AI is boosting performance across four insurance subsectors
Artificial intelligence is already moving the industry forward. A recent McKinsey & Company view estimates generative AI could create US$50-$70 billion in revenue for insurers, with the biggest lift in marketing and sales, customer operations, and software engineering. That value won't appear on its own-it comes from clear use cases and disciplined execution.
Source: McKinsey Insurance Insights
Brokers
Brokers are already using AI to raise conversion through automated submissions, appetite matching, and targeted cross-sell. Next up: agentic systems that handle renewals for simple risks with minimal human touch. Expect less swivel-chair work as quoting, forms, and carrier endorsements get streamlined.
- Stand up automated intake to clean and structure submissions before they hit producers.
- Use appetite matching to route risks to the right markets and cut no-quote waste.
- Deploy renewal bots for small commercial with human review on exceptions only.
- Instrument every step: track bind rate, time to quote, remarket rate, and add-on uptake.
- Consolidate first-party data across your book; larger, cleaner datasets will attract producers and carriers.
Managing General Agents (MGAs)
Premiums placed through MGAs in the U.S. grew about 14% annually over the past decade, nearly doubling from $47B in 2020 to $97B in 2024. MGAs are pushing data and tech forward, and AI is speeding up intake and risk scoring. Simple, lower-risk policies are already being underwritten and quoted with limited human input.
- Unify proprietary data across intake, underwriting notes, and loss history to build stronger risk signals.
- Deploy AI triage: segment risks by complexity, straight-through process the simple, escalate the complex.
- Add continuous learning loops: compare quoted vs. bound vs. loss outcomes to refine models.
- Strengthen APIs with carriers and brokers to shorten cycle time and improve bind ratios.
- Operationalize governance: version models, monitor drift, and document decisions for audits.
Software providers
Carriers are moving away from monolithic AI stacks to modular setups that match models to use cases. That means open standards, easy integrations, and the ability to plug tools into core systems without re-platforming-or getting locked into a single vendor.
- Offer a model-agnostic layer: support multiple foundation models, retrieval methods, and vector stores.
- Ship connectors for core policy, billing, and claims systems-and keep them lightweight.
- Provide privacy and compliance guardrails: PHI/PII redaction, audit logs, and data residency controls.
- Include an evaluation suite with transparent metrics by use case (e.g., FNOL notes, fraud flags, coding tasks).
- Make pricing predictable and modular so carriers can expand pilots that prove value.
Third-Party Administrators (TPAs)
TPAs sit on rich transaction-level data and can use AI to speed high-volume workflows and improve service. The catch: many contracts rely on headcount or activity-based pricing, so automation can compress revenue if pricing doesn't evolve. As automation narrows historical complexity advantages, cost discipline and continuous improvement matter even more.
- Shift pricing toward outcomes: per-claim tiers tied to cycle time, leakage, and customer satisfaction.
- Offer value-share models for improved recovery, subrogation, or fraud savings.
- Productize analytics: sell proactive insights (severity prediction, litigation risk) as add-ons.
- Automate first notice, document handling, and correspondence; keep humans on exceptions and negotiation.
- Renegotiate SLAs to reward speed, quality, and total cost of risk-not seats and clicks.
Where the near-term gains are
The quickest wins are in submission intake, renewal automation for small commercial, document understanding, and agent/adjuster co-pilots. Marketing and sales, customer operations, and software engineering show the strongest upside today. Start with one process, baseline metrics, then scale the playbook.
How to get moving (without stalling on strategy)
- Pick 3 use cases with clear KPIs: time-to-quote, bind rate, loss ratio lift, handling time, or NPS.
- Secure data foundations early: consent, retention rules, PHI/PII treatment, and lineage.
- Run 90-day pilots with weekly reviews, then expand based on measured results.
- Train frontline teams on prompts, SOPs, and exception handling; bake changes into workflows.
- Plan for model choice and change: avoid single-vendor lock-in and keep evaluation ongoing.
If you're upskilling teams for these use cases, you can explore practical programs here: Complete AI Training - Courses by Job.
Your membership also unlocks:
