Insurance (AI Operations) · Deep dive

Pace

An 'agentic process outsourcer' for insurance — AI agents that navigate carriers' internal apps, read documents, and make phone calls to run the back-office work insurers historically shipped to BPOs, priced by the workflow rather than the seat.

emerging

The question that decides it: Pace's wedge is a per-workflow agentic outsourcer that beats offshore BPOs on cost, speed and error rate for mid-back-office tasks — can it hold enterprise carriers as their own AI ops teams industrialise the same computer-use agents in-house, before Sixfold/Federato/Cytora push down-market from underwriting into servicing and Guidewire/Duck Creek/Applied Systems bundle agentic ops into contracts carriers already renew?

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

Snapshot

Pace is an “agentic process outsourcer” for insurance — AI agents that log into a carrier’s policy admin system, read documents, place phone calls, and complete the mid-back-office workflows (intake, servicing, endorsements, renewals, claims triage, data entry) carriers historically shipped to Genpact, EXL and WNS. Founded 2024 in New York by Jamie Cuffe, launched July 23, 2025; inside nine months raised a $10M Sequoia Series A (Jan 27, 2026) and a $46M Series B co-led by Thrive and Sequoia at a reported $375M valuation (May 27, 2026), with Emergence and Pruven participating. It matters because Pace is the sharpest expression yet of the bet that generative AI collapses insurance BPO the way SaaS collapsed on-prem — 250,000+ workflows completed and named deployments at Prudential, Palomar, Convex, WTW, Newfront and The Mutual Group.

Founding story

Cuffe’s biography reads engineered for the pitch. His father spent his career in Lloyd’s of London operations — reinsurer, then Lloyd’s cover holder — so Cuffe grew up between London, New York and Bermuda watching the plumbing. Princeton, a stint at Sequoia as investor/scout, then consumer social startup Cheer (sold to Retool in 2020), then Retool as Head of Self-Serve and New Products. Retool matters technically: Cuffe’s seat put him in front of the first wave of teams wiring LLMs into internal apps to automate ops. Bryan Schreier at Sequoia had backed Cheer and stayed close; when Cuffe started Pace in 2024, Schreier led the Series A almost immediately.

The thesis (Sequoia podcast, 2026; The Insurer TV, Jul 15, 2026): insurance is uniquely suited to agentic AI because its cost structure is dominated by structured, rule-heavy, document-shuffling work carriers already outsourced once at a per-transaction fee. Live buyer (the head of ops with a BPO invoice), live metric (cost per workflow, error rate, cycle time), no political fight about automation — that was won a decade ago. The wedge is that this time the work does not need to leave the building.

How it works

A Pace deployment starts with a workflow, not a data model. The carrier picks a high-volume, well-defined process — Palomar’s example: inbound service requests for document requests, simple cancellations and non-premium endorsements — and Pace scopes it as a discrete agent that sits on top of the existing stack, logs into the same policy admin system a human CSR would, reads the same emails and PDFs, places outbound calls (Pace agents make calls, not just handle them), and writes back into the same systems of record. Nothing is ripped out.

Under the hood: foundation models orchestrating browser and desktop automation against carrier apps, document understanding for ACORD forms and loss runs, and a voice layer for calls. Each agent is deployed against a specific KPI and iterated in production; Pace claims a 90-day time-to-first-agent at Palomar, after which the carrier’s ops team took control of SOPs and testing. That handoff — Pace ships and instruments the agent, carrier ops owns it — is the business model dressed as software. The uncomfortable question is how much of the “agent” is still a Pace engineer in a Slack channel when the workflow breaks at 2am.

Product and business overview

Four workflow families: submission intake (broker emails, ACORD forms, SOVs, loss runs); policy servicing (endorsements, cancellations, document requests, COIs — the Palomar wedge, voice and email); claims handling (FNOL intake, document collection, triage); and new business + renewals ingestion (Convex’s use case). Named deployments: Prudential (Dec 3, 2025, personal and worksite life) on acquisition and servicing; Palomar (specialty P&C, publicly listed) on servicing; Convex (Lloyd’s-adjacent specialty) on renewals ingestion; WTW (top-three global broker) as an operating-model shift; Newfront and The Mutual Group on broker/mutual automation. Notably absent: heavy claims adjudication, regulated pricing, bind authority — the boundary is squarely on ops, not underwriting. That is the deliberate difference from Sixfold, Federato and Kalepa, all of whom fight for a seat inside the underwriter’s decision.

Business model and pricing

No published price list, but the pitch is explicit: priced per workflow completed, benchmarked against the offshore BPO rate for the same task. TAM framing (Fortune, Jan 27, 2026; Sequoia, 2026) is $70B annual insurance BPO spend inside a ~$400B financial-services BPO market. That maps onto an existing carrier budget line — the sale is a swap, not a new expense — and invites the customer to expect the invoice to come down as the agent improves, because BPO contracts already work that way. Closer to a services-consumption meter than a per-seat subscription, which lets Pace claim carriers “scale their agentic workforce to tens of millions of tasks” without a seat ceiling. The corollary the deck does not dwell on: per-workflow revenue is directly coupled to how much of the customer’s work is genuinely replaced. If the agent stalls at 50% resolution, revenue stalls with it.

