Insurance · Deep dive

Pace

An NYC insurance-native AI ops company selling per-transaction agents that eat submission intake, policy servicing and FNOL — pitched as an AI-native replacement for the $70B insurance BPO stack, and racing incumbent core vendors to own the workflow.

emerging

The question that decides it: Does Pace's horizontal insurance-agent layer survive Duck Creek and Guidewire bundling native AI submission intake, triage and FNOL into the policy-admin systems they already own — or does the value pool collapse into the core suites within 24 months, leaving Pace fighting for whatever workflows the incumbents choose not to build?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Pace is a New York-based AI operations company selling autonomous agents that do insurance back-office work end-to-end — submission intake, policy servicing, endorsements, renewals, first-notice-of-loss, premium audits, bordereaux. Founded in 2024 by ex-Retool operator Jamie Cuffe, it raised a $10M Series A led by Sequoia in January 2026 and, four months later, a $46M Series B co-led by Thrive Capital and Sequoia at a reported $375M post-money — roughly $56M raised in total. It counts Prudential, WTW, Newfront, The Mutual Group, Palomar, Convex US and Ryze Claim Solutions among named customers, and says its agents have completed more than 250,000 insurance workflows since launch. The pitch is blunt: replace the $70B insurance BPO industry with software that gets paid per successful transaction, not per seat. The question is whether the core-suite vendors let it.

Founding story

Cuffe’s biography is unusually on-nose for this bet. He grew up between London, New York and Bermuda while his father ran operations for a Lloyd’s of London cover holder, so the sights and smells of insurance operations — the piles of ACORD forms, the loss runs, the offshore teams handling them — were the family business. He went to Princeton, then into software: his first startup, Cheer, sold to Retool in 2020, after which he ran self-serve and new products at Retool during the Retool AI build-out. That is where the thesis crystallised. Watching enterprise operators wire LLMs into internal tools convinced him the economics of insurance BPO — an army of humans re-keying data between broker email, policy admin systems and rating engines — could be flipped by agents that navigate the same apps.

Sequoia partner Bryan Schreier, who had backed Cuffe at Cheer, wrote the first cheque. Pace launched publicly in mid-2025 and, in Cuffe’s own framing, is not a co-pilot vendor: it sells outcomes, not tools. The founding pitch is that a $70B insurance-BPO industry — roughly $400B if you include adjacent financial-services ops — is uniquely well suited to agentic AI because the tasks are high-volume, rules-heavy, document-bound and, until now, only economical to run with cheap offshore labor.

The frequently repeated attribution that Cuffe was previously at Meta and Anthem does not hold up in the public record; his tenure is at Cheer and Retool, and the insurance angle came from his father’s career, not a prior carrier job.

How it works

Pace deploys what it calls agents — LLM-driven workers that log into a customer’s real systems and do the job a human ops analyst would do. In the submission-intake flow, an agent monitors a broker inbox, parses the incoming email plus attached ACORD forms and loss runs, extracts the data, checks it against the carrier’s appetite and guidelines, keys it into the underwriting workbench or policy admin system (Guidewire, Duck Creek, Vertafore, or the carrier’s own portal), pulls quotes from the rating engine and returns them. In policy servicing, agents process endorsements, non-premium changes, audits and renewals. In claims, agents handle FNOL intake — from initial report through file assembly, coverage checks and triage/assignment — and, in some accounts, make outbound phone calls to gather missing information.

Two design choices matter. First, Pace agents work over legacy interfaces, mainframes and desktop apps as well as APIs, because the reality of a Tier-1 carrier is that most of the labor lives in green-screen systems no vendor is rewriting. Second, sensitive actions are logged and human-verified before they touch production, and that verification loop feeds back into the agent’s working instructions — Pace’s version of the standard “human-in-the-loop supervised learning” pattern. The company claims 98–99% accuracy per action, with some deployments above 99.9%; those are self-reported and not independently audited.

The unit of production is a completed workflow, not an FTE. That is what lets Pace charge per successful transaction and undercut a BPO invoice priced per hour.

Product and business overview

The company markets a single AI operations platform with three primary workstreams. Submission Intake covers broker-email parsing, ACORD/loss-run ingestion, appetite checks, data entry into underwriting/rating systems and quote return, for both commercial and personal lines. Policy Servicing covers endorsements, premium audits, NIGO (not-in-good-order) resolution, underwriting data collection, renewals, binding and bordereaux. Claims covers FNOL and death-notice intake, claim-file preparation, triage and assignment against adjusting guidelines. Around all three sits an operator console for exception review, audit trails and the human-verification loop.

The customer set is deliberately across the value chain: retail broker (Newfront), global broker/consultancy (WTW), primary specialty carrier (Palomar), diversified insurer (Prudential, The Mutual Group), specialty London-market-adjacent carrier (Convex US) and a claims TPA (Ryze). That breadth is a feature — it signals horizontal applicability — and a risk, because none of those buyers looks like another and each integration is bespoke.

Pace describes itself as an “agentic process outsourcer” (APO), a label Cuffe coined to reframe the category away from software vendors and toward the BPO budget line. The rhetorical move matters: it aims Pace’s price comparison at Genpact and EXL invoices, not at other insurtech SaaS seats.

Business model and pricing

Pace does not publish a price list. What has been consistently disclosed, including in its own materials and in the Series B commentary, is that pricing is performance-based on completed transactions — customers pay for the outcome (a submission entered, a policy endorsed, an FNOL triaged) rather than a per-seat licence or a fixed annual platform fee. That is the intended anchor against BPO economics, where the customer pays for human hours regardless of throughput.

Economically, the model implies gross-margin structure closer to a software company than to a BPO, provided the agents actually complete work autonomously at the claimed 98–99% rate. Every action that kicks out to human review erodes margin, so the company’s forward-deployed engineering team — which Cuffe has said is the primary destination for Series B capital — is really a margin lever as much as a customer-success function. Revenue is not disclosed and the 250,000-workflows figure through May 2026 is a volume metric, not a dollar one; at typical BPO price points that would imply low-single-digit millions of run-rate revenue if fully priced as replacement work, but the mix is unknown.

Traction over time

The velocity is unusual — two rounds and a $375M mark inside 18 months of founding — and reflects both the fundraising climate for enterprise AI in 2026 and the strategic value of the customer logos more than any disclosed revenue.

Market analysis

The stated wedge is the insurance BPO services market, which Mordor Intelligence sizes at roughly $68.4B in 2026, growing to ~$93.1B by 2031 (6.4% CAGR); Cuffe cites ~$70B in his own materials and ~$400B if you include adjacent financial-services operations. Fortune Business Insights pegs the broader AI-in-insurance market at ~$13.5B in 2026 growing to ~$154B by 2034 (35.7% CAGR). Both numbers are large; both have been sold repeatedly.

The structural forces are real. Insurance is a document-heavy industry running on legacy systems, with margin pressure in personal lines and rate volatility in specialty, and carriers have already outsourced the low-value ops work overseas. That is precisely the workload agents can now attack — unstructured inputs, deterministic outputs, high volume. What is genuinely different in 2026 is that model quality and integration tooling have crossed the threshold where an agent can plausibly complete an ACORD-to-quote workflow at accuracy the underwriter will tolerate. The pull from carriers is real enough that Prudential, WTW and Palomar are quoted in Pace’s own announcements — those are not experimental logos.

The honest counterweight: the entire insurance-AI cohort is chasing this same thesis with the same LLMs, and the true TAM Pace can capture is not the BPO number but whatever slice the core-suite vendors and the BPOs themselves do not defend. That is probably a large business, but a smaller and more contested one than the $70B slide suggests.

Competitive intel

The competitor with the most immediate strategic weight is Duck Creek, owned by Vista. On 28 April 2026 it launched an insurance-native Agentic AI Platform, and it has separately acquired underwriting-workbench Send to build what it describes as the industry’s only “agentic underwriting-to-core” stack. That places native AI agents inside the policy administration system Pace has to integrate around. Guidewire is the same story in a different flavour — its Insight AI and marketplace push agents next to the core suite that most Tier-1 P&C carriers already run. The bundling risk is not hypothetical; it is announced.

Beside the core suites, the field is crowded. Cytora, acquired by Applied Systems in 2025, hits Pace’s submission-intake wedge with a broker distribution channel Pace lacks. Sixfold launched a straight-through quote-and-bind agent in June 2026, competing for the intake-to-quote budget. Federato (Series D, ~$180M raised) sells an AI underwriting workbench; Kalepa, FurtherAI, Comulate and Convr each own a slice. Every one of these is well-capitalised and shipping.

Then there are the incumbents whose revenue Pace explicitly wants to take: Genpact, EXL, WNS, Accenture, Cognizant, Xceedance, Sedgwick. Collectively they earn a large fraction of that $68B BPO spend. The threat they pose is not that they build a better agent — they will not — but that they buy or license the same underlying models, absorb the productivity gain, and re-price their own contracts down to keep the account. That defence is boring, cash-generative and hard to disrupt without a compelling switching event on the customer side.

History and evolution

• 2024: Pace founded in New York by Jamie Cuffe.
• Mid-2025: Public launch as an agentic process outsourcer; first commercial deals.
• Dec 2025: Prudential publicly names Pace as an agentic-AI partner for insurance operations.
• 27 Jan 2026: $10M Series A led by Sequoia; Fortune covers the round.
• Q1 2026: Palomar case study published: 90%+ end-to-end resolution across three of the carrier’s highest-volume service flows.
• 28 Apr 2026: Duck Creek launches its insurance-native Agentic AI Platform — the clearest visible sign that the core-suite bundling risk is now live.
• 27 May 2026: $46M Series B co-led by Thrive and Sequoia at a reported $375M post-money.
• Jun 2026: Sixfold launches a competing straight-through underwriting agent.
• Mid-2026: Ryze Claim Solutions reports ~30% cycle-time reduction with Pace in claims; Convex US live in data ingestion for new business and renewals.

The stumbles, so far, are more about what has not happened than what has: no independent audit of the accuracy claims, no disclosed revenue, no named enterprise-scale contract at the multi-hundred-FTE-replacement size the BPO framing implies. That is normal for an 18-month-old company; it is also the gap the next 12 months have to close.

What people say

The case for. Trade press and investors have been unusually credulous, which is a signal in itself. The Insurer’s interview with Cuffe after the Series B framed the raise as validation that “AI agents completing insurance operations end-to-end” — not co-pilots — are unlocking products and markets carriers previously could not economically serve. Forbes’s Anna Tong wrote up the round as a real-work agentic bet on the boring parts of insurance, not another chatbot. The recurring positive theme from carriers is throughput plus consistency: Palomar publicly credits Pace with resolving more than 90% of three high-volume service flows end-to-end; Ryze cites a ~30% reduction in claims cycle time; Prudential positioned Pace as a way to compress policy servicing and issuance work. The customer set — Prudential, WTW, Newfront, The Mutual Group, Palomar, Convex US — is more Tier-1 than an 18-month-old company usually has by Series B.

The complaints. The critical read has three planks and every one is a real one. First, the crowded field: FurtherAI, Sixfold, Cytora, Federato, Convr, Kalepa, Comulate and half a dozen quieter startups are all pitching insurers on some version of “agentic AI for the workflow” — competitors like FurtherAI publish direct comparison content aimed at buyers evaluating Pace, and the differentiation between them is largely execution and go-to-market, not defensible IP. Second, the “wrapper on GPT” critique, which trails every horizontal agent company in 2026: Pace’s moat is workflow depth, forward-deployed engineering and customer trust, not proprietary models — and if those advantages can be replicated by a rival with better distribution, the price point compresses. Third, and most seriously, the incumbent-vendor bundling risk: Duck Creek’s April 2026 Agentic AI launch and its Send acquisition, plus Guidewire’s marketplace push, mean the workflows Pace targets are being productised inside the systems Pace integrates with. On the other side, the BPOs Pace wants to displace — Genpact, EXL, WNS — are not going to sit still while their most profitable contracts get repriced per transaction; they are more likely to build or buy the same capability and defend the account. Employee-side signal is limited (no meaningful Glassdoor footprint at 28 heads) but hiring is aggressive, and forward-deployed engineering roles dominate the openings, which is consistent with a company that is genuinely delivering per-transaction outcomes rather than shipping shrink-wrap software — and also consistent with a company whose margin depends on how much manual configuration each new customer requires.

Outlook: the open question

Pace’s answer resolves on whether Duck Creek and Guidewire choose — or fail — to make native AI submission intake, triage and FNOL a first-class part of the core policy-admin suites they already sell. If they do, and if it works, Pace’s addressable surface at every Duck Creek and Guidewire shop shrinks to the exception workflows the core vendors decline to build. In that world Pace is a specialist add-on, worth something, but not a $70B-BPO-replacement business at a $375M valuation. If the core vendors under-deliver — which is the historic default for enterprise incumbents facing AI-native challengers — Pace has a real shot at being the horizontal ops layer the way ServiceNow became the horizontal workflow layer for IT.

Three things have to be true for the bullish case to hold. First, the accuracy claims have to survive independent scrutiny at Tier-1 carrier scale: Prudential-sized deployments generating audit trails a regulator can inspect. Second, at least one lighthouse customer needs to migrate a meaningful multi-hundred-FTE BPO contract onto Pace and disclose the economics, because the “we replace Genpact” pitch is a slide until the numbers exist. Third, Pace has to move faster than Duck Creek and Guidewire can bundle — which is a genuinely winnable race, because core vendors are historically slow and their AI launches are often marketing before they are product, but not a race Pace has yet won.

