Contact center AI for insurers: which tool for which job

Which tool for which job?

Your job	Documented example	The reference to read	What it documents
QA on far more calls than manual sampling	Observe.AI	Trupanion	Trupanion moved from about three QA evaluations per agent per month to Auto QA across member experience, sales and retention teams, with one dedicated QA lead. 2
Automated QA scorecards and coaching time	Observe.AI	American National Insurance	American National Insurance automated QA across more than one million calls with tailored scorecards, and piloted real-time agent assistance for reinstatement. 5
Speech analytics for compliance and service quality	CallMiner	VitalityHealth	VitalityHealth analyses all calls with CallMiner Eureka, run as a managed service by Davies Consulting, and scores them on regulatory, service excellence and process assurance. 6
Workforce scheduling for a support team	Assembled	Marshmallow	Marshmallow rebuilt scheduling in Assembled: schedule automation went from 0% to 95%, and two planners now cover roughly 370 agents. 7
Making offshore agents easier to understand	Sanas	Staysure	Staysure piloted Sanas accent translation with nine Sri Lankan agents on medical and renewal calls, tracking CSAT, NPS and positive communication mentions. 4

What do insurers actually use contact center AI for?

The documented examples fall into four jobs: reviewing calls for quality and coaching, automating caller verification, scheduling staff, and making offshore agents easier to understand. One insurer, Arbella, replaced its whole contact center stack. Every example below is a vendor-published story or a vendor customer page, so read them as workflows to investigate, not as a ranking.

As of 2026-11-14, usedby lists 27 Insurance & InsurTech companies with a public contact center AI line: 13 CONFIRMED and 14 OBSERVED. The count shows how many have a public page, not which tool is better. Several lines name a customer but no task.

189

Which insurers use AI to review every call and coach agents?

Trupanion, a pet insurer, disbanded its QA team in 2019 after reviewing only about three calls per agent per month. It now runs Observe.AI Auto QA and Moments across member experience, sales and retention, with a dedicated QA lead who also does manual reviews. Trupanion reports a 5% retention increase and says its cancellation save rate rose from 20% to 25% in one cited period; it links Observe.AI to coaching targets, not to a proven cause.

American National Insurance had the same sampling limit, three calls per agent, and automated QA across more than one million calls with tailored scorecards. Supervisors moved from reviewing to coaching, and the team piloted real-time assistance for reinstatement calls. The source does not report the pilot's result.

VitalityHealth, a health insurer with more than 550 advisors and over a million calls a year, uses CallMiner Eureka through Davies Consulting's managed service. Every call is transcribed and checked against what advisors were expected to say, in three areas: regulatory, service excellence and process assurance, with results in the coaching system within 24 hours. usedby lists VitalityHealth at OBSERVED level, so treat this CallMiner-hosted story as a described workflow.

For a smaller insurer, the shared change is moving from sampling to full coverage. The criterion to test (an editorial suggestion, not a vendor claim) is whether automated scores match your own supervisors' scoring on the same calls. Observe.AI states inference-only use of data and regional storage in US, CA, EU and AU.

256101112

Can an AI agent handle caller verification for an insurer?

Clearwater Benefits, which administers health benefits, deployed Observe.AI's VoiceAI Agent, named Amy, to verify members and then providers in English and Spanish. Failed first attempts get a second pass before a human takes over, and escalations arrive with context. Clearwater reports 80%+ automated member verification, verification of about 90 seconds against more than two minutes before, and average handle time down from roughly eight to 5.5 minutes. Those are vendor-published figures for one workflow.

Arbella Insurance is the broad case: it replaced UCaaS, CCaaS and CRM in one rollout and moved to Talkdesk's insurance experience cloud with Autopilot, Copilot and QM Assist. The source names these products without a per-product task, so the platform bet cannot be separated from its parts. Talkdesk's listed price for its Industry Experience Clouds tier, which includes Insurance, is $225 per user per month.

91314

How do insurers schedule support staff with AI?

NEXT Insurance managed 110+ agents across internal and BPO teams in spreadsheets before choosing Assembled for forecasting, scheduling, monitoring and reporting. Most of the setup was done in a month, per the source. Adherence rose from 68% to 85%, which the source translates into $957K in savings; that is a vendor-published figure. NEXT runs Talkdesk, and the Assembled integration was a selection factor.

Marshmallow, a UK car insurtech, shows the catch: it already had Assembled but had not configured handle times, SLA targets or shrinkage, so the tool worked like a spreadsheet. After a rebuild, schedule automation went from 0% to 95%, and two planners now cover roughly 370 agents with a two-month horizon instead of 80 FTE and four weeks. Its contact platform is Intercom, which the planner cited as a reason against NICE and Genesys.

3715

Does accent translation help insurance sales, retention and sensitive calls?

Staysure, a UK travel insurer, ran a proof of concept from August 2025 with nine Sri Lankan agents at a BPO partner, after customers disengaged in medical conversations. Positive communication mentions rose from 12 to 39 and CSAT from 65-66 to 68. Sanas runs a locally running pre-trained model, which Staysure says helped its data protection review. The sample is small and the CSAT move modest.

Oncourse Home Solutions, a home warranty provider, ran a five-week controlled pilot at two BPO sites, in the Philippines and Colombia, against matched control groups and each agent's own baseline. It reports inbound sales conversion from 9.84% to 17.26% and a 4.3-point rise in plan save rate. The control design is the transferable part; the results belong to that pilot. Sanas lists AWS Marketplace prices of $72,000 per 12 months for 100 users (Accent Translation) and $18,000 (Speech Enhancement).

41617

Which US insurers are named, and at what proof level?

CONFIRMED with a described task: Clearwater Benefits (Observe.AI, verification), NEXT Insurance (Assembled, scheduling), Oncourse Home Solutions (Sanas, sales and retention pilot) and Arbella Insurance (Talkdesk, platform replacement). Northwestern Mutual is CONFIRMED on Sprinklr, but its record concerns a social media campaign, not a contact center task. OneMain Solutions is CONFIRMED on Glia with no task attached.

AmTrust (Balto) and New York Life (CallMiner) are OBSERVED only: the vendor shows them as customers, with no task attached. Balto's homepage also shows anonymous insurance-company results, which cannot be tied to AmTrust or any named insurer. All eight are US companies; this says nothing about preference in your market.

3913161819202122

What should a smaller insurer check before copying one of these?

Check the baseline first. Trupanion reviewed about three calls per agent; NEXT managed 110+ agents; Marshmallow planned 80 FTE. A team well below those volumes may not get the same payoff. Then check your telephony: Observe.AI, Balto and Sanas each list Genesys and Five9 integrations, and Observe.AI also lists Talkdesk and Amazon Connect. Pilot shape matters too, and Oncourse's controlled design is a useful model.

Remaining limits: no source compares tools head to head, none reports a legal or regulatory outcome, and VitalityHealth's regulatory scoring is a QA category, not a compliance result. Observe.AI prices by custom quote; Assembled pricing is not verified here.

61623242526