What’s the Future of Vertical SaaS in an AGI World? Jamie Cuffe, CEO of Pace
In a Nutshell
Pace, led by CEO Jamie Cuffe, is an agentic AI outsourcer automating insurance carriers' back-office operations like submission intake and claims processing, replacing low-margin BPOs (10-15% margins) with scalable AI agents targeting 80% margins and superhuman speed/accuracy on mission-critical SOPs. Forward-deployed engineers ensure 100% pilot-to-production success with top insurers by closing gaps on-site, driving 50-75% cost savings, 24/7 scalability, and full observability. Long-term vision: emulate Constellation Software by aggregating AI services across $400B BFSI back-office markets, prioritizing execution, trust ("be the rock"), and high-urgency culture.
These notes were generated by AI and may contain inaccuracies.
Building AI agents to take any standard operating procedure and run that process end-to-end is the direction at Pace.
Jamie Cuff, founder and CEO of Pace, explores how AI agents are revolutionizing insurance and the future of work in high-complexity, highly regulated environments. Breakthrough is architecting AI to handle nuanced end-to-end processes requiring human touch and judgment. Discussion covers building trust with biggest and oldest companies, forward deployed engineers as secret weapon for AI companies, and AI replacing BPOs rewiring sectors of the economy.
Pace is the agentic process outsourcer for insurance. Shift from onshore to BPO work over decades; thesis is shift to outsourcing to AI over next decade. Focus exclusively on insurance carriers, automating critical back office operations traditionally outsourced to BPOs.
Top of class at Princeton, investor at Sequoia, rising star at Retool, then insurance BPOs. Grew up around insurance: born in London, grew up in New York, London, and Bermuda (insurance capitals). Dad ran operations for a reinsurer and Lloyd's cover holder in London. Saw problems at Retool where many customers from financial services, especially legacy insurers using Retool for custom internal tooling (e.g., Ethoses, Vouchers, Progressive, Bircher Halfway). Forward deployed engineering gave front-row seat to problems like policy administration, claims submission intake, core workflows. Software makes BPO 10% better (e.g., OCR, RPA). Post-ChatGPT, opportunity to completely replace vertical services. Insurance good starting point: lingua franca is emails and PDFs due to intermediary players, no common system, lack of APIs. AI agents bridge gap created by manual data entry from documents/emails.
BPOs are tens of billions dollar public companies at 10-15% gross margins with tens to hundreds of thousands of people. That's failure case for Pace: multi-billion BPO without changing economics. Goal: 80% gross margins with hundreds of thousands of AI agents and small team of insurance experts, across many verticals. Path to high gross margins essential; flywheel where product improvements increase gross margins. Even if lower today, margins shift up as tackling harder tasks.
Focus on mission-critical processes outsourced end-to-end. Prior tech good for data extraction but not full problem (reasoning, human judgment, writing back to systems). Avoid workflow DAG builders with AI nodes in boxes plus code layer—misses AI point, short on AI. Build for AI agents to run any standard operating procedure end-to-end.
Vast majority of workflows: no human in the loop, AI end-to-end. Operations team for most complex; escalate to customers. Primarily used in onboarding: label examples for 90% accuracy to 99.5%+ required. Self-improvement loop reduces human involvement over time; ops team moves to next task.
Focus on already outsourced at scale: codified SOPs, accuracy checks (internal review or QA sampling), easier change management (spin down BPO vs. retrain W2 workforce). High-volume use cases for biggest ROI. Start top of funnel: submission intake (extract data, run business logic, write to new business underwriting platforms). Claims side: first notice of loss or claims intake. Expand to policy servicing lifecycle (endorsement processing, policy admin, billing) and claims (intake to payout, QA).
BPO spend often eight figures+. Achieved 50-75% cost savings quickly. Beyond cost: superhuman speed closes business faster (submissions go to multiple carriers). Scalability: handles seasonality (e.g., open enrollment, hurricane season)—AI works 24/7, cranks backlog. Observability: past BPO black box; now see task stages, reasoning, bottlenecks via software workflows—enables redesign.
95% AI pilots fail (MIT studies), but portfolio companies succeed. Pace: 100% pilot success to production. Core value: closing the distance. Forward deployed engineering: on-site, fly to customers, deep insurance focus. Partnership proves value quickly, counters demo letdowns. Forward deployed engineers profile: former founders (engineers wanting commercial or vice versa)—do whatever to succeed (prompt tuning, evals, integrations).
Hire both AI experts and insurance experts, sit closely with customers for constellation learning. Insurance from top-five carriers on mission-critical ops; AI from cutting-edge (e.g., former YC founders). Few pre-ramped on both; hires pick up quickly, resonate with thesis.
Short-term: more deployed engineers for customer flywheels (expand use cases, BPO budgets) and internal (ship code to product, tight feedback loops). Deployment times down, leverage up as push to software. Long-term: software codified workflows; customers build in product (e.g., from 1-2 to 9-10 workflows via business analysts). Team balance: 80% engineering, forward deployed at top, GTM. Okay long-term if makes customers successful.
Seeing a big forward deployed engineering team constantly increasing leverage to do more is a great story for public markets. Check back in 10 years as we're going public.
Bottoms-up distribution vs top-down: be pegged at one end, not stuck in the middle. ChatGPT and OpenEvidence are bottoms-up tools people love and adopt. Harvey is a solution for law, Sierra for customer communications, Pace for insurance.
In top-down solutions, sell trust. Customers see AI magic and potential but also issues. Role is to say, "Trust us, we'll make sure you're good, AI will benefit your business, we'll do whatever it takes."
