Insurers rate their own AI progress highly, but few have rebuilt the processes that matter. In its study Unlocking AI value in insurance, published on 29 September 2026, KPMG found that 44% of insurance leaders place their organisation in the top quartile for AI transformation, while none has fully redesigned sales and distribution or underwriting around AI and only 11% report strong data foundations to scale AI beyond pilots. Seen from the contact centre, that gap between confidence and redesign is easy to recognise.
What KPMG found#
The study is KPMG International's, announced in a press release on 29 September 2026. According to the release, the research was conducted between 20 and 29 May 2026 with insurance leaders across 20 countries and six regions, at organisations with 500 or more employees, covering all major insurance sub-sectors. The release does not state the number of respondents, so we do not give one. It also draws on KPMG's separate Transforming the Enterprise 2026 research among senior insurance transformation leaders.
| Item | What the release states |
|---|---|
| Publisher | KPMG International, press release |
| Date | 29 September 2026 |
| Fieldwork | 20–29 May 2026 |
| Respondents | Insurance leaders, 20 countries, six regions, organisations with 500+ employees; respondent count not stated |
| Self-assessment | 44% place their organisation in the top quartile for AI transformation |
| Process redesign | None has fully redesigned sales and distribution or underwriting around AI; 3% have done so in policy servicing and claims |
| Data foundations | 11% have strong data foundations and governance to scale AI beyond pilots |
| Type | Leader survey, self-reported |
Source: KPMG press release, 29 September 2026. We have not read the full report; figures as published in the release.
The release adds context to those headlines. Most respondents, 71%, mainly use AI for content generation or routine task automation, and 11% say they have a very clear view of the return on their AI investment. KPMG's own commentary frames the finding as a gap between activity and transformation: insurers are busy with AI, but mostly to make existing work more efficient rather than to change how the business runs.
Why confidence runs ahead of redesign#
A firm can deploy many AI tools without changing a single end-to-end process. A model drafts letters, another summarises claims notes, a chatbot answers policy questions. Each looks like progress, and together they can justify a top-quartile self-assessment. But the process itself, from the customer's first contact to the business outcome, stays as it was. We read KPMG's near-zero figures for full redesign of distribution and underwriting as saying exactly that.
The data finding explains part of the reason. Scaling AI beyond pilots requires data that can be trusted, shared and acted on. Only 11% of respondents say they have that foundation. In customer contact the consequence is concrete: an automated conversation is only as good as the policy, payment and customer data it receives, and only as useful as the systems that receive its outcome.
What we see in insurance contact#
We work with insurers on several processes: contact during the life of the policy, renewals and cross-sell on warm bases, receivables monitoring, and booking medical benefits for insured persons. Each of them is an end-to-end process, not a tool, and each depends on the data and integration questions KPMG raises.
The policyholder between sale and renewal#
In our experience, our home market has little culture of staying in touch with the policyholder for the whole life of the policy. Many insurers wake up at renewal, often by e-mail alone. A redesigned process would treat the policy period as a relationship: reminders, confirmations and offers at moments that matter to the customer, not only to the sales calendar. That is a process change, not a technology purchase, and it is the kind of redesign KPMG finds rare.
Receivables monitoring#
Unpaid premiums are an early signal, and in our view a timely call changes the outcome more than a late letter. In soft collections for an insurer we measured a first contact closure rate of 61.9%: in that share of cases a concrete commitment was obtained in the first contact (historical, one process). That result depended on the process around the call: which policies were assigned, when, with what data, and what happened to the outcome afterwards.
Medical benefits for insured persons#
In medical insurance, the insured person asks for an appointment and someone has to find one. In this process Vee takes the request, then calls the partner clinics' receptions and IVRs until it finds a slot that matches the person's preferences. It is a clear example of AI completing a task across organisational boundaries, and it only works because the process was designed end to end.
From pilot to production is an operating question#
KPMG's phrase about scaling AI beyond pilots names the step where, in our experience, insurance projects most often stall. The step is rarely blocked by the model. It is blocked by the work around it: who owns the process on the insurer's side, who approves the content, which data flows are allowed under the data-processing agreement and impact assessment, and who acts on the outcomes each morning.
Our delivery path is built around that step. It starts with a diagnostic that includes listening to the insurer's own call recordings and, when it is true, saying that a process will not pay back. It continues through a three-part specification covering dialogue, integration and telecoms, testing, and operating rules agreed before launch. Then comes a controlled pilot on a sample, never the whole base, followed by managed production: intensive at first, then weekly, then monthly. Years later we still meet clients every week.
Scaling then happens the way KPMG's leaders would recognise: one process proven, then the next portfolio, the next line of business, sometimes the next country. It is slower than buying a platform for the whole organisation, and, in our view, far more likely to reach production. We think the redesign KPMG finds missing is best built this way, one end-to-end process at a time, with the data foundation improved along the way rather than demanded up front.
Where we would not stretch the findings#
The study does not measure the contact centre, and it does not say that redesigning customer contact will close the gap it describes. The largest economic question in insurance, the claims ratio, is affected only modestly by customer contact, and we do not promise otherwise. What contact can change is reach, timeliness and the share of cases that reach a clear outcome. Those are smaller than an underwriting transformation, but they are measurable within weeks rather than years.
Nor should the 44% be read as a sign of complacency. In our view, self-assessments in fast-moving fields often run ahead of evidence, and KPMG's own framing suggests leaders know it. The useful reading is that the next phase of value will come from redesigning processes, and that data foundations are the bottleneck.
What it means for insurers' customer contact#
- Start with one process, end to end. A renewal campaign or receivables monitoring for one line of business is a better first step than a tool rolled out across many.
- Fix the data path first. Agree what data the conversation receives, in what format and how often, and where its outcome is written back.
- Test on a sample. A controlled pilot on part of the base, with a comparator on the same portfolios, shows the incremental effect before scaling.
- Keep regulated wording approved. In regulated processes, replies come from a closed, versioned library, and every call can be reconstructed from its recording, logs, statuses and script version.
- Disclose the AI. Vee always introduces itself as AI, in line with Article 50 of the AI Act, which has applied since 2 August 2026.
Questions to ask a supplier#
- Which insurance process will you run end to end, and where does it start and finish?
- What policy and payment data do you need, and how is the outcome returned to our systems?
- Which comparator will show the effect, on which portfolios?
- What happens when a policyholder raises a claim, a complaint or a hardship during the call?
- Who listens to calls after launch, and how fast is an approved change live?
KPMG's study shows insurers moving quickly with AI and slowly with redesign. In customer contact, redesign can start small: one process, one portfolio, a clear outcome. That is how we work with insurers, and how we would suggest any insurer tests a supplier.
Follow The First Second
Occasional issues on the mechanics and economics of AI-powered customer contact. No noise, no vendor news.
Follow via RSS