The bear case is easier to sketch. Duck Creek’s Send-plus-Agentic-AI stack is real and shipping; Guidewire’s marketplace has distribution Pace does not; Cytora inside Applied Systems has broker channels Pace does not; Sixfold and FurtherAI are perfectly capable competitors at the top of the funnel; and Genpact and EXL can adopt the same LLMs and re-price. If any two of those pressures land, the $70B TAM narrative collapses into a much narrower per-transaction niche, and Pace has to justify a $375M mark on run-rate revenue that has not yet been disclosed. The Series B buys 18–24 months to prove the answer. The next raise, and whether it is priced on revenue or on story, will tell you which world we are in.

How a challenger would attack it

Turn its own playbook on it before the moat exists. Pace is 28 people, 18 months old, and its differentiation is admittedly execution, not IP — the models are rented, the workflows are learnable, and FurtherAI already publishes comparison content aimed at Pace’s evaluators. The cleanest attack exploits Pace’s structural bind: every customer is bespoke (a retail broker, a Tier-1 life insurer, a claims TPA — none looks like the others), so forward-deployed engineers are both the product and the margin. A challenger that picks one buyer shape — say, specialty MGAs on a single core system — ships a productised agent with two-week deployment, and prices per transaction below Pace’s rate wins on cycle time while Pace’s team is buried in Prudential’s green screens. The second vector is verification: Pace’s 98-99% accuracy claims are self-reported with no independent audit, and its human-review loop is invisible to buyers. A rival that ships regulator-grade audit trails and publishes third-party accuracy attestation weaponises the disclosure gap in every procurement process. Third, distribution judo: Cytora-inside-Applied shows the move — sell through the channel that already owns the buyer. A challenger that becomes Duck Creek’s or Guidewire’s blessed marketplace agent gets handed the exact accounts Pace must win one bespoke integration at a time.

Same playbook, new buyer

Aim the agentic-process-outsourcer model at ops books the insurance-AI cohort is ignoring. Every named competitor — Sixfold, Federato, Cytora, Kalepa — is fighting over P&C submission intake and underwriting, the most crowded slice of the thesis. The identical mechanics (document-heavy, rules-bound, high-volume work done today by offshore FTEs, priced per completed transaction) apply untouched in reinsurance back-office — treaty accounting, claims bordereaux reconciliation between cedents and reinsurers, where the London market still runs on spreadsheets — and in TPA-land beyond Ryze: workers’ comp medical bill review and benefits administration, where the buyer is a claims administrator, not a carrier, and no Duck Creek or Guidewire bundling threat exists because there is no core suite to bundle into. The third shift is geography: European and Lloyd’s-market operations, where Cuffe’s cover-holder upbringing is the founding story but not yet the go-to-market, and where US-focused rivals have no regulatory or format fluency. Pace can’t chase all of this: its Series B mandate is scaling US carrier logos toward the revenue its $375M mark requires, its 28-person team is the constraint, and every bespoke vertical it adds worsens the margin problem its investors are already watching.

Sources and further reading

• Forbes — “Thrive And Sequoia Back Pace With $46 Million To Automate Insurance’s Back Office” (Anna Tong, 27 May 2026)
• The Insurer — “Pace’s Cuffe says AI agents are unlocking products and markets carriers couldn’t previously serve after $46M raise” (28 May 2026)
• Fortune — “Exclusive: Pace raises $10 million from Sequoia as enterprise AI collides with insurance” (27 Jan 2026)
• BusinessWire — “Pace Raises $46 Million Series B to Build the Next Century of Insurance” (27 May 2026)
• Sequoia Capital podcast — “What’s the Future of Vertical SaaS in an AGI World? Jamie Cuffe, CEO of Pace” (2026)
• Duck Creek — “Duck Creek Launches Insurance-Native Agentic AI Platform” (28 Apr 2026)
• Duck Creek — “Duck Creek Acquires Send, Creating the Industry’s Only Agentic Underwriting-to-Core Platform” (2026)
• Pace — “Case Study: Palomar taps Pace to automate specialty insurance operations” (2026)
• Financial Content / BusinessWire — “Pace selected by Prudential Financial to help automate its insurance operations with agentic AI” (4 Dec 2025)
• Mordor Intelligence — “Insurance BPO Services Market Size, Share, Report 2031” (2026)
• Fortune Business Insights — “AI in Insurance Market Size” (2026)