Capacity for the work that needs people
Reduce avoidable administration and rework so experienced colleagues can focus on complex risks, claims and customer conversations.
Talk to Us
Human capability for insurers
Your teams assess risk, resolve claims and answer policyholders when it matters. AI should help them do that work well.
Build the judgment, practical skills and ability to improve workflows that turn AI investment into useful change.
Why AI Capability Matters
An underwriter has another incomplete submission. A claims handler has someone waiting for an explanation. Learning a tool, checking its output and changing routines must fit alongside the work your teams already carry.
Reduce avoidable administration and rework so experienced colleagues can focus on complex risks, claims and customer conversations.
Help teams check policy wording, source records and missing facts before relying on an AI-generated answer.
Protect policyholder information, recognize fairness concerns and make human accountability clear.
Connect AI use to turnaround, service quality and operating performance, with evidence behind investment decisions.
Responsible & Ethical AI
Approved tools and technical controls establish boundaries. People put them into practice with every submission, claim and policyholder conversation.
When the approved route does not meet the task, staff may improvise. Address shadow AI with clear information boundaries, usable alternatives and a way to ask for help.
Recognize sensitive customer records and claim details. Check that the tool is approved for the information and purpose before entering it. Practice when to stop, use synthetic material or request support.
Compare an answer with the applicable policy wording, endorsements and source evidence. Look for omitted facts, conflicting records and unsupported reasoning. A fluent explanation is not proof of accuracy.
Make the authorized decision owner and escalation route explicit. Staff need judgment to accept, revise, reject or escalate an output within insurer policies and delegated authority. Qualified teams remain responsible for model validation and fairness testing.
The NAIC model bulletin and New York underwriting/pricing guidance include training within broader governance expectations. Application depends on jurisdiction and use. Training does not establish compliance. NAIC · NYDFS
Adoption and implementation challenges
Choose a challenge. Explore the capability, relevant programs and progress your insurance teams can demonstrate.
Submissions, claim files and policyholder updates already fill the day. Learning AI needs to make that work easier, including the time spent checking it.[2]
Prepare submission summaries and claims briefs with clear instructions, evidence and review criteria.
Time to a usable file, checking time and rework. Decide how recovered capacity will serve your team or policyholders.
Which recurring task most needs to improve?
A loss summary can omit a material fact. A coverage explanation can miss an endorsement. Teams need to verify the applicable wording and evidence.[5]
Check sources, recognize uncertainty and fairness concerns, and escalate within delegated authority.
Source-check quality, omissions and corrections. Qualified teams validate models and monitor fair treatment.
Can your team explain why the answer is fit to use?
A claim needs a summary. The easiest tool is not approved. Unclear information boundaries or support routes leave staff guessing.[4]
Apply insurer policy to the tool, customer information, review and escalation.
Decisions in realistic tool-use scenarios, concerns resolved and time to get practical support.
Does the claims handler know where to get help?
A faster summary still leaves missing evidence, duplicate entry and an unclear owner. The policyholder experiences every handoff and delay.[7]
Test a repeatable claims-handoff brief with evidence, review criteria and an exception route.
End-to-end turnaround, avoidable handoffs, missing-information requests and repeat policyholder contact.
Where does the claim stall after AI finishes its part?
One underwriting team has confident users; another is unsure what is appropriate. A demonstration does not give every colleague a dependable method.[3]
Make context, instructions and source checks teachable and repeatable.
Can different colleagues produce acceptable work, explain their checks and improve through feedback?
What does your strongest user need to share?
More logins and faster drafts do not show whether claims need less rework or policyholders get clearer answers. One team’s gain can move work elsewhere.[1]
Set a baseline, name the owner and measure the insurance outcome.
Submission turnaround, claim cycle time, rework, repeat contact and redeployed capacity. Finance validates financial attribution.
What should improve for your policyholders or operation?
Organizational Benefits
Consider a commercial property claim. AI can help prepare and organize work. The policyholder benefits when facts, decisions and next steps move together.
Illustrative claims workflow
AI can assistOrganize the initial account of the loss.
AI can assistSurface missing fields and conflicting details.
AI can assistPrepare a source-linked file for review.
AI can assistDraft a clear update and next-step summary.
Measure the whole journey. Track turnaround alongside source-check quality, avoidable handoffs, rework and policyholder experience. Time saved is not automatically a cash saving.
Talk about your team's workflowAgency. Capability. Imagination.
Our framework connects ownership, reliable practice and the ability to improve work. People need all three to turn access into useful change.
“What do I own?”