Traction over time

Nine months from Series A to a $46M Series B at ~$375M is not normal — it is a bet on founder-market fit as much as on numbers. The 250,000-workflows figure is the most concrete data point, but Pace has not broken it down by customer or workflow type, and 28 employees is a very small number to be running production automation at Prudential.

Market analysis

Estimates depend on scope. IMARC pegs global insurance BPO at ~$7.5B in 2024 growing to ~$10.4B by 2033 (3.6% CAGR). Research and Markets puts it at $15.2B in 2025 rising to $25.5B by 2030 (10.8% CAGR). Mordor Intelligence uses the broadest cut: $68.4B in 2026 to $93.1B by 2031 (6.4% CAGR). Pace’s own $70B framing aligns with Mordor, and Sequoia extends it to a ~$400B financial-services BPO market.

The structural forces move Pace’s way: combined ratios stay under pressure, a wave of underwriter and CSR retirements is straining carrier operations, foundation models finally make unstructured-document work cheap, and Conning’s 2025 survey found early or full LLM adoption among insurers jumped from 18% to 63% in twelve months. The counterforce is that the same cheapness lowers the barrier for everyone — carriers, incumbent vendors, and the BPOs themselves. Genpact and EXL are not going to sit still while a 28-person startup lifts their book.

Competitive intel

The set (sidebar has full profiles) splits three ways. Agentic-underwriting startups — Sixfold ($50M+, Guidewire-backed), Federato ($80M), Kalepa ($16M) — start upstream but are pushing horizontally into servicing and claims; if Sixfold’s autonomous underwriter learns endorsements, it lands on Pace’s core workflow. Incumbent AI-in-insurance vendors — Cytora (Applied Systems, Autopilot launched March 2026), Ushur ($108M), Roots Automation ($22M Series B) — are already inside the buyer. Cytora inside Applied is the sharpest structural threat: ingestion and (soon) servicing shipped as part of the system carriers already run. And the incumbent BPOs — Genpact, EXL, WNS, Cognizant, TCS — are the loudest, best-resourced competitor Pace never mentions. They own the invoice line and can renegotiate a five-year MSA faster than procurement can onboard a new vendor. Pace’s real defence is speed: it wins if it can prove three workflows per customer in the time a BPO takes to draft a change order.

History and evolution

• 2024 — Founded in New York by Jamie Cuffe; small pre-seed build, no public press.
• Jul 23, 2025 — Public launch as the “agentic process outsourcer for insurance”; first named customers The Mutual Group and Newfront.
• Dec 3, 2025 — Prudential Financial deployment announced (BusinessWire).
• Jan 27, 2026 — $10M Series A led by Sequoia; Fortune exclusive.
• Q1 2026 — Palomar case study: 90-day first agent, 90%+ resolution.
• May 27, 2026 — $46M Series B co-led by Thrive and Sequoia, with Emergence and Pruven; ~$375M valuation, ~28 employees; Convex and WTW named.
• Jul 15, 2026 — Cuffe interview on The Insurer TV.

The stumbles are unwritten — the company is barely a year public. The tests to watch: whether Prudential and WTW scale from an initial workflow into an eight-figure MSA, and whether the 28-person headcount can support the deployment cadence a Sequoia/Thrive-backed enterprise business is expected to hit.

What people say

The case for

Public commentary is uniformly bullish, as expected for a nine-month-old company with a Sequoia partner and a Thrive lead. Fortune (Jan 27, 2026) framed the Series A as “enterprise AI colliding with insurance”; Forbes’ Anna Tong (May 27, 2026) covered the $46M round as the archetypal Thrive-Sequoia bet on the AI-BPO thesis. The Palomar case study reports agents running three highest-volume service flows across voice and email with 90%+ resolution and no additional CSR headcount — a specific, testable number rather than a vague endorsement. Trade press (The Insurer, Jul 15, 2026) frames Pace as one of two or three AI insurance names investors will underwrite at growth multiples right now.

The complaints

No Glassdoor, G2 or trade-press criticism yet — the company is too new. That absence is itself the honest complaint: every publication has been an announcement or a founder interview, not independent review. The substantive cuts: Valuation-to-fundamentals — ~$375M on a 28-person company ten months post-launch prices in a lot of story. Workflow density — 250,000 workflows across six-plus carriers over ten months is real but averages out to something a small ops team with a smart wrapper around foundation models could run. In-house risk — the buyers are exactly the enterprises with the biggest incentive to build the next agent themselves, and per-workflow pricing gives them a clean unbundling line. BPO counter-attack — Genpact and EXL will arrive with a bundled “AI + delivery centre” price CFOs find hard to refuse. Founder concentration — public materials name Cuffe only; no disclosed CTO or head of AI, thin governance for production agents inside Prudential.

