When insurers should scale AI, focus on workflows and outcomes
AI scale in P&C hinges on workflow readiness, not model hype. The carriers that win build governance, exception handling, and human oversight into the process.
Underwriting desks, claims queues, and agency operations are where AI pilots in property and casualty either survive or become stranded experiments. The models are not usually the problem. The workflow is not ready, the data is messy, and no one has decided who owns the last mile when the system makes a recommendation that matters.
Readiness starts with outcomes, not novelty
James Thom, Vertafore’s chief product officer, says the clearest signal that a carrier is ready to scale AI is a shift away from technical novelty and toward business outcomes, expected impact, and process change. Insurance organizations often chase interesting problems instead of the painful ones: the renewal bottleneck, the endorsement backlog, the reconciliation grind, the overloaded inbox that slows service teams every day.
On June 22, 2026, Carrier Management focused on the harder question: whether a pilot was ready to become part of day-to-day business. That standard exposes whether the carrier has defined the workflow, cleaned the data, and built enough governance to let AI operate inside production processes rather than beside them.
The scale test is an operating model test
BCG’s 2026 property-and-casualty research argues that CEOs have to redesign core workflows such as underwriting and claims end to end, with autonomous AI agents working under human oversight. The heavy lift is not the model itself. Most of the redesign effort has to go into people and processes: upskilling, culture, change management, and governance.
That is the part many software buyers still underprice. They budget for the AI feature and forget the operational scaffolding around it. If underwriting still depends on scattered email threads, if claims handlers do not trust the system’s recommendations, or if exception handling lives in no one’s process map, the AI layer becomes decorative. Production scale demands clear handoffs, auditability, and a decision path for every case that does not fit the script.
BCG’s broader insurance strategy work also focuses use-case discipline. For a global commercial P&C carrier, the most valuable AI bet may be materially improving speed to quote in underwriting. For a retail P&C carrier, the priority may be fixing customer service that has lagged for years.
Where to start: the bottlenecks that actually move the business
- speed to quote in underwriting
- costs and satisfaction in claims
- renewals that require too much manual follow-up
- new business intake that burns time on repetitive data entry
- endorsements and reconciliation that trap skilled staff in low-value work
The strongest AI programs in P&C do not begin with vague promises about transformation. They start with one of a few concrete pressure points:
Those are the workflows where AI can compress cycle time and free experienced people for judgment calls. They are also the places where bad process design is easiest to see. If an insurer cannot explain how an AI agent escalates an exception, routes a task, or preserves a decision trail, the carrier is not ready to scale it.
Vertafore’s product direction in 2026 reflects that practical view. The company has emphasized AI embedded in real workflows, not isolated features bolted onto the side of a platform. On June 30, 2026, Vertafore announced new Velocity AI agents for AgencyOne, built into workflows and designed to keep professionals in control for review, approval, and exception handling.
Why embedded AI beats a standalone demo
The difference between embedded AI and a standalone feature is not cosmetic. Embedded AI can sit where the work already happens, which means less context switching, fewer duplicate steps, and a cleaner audit trail. Standalone tools often impress in a demo and fail in implementation because they leave the carrier with too many handoffs and too little accountability.
In its 2025 and 2026 messaging, Thom repeatedly pointed to renewals, new business, endorsements, reconciliation, and email handling as the work that matters most. Service teams burn hours there, errors multiply, and customer experience gets damaged by delay.
Vertafore’s Project Impact put a benchmark on that capacity: up to 45 minutes of time savings per day for service professionals. It only matters if the carrier can hold that time savings inside the live workflow rather than bleeding it away in rework and exception cleanup.
The questions software buyers should ask before they scale
- Is the underlying data clean enough to support automated recommendations?
- What happens when the AI is wrong, uncertain, or missing context?
- Which exceptions require human approval, and where does that approval live?
- Can the workflow produce an audit trail that compliance and leadership will trust?
- Does the AI integrate with core systems, or does it create another swivel-chair process?
- Who owns the decision when AI is embedded in underwriting or claims?
The right vendor conversation should feel less like a product tour and more like an operational review. Before a carrier scales AI, the buyer should be able to answer a short list of hard questions:
They expose whether the carrier has the governance discipline BCG identifies as essential, including human oversight, upskilling, and process redesign.
This article was produced by Prism’s automated news system from verified source data, official records, and press releases, then run through automated quality and moderation checks before publishing. The system is built and supervised by the people who set the standards it runs under. Read our full AI policy.
Did this article answer your question?