A decision I can explainPeople need to see where AI fits in their role, what they control and when to seek review. They direct its use within insurer policies and delegated authority, staying responsible for the result.
Build ownership of AI use and its outcomes.“What must I check?”
Evidence behind the summaryAccess reaches the team. Reliable practice requires useful context, source verification, privacy awareness and shared quality standards.
Build skills people can demonstrate at work.“What could work better?”
A clearer path to resolutionA task gets faster. The larger opportunity is a smoother claim, a clearer submission or fewer repeat calls. Teams need room to rethink the whole process.
Build the ability to improve the work.Capability building complements validated models, suitable tools, data quality, integration and insurer-specific controls.
Published insurer experience
These examples pair AI with changes to work, controls and people. Their value is in the conditions behind the reported result.
Zurich North America / Underwriting
An initial rollout to 16 U.S. middle-market underwriters used AI to summarize submission material.
Build verification into the work.Company estimate from an initial rollout, not a controlled evaluation or a result for all Zurich submissions.
Read Zurich's accountAviva UK / Claims transformation
A broad claims transformation combined AI, workflow changes and more than 40,000 hours of staff training.
Redesign the journey and develop the people.McKinsey participated in delivery. The case does not isolate the effect of AI or training.
Read the Aviva caseAllianz Australia / Food-spoilage claims
Agent-based automation handled food-spoilage claims under AUD 500. Potential rejections went to experienced handlers.
Define the scope and the human exception route.Company-reported result for a narrow claims category, not claims processing overall.
Read Allianz's accountThese are insurer and implementation-partner reports, not humanskills.ai program outcomes.
Develop the people behind the implementation
We call this Directing Intelligence: deciding where AI can help, giving it purpose and context, evaluating its work and owning the result.
Match the learning to the responsibilities your underwriters, claims handlers, service colleagues and leaders carry.
Human-First AI Literacy and Human × AI Fluency
Understand AI, give useful context and evaluate its work. Build a shared language for appropriate use across underwriting, claims and service.
Compare an AI-generated submission summary with synthetic source records. Identify missing facts and explain what needs review.
Explore AI FoundationsPut responsible use into everyday practice
Practice judgment about sensitive information, evidence, fairness, disclosure and accountability within your insurer's governance framework.
Work through an unapproved-tool request or unsupported coverage explanation. Decide what to protect, verify or escalate.
Plan responsible AI learningMake useful practice repeatable
Develop reusable approaches to analyzing information, preparing work and improving a process with AI. Test and refine methods your colleagues can share.
Create a claims-handoff brief with required evidence, review criteria and an exception route. Examine how it could reduce avoidable rework.
Explore an AI Agility cohortVirgil, our Agentic Learning Guide, walks participants through exercises adapted to their work and goals. Use synthetic or approved nonsensitive material. The programs build transferable capability alongside insurer-specific policies, systems training and professional expertise.
Delivered through our higher education partners
humanskills.ai brings the programs and guided practice. Our higher education partners connect your organization with learning opportunities, private cohorts and enrollment arrangements.
Agree on protected learning time and practical support so participation fits the responsibilities your teams already carry.
Let's make it concrete
Start with one team and a priority that matters to your policyholders, people or operations.
The insurance evidence behind the page
Reviewed September 17, 2026. Written for U.S. carrier leaders, with property and casualty examples leading. International cases are labeled by market. Sources do not imply partnership or endorsement; surveys and company reports provide different kinds of evidence.
June 24, 2026. Global survey of 263 senior insurance executives across P&C and life/annuity. Self-reported findings, not a U.S.-only industry census.
April 4, 2025. Analysis includes a June 2024 survey of 200 U.S. insurance executives. Workforce readiness and business participation matter to scaling.
July 22, 2026. Analysis of changing work, expertise and accountability as AI adoption grows; not a measured training-program evaluation.
Adopted December 4, 2023. A model for state supervisory expectations, including governance, controls, training and vendor oversight. Application depends on the state.
July 11, 2024. New York guidance for AI and external consumer data in underwriting and pricing, including fairness, transparency and role-appropriate training.
July 9, 2025. Company account of an initial rollout to 16 underwriters. The reported 60-minute average saving per submission is an estimate.
Undated implementation case. Reports broad changes to technology, work and skills. McKinsey participated in delivery; results do not isolate AI or training effects.
March 18, 2026. Company-reported Australian food-spoilage claims example. Timing applies to claims under AUD 500, with human escalation of potential rejections.
Insurance scenes are AI-generated illustrations, not photographs of customers or the cited insurers.