Outlook: the open question

Pace is the cleanest live AI-eats-BPO bet in enterprise software, run by a founder with unusually good buyer fit. Wedge is real, customers named, pricing honest, cap table (Sequoia + Thrive + Emergence) funds the next two rounds regardless of the tape. Pace wins if per-workflow agentic operations is the durable form factor — carriers keep swapping BPO invoices for Pace invoices rather than building in-house, the 90-day Palomar motion generalises across ten more carriers before Sixfold, Cytora or Applied bundle equivalent capability into the renewal, and 250,000 workflows compounds to tens of millions in 2026 without the 28-person team snapping. It struggles if (a) carriers decide the moat is the workflow logic and stand up internal ops-engineering that keeps only the hard 20% on Pace; (b) Applied, Guidewire or Duck Creek ship “good enough” agentic servicing inside the policy-admin renewal; or (c) Genpact and EXL redraw MSAs with AI pricing that closes the ROI gap. The tell over the next twelve months: whether Prudential and WTW expand from one workflow into ten, and whether Pace names a CTO and doubles headcount without losing operational velocity.

How to attack it

The wedge is not to out-build Pace on insurance — it is to out-flank it on the workflow-instrumentation layer Pace hides from customers. Foundation models plus computer-use APIs give any credible team the raw capability; the real moat is the tooling to spec, deploy, evaluate and iterate agents against carrier production traffic — the observability layer that turns “AI works in demo” into “AI works overnight without a human in a Slack channel”. A credible attacker builds that layer and either sells it as infrastructure (an “agentic ops platform” carriers use to build their own agents) or uses it to undercut Pace’s per-workflow pricing with a lighter services layer. This exploits Pace’s largest exposure: enterprise buyers with real budget prefer to own operational tooling and rent capability, not the reverse.

Second vector: vertical BPO where insurance isn’t. The same agentic-BPO thesis in health-plan ops (prior auth, appeals, member services), government back-office (Medicaid, DMV, SNAP), or freight/logistics documentation (bills of lading, customs, chargebacks) has the same set-up — large BPO invoice, per-workflow price, no dominant agentic vendor — and none has a founder with a Lloyd’s-of-London father and Sequoia on speed-dial. A Q3 2026 start-up has a two-year head start because Pace is committed to widening within insurance carriers first.

Exploitable weaknesses, each sourced: valuation exposure (~$375M at 28 employees, Forbes May 27, 2026); founder concentration (Cuffe named alone; no disclosed CTO); workflow narrowness (deployments cluster in servicing and intake; adjudication, regulated pricing, bind authority off-limits); incumbent bundling (Cytora Autopilot inside Applied, March 2026; Guidewire backs Sixfold); BPO counter-price (Genpact, EXL, WNS can bundle AI pricing to close the ROI gap without rip-and-replace).

Adjacent-segment play

Take the same agent stack out of carriers and into insurance’s downstream buyers — brokers, MGAs, TPAs — where the same workflows run without the AI budget. Pace has one broker (Newfront) and one global broker mid-shift (WTW), but its centre of gravity is carriers. The mid-market broker and MGA layer — thousands of firms doing certificates, endorsements, submissions and renewals by hand — has the same document chaos and phone-heavy workflows but a very different sales motion (SMB, self-serve, per-seat) a purpose-built entrant could own. No dominant agentic-BPO for brokers exists yet.

Second adjacency: health payer operations — prior auth, appeals, member services, claims intake, provider data — where workflow shapes rhyme exactly but buyer, regulatory frame and sales motion are different enough that Pace is unlikely to redirect. Cohere Health and Anterior are the closest analogues; neither is set up as a per-workflow “agentic BPO” against Cognizant and Optum’s back-office invoice.

What does not work: out of financial services entirely. The wedge depends on a large existing BPO invoice, high document density with structured schemas, and a buyer sophisticated enough to trust an agent inside systems of record. Retail, construction and hospitality fail on at least one. Pace generalises across regulated document-heavy verticals — not every industry with a back office.

Sources and further reading

• Pace Raises $46 Million Series B to Build the Next Century of Insurance (BusinessWire, May 27, 2026)
• Thrive And Sequoia Back Pace With $46 Million To Automate Insurance’s Back Office (Forbes, May 27, 2026)
• Exclusive: Pace raises $10 million from Sequoia as enterprise AI collides with insurance (Fortune, January 27, 2026)
• Partnering with Pace: Making Work Weightless (Sequoia Capital, January 2026)
• Pace selected by Prudential Financial to help automate its insurance operations with agentic AI (BusinessWire, December 3, 2025)
• Case Study: Palomar taps Pace to automate specialty insurance operations (Pace, 2026)
• Agentic AI enabling carriers to ‘do significantly more with less’: Pace’s Cuffe (The Insurer TV, July 15, 2026)
• What’s the Future of Vertical SaaS in an AGI World? Jamie Cuffe, CEO of Pace (Sequoia Capital podcast, 2026)
• Insurance BPO Services Market Size, Share, Report 2031 (Mordor Intelligence, 2026)
• AI’s Everywhere in Insurtech: What 188 Startup Applications Reveal (InsurTech NY, 2026)