

By September, the company was publishing engineering work on improving extraction agents using historical runs and feedback. The subject had moved from whether a model could read documents to how a production system could learn which fields share evidence, choose models and validate corrections. For Cuffe, the challenge is making the useful result repeatable after the demonstration ends.

A longer clock

The financing accelerated in 2026. Pace announced $10 million led by Sequoia in January, then a $46 million Series B co-led by Thrive and Sequoia in May, with Emergence and Pruven participating. Cuffe said he had passed on higher valuations in favor of the partners he wanted. He has described an intention to build the company over decades.

There is an earlier clue to that timescale. Asked in 2018 where he would go if he could travel through time, Cuffe chose Bell Labs during the transistor breakthroughs. What appealed was the prospect of sustained scientific work with consequences beyond the next business milestone. Even as a student, his imagined destination had a longer clock.

Pace places that ambition inside a very practical setting: the request waiting for a reply, the policy waiting for an update, the customer waiting for a decision. Cuffe has returned to the industry he grew up around with an engineer’s attention to its unfinished jobs. His wager rests on what becomes possible when those jobs finally get done.