TechTalk: AI washing creates ‘massive expectation gap’ across insurance provision and liability

The race to innovate in insurance technology has supposedly been super-charged by the AI boom, with progression in this field consistently forming large parts of corporate strategy updates.

And while many initiatives have begun with fanfare, the measurable progress originating from these AI integrations has been somewhat harder to identify.

A recent report from Simplifai, released 1 April 2026, analysed data from McKinsey, EY, Deloitte and Swiss Re, revealing that while 99% of insurers now had generative AI initiatives in place, and 83% of carriers were spending more than $5m (£3.76m) a year on such initiatives, only 42% had deployed AI into active business functions.

Additionally, the returns on these initiatives were unconvincing from a return on investment perspective, with fewer than 15% of insurers reporting a measurable impact on combined ratio, cycle time or loss ratio from AI implementations.

As this report evidenced a failure to meet heady expectations, some in the industry drew parallels with greenwashing – a term coined to define a situation where a company overemphasises the environmentally beneficial impact of an offering to appear more favourably to external stakeholders.

One such commentator was Michael James, technology consulting director at Simplify Consulting, who told Insurance Times that the insurance sector now “risks doing something similar with AI”.

The critical difference to greenwashing, however, is that the pressure to look good is not coming from customers.

He said: “Our own research is telling us that customers are concerned about their insurance provider using AI. The pressure to use AI is coming from competitors, boards, investors and the fear of being left behind, which is why much of the current market noise is better described as AI gloss than AI strategy.”

Coined as AI washing, this new phenomenon refers to the misrepresentation or overstated efficacy of AI within a service or offering.

Claud Bilbao, vice president for underwriting and distribution at Cowbell, said much of the AI washing in the insurance market was simply “dressed up” rule-based automation that did not improve understanding of a client’s risk profile or provide proactive guidance.

Bilbao explained that the trend was becoming a “significant issue” to the point that regulators such as the FCA were “increasingly scrutinising false or exaggerated tech claims”.

And as more firms strive to incorporate AI into their systems, there is the risk that AI washing could distort how insurance is provided, managed and governed across the sector.

An ‘abstract compliance exercise’

A central issue underpinning AI washing is that the speed and scale of AI adoption has, in many cases, outpaced the development of governance frameworks.

This is according to Thomas Barrett, partner at law firm Weightmans, who told Insurance Times that the “risk is that businesses are accumulating latent liability because adoption has been prioritised over robust governance and oversight”.

Indeed, EY-Parthenon’s CEO Outlook Survey of global chief executives, which polled 51 leaders the UK financial services sector across March and April 2026, revealed that only 8% of UK respondents believed current AI frameworks provided clear guidance to support strategy and innovation – compared to 19% globally.

Compared to other technologies, Barrett explained that many AI systems are in fact being adopted with “relatively limited scrutiny” as “boards lack the technical expertise needed to interrogate them properly, or because supplier claims are accepted at face value”.

He continued: “That is particularly concerning given the potential for bias and opaque black box decision-making.

“In many respects, businesses are deploying AI before the governance, compliance and testing frameworks around it are sufficiently matured. Given the pace of adoption and limited independent testing of many systems, there is a real risk that some organisations are already accumulating liabilities that simply have not surfaced yet.”

For Somesh Mukherjee, vice president for innovation, data and AI at Acord Group, the “single greatest risk” firms face is treating AI governance as an “abstract compliance exercise”.

This is especially pertinent as the industry moves from “high-level ethics concerns” into an era in which “technical accountability will be at the forefront”, he added.

Mukherjee explained: “If you give an AI agent the autonomy to handle submissions, price risk or triage claims, you have to look at the operational plumbing.

“First, you need absolute model traceability. If a model drifts or its data lineage is muddy, you won’t survive a modern regulatory audit. Second, we have to eliminate the illusion that we can outsource this liability.

“You can buy the software, but you cannot buy your way out of the risk.”

Downstream risks

Insurers are also increasingly becoming exposed to downstream liability risks created by overhyped systems, especially as regulation struggles to keep pace with AI adoption.

“Helping clients understand that they own the liability for the AI tools they deploy is one of the most important conversations a broker can have right now.”

Speaking to Insurance Times, Bilbao said that the UK’s “sector-by-sector approach” to AI regulation was leaving SMEs exposed to the “massive expectation gap” between what AI promises and what it can deliver.

“If that (AI) system is poorly implemented, makes up facts or acts on biased data, then the business is left completely exposed,” he continued.

“From an insurance perspective, this expectation gap translates directly into claims. It quickly morphs from a tech problem into a professional indemnity (PI) issue, a directors and officers (D&O) claim for failing to govern operational risk or a regulatory fine for misrepresentation.”

Where the liability sits, he explained, is “worryingly murky” as there are not yet enough test cases – a scenario creating a complex overlap for brokers.

Bilbao added: “An incident might start as a technology failure, but it quickly spills over into PI exposure if bad advice was given, or a cyber claim if data privacy was breached.

“Helping clients understand that they own the liability for the AI tools they deploy is one of the most important conversations a broker can have right now.”

Leveraging responsibility

With many firms supposedly overstating their AI credentials, James stressed that the focus should be on applying and governing the technology responsibly to turn it into a “measurable operational advantage”.

For example, Zurich UK reinforced its commitment to ethical employment practices when adopting AI by formally signing the Community Union’s Fair Future for Finance Responsible AI Charter on 18 May 2026.

Believing that transparency remains critical in the adoption of AI, Asha Kumar, chief technology officer at Zurich UK, told Insurance Times that boards and senior leaders need a stronger understanding of AI-specific risks “to ensure they are embedded into governance and operating models”.

“Leadership must recognise that risk is not purely technical, but is also driven by how AI is used in real-world contexts,” she continued

“Strengthening oversight requires a balanced approach across technology, process and people, which is supported by training, structured risk assessments and governance forums that bring together business owners and risk experts.

”This ensures that as AI scales within a business, leadership retains confidence that both traditional and emerging risks are being managed effectively.”

Elizabeth Wooliston, chief of markets at Artificial, added: “At board level, digitalisation means staying involved beyond the initial approval

“The organisations that get value from digital change are typically very clear about what they’re trying to improve and they remain engaged as those initiatives move from design into delivery.”

‘Unglamorous’ opportunities

For Wooliston, some of the “most underused opportunities” for digitalisation “sit in relatively unglamorous areas”

This includes structured data, clearer decision frameworks and better integration between systems that already exist.

While she admitted that these areas “don’t attract headlines” they do “make an immediate difference to how consistently underwriting intent translates into outcomes”.

Echoing this view, James warned that “underwriting often attracts the loudest AI claims” but should also “attract the greatest caution”.

The strongest near-term AI opportunity for insurers lies in operational efficiency, he explained, in which “claims handling, service delivery, case triage, engineering productivity and customer acquisition all offer more immediate and measurable potential”.

He concluded: “This matters because regulatory attention in the UK remains firmly focused on consumer understanding, claims handling and service quality, which makes disciplined operational improvement more valuable than the broad AI theatre.”