Forward deployed engineers build customer confidence and create feedback loops, solving real problems where rubber meets the road, going deeper into workflows.
Insurance is the business of trust.
Pace value: Be the rock – customers are there for their customers on hardest days, Pace is there for customers every day. Work with most trusted brands, top insurers, build trust, credibility, results, then expand. Like Sierra and Harvey: succeed with top customers (highest volumes, ROI), product speaks for itself.
Values: Close the gap, Be the rock, Set the pace.
Set the pace reflects high urgency, impatience, luck of current moment building business. Amazing time, lucky to work with companies, charting path to make future reality. Company named Pace about speed, value for customers and internally, step toward bigger mission.
Inspired by internship learning Andrew Reed's favorite company: Constellation Software (largest in Canada, ~$70-80B, compounded massively; technically favorite non-Sequoia public company).
Mark Leonard built for 30 years: aggregate niche verticals not scaled enough individually, but federated model creates legendary business.
Opportunity: do same in services market (orders of magnitude larger than software). BFSI BO spend ~$400B/year (same as global cloud software), $70B in insurance. Be platform aggregating niche vertical services across BFSI and beyond, transforming industry economics.
Vertical AI targets sneaky big markets (not household B2B SaaS). BPO interface clean: defined objectives, built-in eval, champion-challenger models.
Customers see AI transforming high-volume repetitive tasks needing human judgment – exactly BPO. Standard operating procedures to model, built-in accuracy checking, faster change management, hard cost/ROI savings.
Baseline BPO accuracy: 5-10% error rates, sometimes much higher. AI can exceed human accuracy/consistency (e.g., claims/policy QA on 300-page docs, hundreds of rules – humans miss on page 298, especially 15th on Friday). AI scales best rep, always perfect attention, seen every corner case (like Waymo drivers).
Engagements start with COO office, identify AI opportunities. Avoid POCs (bad rap in AI); get to success fast on real data.
- Define success criteria upfront (encourage hardest ones).
- Fly on-site, partner closely.
- Take 50-100 page SOPs (60 steps, highlights/red boxes for data extraction/entry).
SOPs well-codified for BPO handoff, but gaps exist – forward deployment closes them quickly.
Convert to agent operating procedures: Notion-like doc with natural language steps + tools (long-context retrieval from complex docs, human reasoning over rules, write to systems of record, web agents for no-API systems like broker/carrier portals).
Do whatever it takes: API if available, web agents otherwise. Fastest to POC/production for many mission-critical use cases by going on-site.
Benefited from reasoning model scaling (document extraction > human reliability). Next:
- Best web agents (critical for insurance CRUD on admin panels, unlike restaurant/flight booking demos).
- RL/reinforcement fine-tuning: Product structured for end-to-end agent decisions (not siloed blocks). Agents follow full procedures, highly gradable for reward functions per workflow/customer at scale. Avoid determinism; model perfect world for RL across tool calls.
Surprises:
- How much fun: build faster (Cursor, Claude, CodeX); codify what was hard in software/LMs (workflows that took decades now easy, offload to agents).
- Advice: Be a forward deployed engineer (best founder training – close to problems, production, building/selling). Do at Pace.
Learned at Retool (David Sher CEO, Brian Shrier): build post-sales motion, zero-to-100 products. Made starting Pace more tractable. Find backers who bet on you repeatedly.
Hot take from Sequoia internship: Sequoia never rests on laurels despite success (NY office screen: "we're only as good as our next investment"). Permeates Pace culture: reset to zero every quarter/new customer – only as good as next customer.
What doing differently this time (some same as Retool):
- Retool as a business: focus on customer value/revenue. Avoid distractions (fundraising, office, employers); spend max time with customers, moving key metrics like revenue growth.
Gross margins are one of the key things they think about a lot. Another big thing is who you work with. In his last company, he had an amazing co-founder. With Pace, he started as a solo founder and built an incredible team.
There's a sentiment that people don't like solo founders for investment, which makes sense at very early stages with existential risk, where a co-founder helps through challenges. Very thankful to his previous co-founder. But once direction is clear and it's all execution risk—as with Pace, where it's clear what they're doing and just execution—there are benefits to starting as a solo founder and building a team.
First engineer was Rachel's second engineer; got Yogi. Second hire was Luis, head of corporate engineering at Loom, also in the Sequoia portfolio. The broader team takes on roles traditionally for co-founders, without involving the team. Team involved in shaping values, critical go-to-market hires, product decisions. This lets him hire a much stronger team and build stronger culture.
Kept working with Sequoia and Ryan and Lauren, which has been fantastic. Amazing to work with a partner who's known him almost a decade. Met Brian in college.
Brian gave a talk on mentoring and building great mentor relationships: lead with value and give up front. He dropped that he was going to Princeton campus to talk with the president about entrepreneurship and asked what to say. Pulled an all-nighter and sent a 20-page report on where Princeton entrepreneurship could use help. That became the basis for much of what followed. Brian has seen every step of his career: from Zoya, starting first company, Retool, many board meetings where he learned a ton, to starting Pace.
Amazing partners like Brian, Ryan, Lauren create a strong board relationship that feels like a co-founder relationship, in the trenches together. This is incredibly unique and makes Pace a much stronger company.
Thank you for joining the show. It's been amazing to have you and hear the full story behind Pace that's just beginning.
Thank you for having me. This is a blast.
Keep Sequoia Capital in your library
Save the videos and channels worth coming back to, and find them again in one place.
