What’s the Future of Vertical SaaS in an AGI World? Jamie Cuffe, CEO of Pace
Jamie Cuffe, founder of Pace, discusses how AI agents are replacing traditional Business Process Outsourcers (BPOs) in the insurance industry, achieving 50-75% cost savings. He details his strategy of using forward-deployed engineers to build trust and ensure success in mission-critical workflows, transforming a low-margin services industry into a high-margin, AI-native business.
The next decade will see a monumental shift from outsourcing work to human-led Business Process Outsourcers (BPOs) to outsourcing to AI agents. At Pace, we are building the agentic process outsourcer for the insurance industry, focusing on automating the critical back-office operations that carriers have traditionally sent offshore.
Instead of merely augmenting outsourced services with software for a 10% improvement, the goal is to completely replace the vertical service provider. This transforms the fundamental economics of the industry, shifting from a 10% gross margin business with hundreds of thousands of people to an 80% gross margin business powered by hundreds of thousands of AI agents and a small team of insurance experts. The flywheel we aim to create is one where every product improvement directly increases gross margins.
Why Insurance is the Ideal Starting Point
The insurance industry is uniquely suited for this transformation because its lingua franca is emails and PDFs. The lack of common systems and APIs between the many intermediary players (brokers, carriers, reinsurers) has historically necessitated a massive human workforce for manual data entry and document handling. This exact challenge makes the industry perfectly positioned for AI agents, which can bridge these gaps and process unstructured data at scale.
Architecting for End-to-End Automation
A common pitfall is to shoehorn AI into small, predefined boxes within a traditional, deterministic workflow builder (e.g., a DAG). This approach is “short on AI” and misses the point. The very reason BPO work exists is that these processes, full of edge cases and requirements for human judgment, are too difficult to codify deterministically.
The correct approach is to build for AI agents that can take a full Standard Operating Procedure (SOP) and execute the process end-to-end.
- Human-in-the-Loop: For the vast majority of production workflows, there is no human in the loop. An internal operations team is used primarily during the onboarding phase to label examples and fine-tune the agent’s accuracy from a baseline of ~90% to the 99.5%+ required for mission-critical operations. This team then moves on to tackle the next most complex task.
- Technical Wishlist: The next frontiers for improvement are more robust web agents capable of handling complex, legacy insurance admin panels, and the application of Reinforcement Learning (RL). Our end-to-end agentic architecture is designed to facilitate RL, as the agent’s full input-to-output process is highly gradable, allowing for the creation of clear reward functions to continuously improve performance.
The Go-to-Market Playbook: Building Trust in a Conservative Industry
When working with massive, regulated companies on their core business operations, the product being sold is trust as much as it is technology. Our pilots have a 100% success rate in moving to production, driven by a set of core principles and a specific operational model.
The Power of Forward-Deployed Engineering
The key to success is “closing the distance” with the customer through a forward-deployed engineering (FDE) model.
- Profile: The ideal FDE is a former founder—an engineer who is commercially minded or a commercial person who is deeply technical. Their mandate is to do whatever it takes to make the customer successful.
- Function: FDEs fly onsite, learn the customer’s exact problems, help with prompt tuning, create clean evaluation sets, and handle systems integration. They don’t just file tickets; they have the ability to ship code directly to the product’s codebase, creating an incredibly tight feedback loop between the customer’s problem and the product’s evolution.
- Strategy: While the long-term goal is to empower customers to build workflows themselves in a software-driven motion, the FDE model is critical in the short and medium term to guarantee customer success and inform product development.
Targeting and Engagement Model
We begin by targeting workflows that are already being handled by a BPO. This provides several advantages:
- Codified Procedures: The process is already documented in an SOP.
- Built-in Evals: The customer already has a QA process to check the BPO’s accuracy.
- Easier Change Management: It is organizationally simpler to spin down a BPO contract than to retrain or reallocate an internal W2 workforce.
The process involves defining success criteria upfront, ingesting the customer’s 50-100 page SOPs, and translating them into “Agent Operating Procedures” in our platform. This allows an agent to execute the workflow using a suite of tools for long-context retrieval, human-like reasoning, and writing back to systems of record, either via API or through web agents.
Redefining Performance and Value
While engagements often start with cost savings—typically 50-75%—customers quickly realize other, more strategic benefits that were previously unattainable.
- Speed: In processes like submission intake, responding to a broker faster directly correlates with winning more business.
- Scalability: AI agents can handle massive, seasonal spikes in volume (e.g., hurricane claims, enrollment periods) 24/7, turning a potential backlog into a solved queue.
- Accuracy & Consistency: For highly complex tasks, like QA on a 300-page claims file against hundreds of rules, an AI agent can be far more accurate and consistent than a human, who is prone to fatigue and error.
- Observability: The “black box” of BPO work becomes a transparent, auditable software workflow, allowing customers to analyze bottlenecks and redesign their operations.
The Bigger Mission: A Constellation for the Services Market
The long-term vision, inspired by Constellation Software’s success in aggregating vertical-specific software companies, is to apply the same model to the services market. The BPO spend for the Banking, Financial Services, and Insurance (BFSI) sector alone is nearly $400 billion annually—roughly the same size as the entire global cloud software market. The opportunity is not just to aggregate these niche vertical services, but to fundamentally transform their economics with AI, building a durable, high-margin platform that spans multiple industries.