BFSI's AI Proof Gap: Big Returns, Bigger Governance Risks

The Revenue Numbers Are Real — And So Is the Exposure

Three major surveys landed within weeks of each other in April and May 2026, and taken together they tell an unusually honest story about where the BFSI sector actually stands with AI. Grant Thornton surveyed 950 C-suite and senior business leaders across 10 industries, including an insurance-specific subgroup of 100 executives. AM Best published a separate study of 152 rated carriers and managing general agents. Insurity polled over 1,000 US consumers on their attitudes toward AI in property and casualty insurance. The picture that emerges is decidedly not the triumphant "AI transformation" narrative common in conference keynotes — it's something more complicated, and more useful.

Start with what's working. Grant Thornton's insurance subgroup found that 52% of insurance executives say AI has driven revenue growth, 62% report improved decision-making insights, and half say it has reduced costs. In a sector where margins are structurally thin and regulatory overhead is high, a majority of firms pointing to genuine commercial returns from AI is significant. The direction of travel is clear.

Then comes the qualifier buried in the same report's governance findings: only 24% of insurance executives are very confident they could pass an independent AI governance review within 90 days. Nearly four in ten — 44% — say governance or compliance challenges have directly contributed to AI project failure or underperformance. Revenue gains and governance fragility, coexisting inside the same organisations. That tension is the defining story of AI in BFSI right now, and it deserves more attention than it typically gets.

The Adoption Curve Doesn't Guarantee Readiness

AM Best's April 2026 survey found that nearly 60% of insurance respondents expect AI to significantly transform their business models within one to three years — yet only about 20% consider their organisations at an advanced stage of AI implementation. Among the rest, 53% described their approach as cautiously keeping pace with the industry, and 27% said they aim to be "successful followers" learning from peers. A sector where 80% of players are deliberately waiting for others to move first creates a narrow window for those willing to commit early.

The barriers are familiar but structural. AM Best identified data readiness (45%), security and privacy (43%), and legacy system integration (41%) as the top implementation challenges — and these aren't three separate problems so much as different faces of the same underlying issue: whether enterprise data environments are mature enough to support AI safely at scale. A BFSI firm with transaction data siloed across a core banking system, a CRM, a claims platform, and a separate policy administration tool cannot train or deploy meaningful AI without first solving the data architecture beneath it. The tools are rarely the bottleneck. The infrastructure nearly always is.

Across the broader marketing data landscape, Salesforce's 2026 State of Marketing report found that 87% of marketers now use generative AI in at least one workflow, up from 51% in 2024. That near-universal adoption rate applies to marketing teams inside BFSI firms too. Adoption alone, though, does not produce returns. McKinsey's data shows that AI-driven personalization can unlock 10–20% revenue uplift in banking, and that those gains accrue specifically to firms that have moved from isolated tool use to systematic integration across content, campaigns, segmentation, and reporting simultaneously — a very different thing from running a few AI-assisted email sequences.

Consumer Trust Is Moving, But Only in One Direction

Insurity's 2026 AI in Insurance Report surfaced something that gets less attention than the enterprise adoption data: consumer attitudes toward AI in insurance have shifted sharply. Consumer support for AI in P&C insurance nearly doubled year-over-year, rising from 20% in 2025 to 39% in 2026. As 84% of consumers now use AI tools at least occasionally in their own lives — for writing, productivity, health queries, financial comparisons — they are genuinely less resistant to AI in the products and services they buy. That's a real shift, not noise.

The goodwill is conditional, though. Nearly half of respondents express distrust when AI is positioned as making autonomous decisions about claims approvals, fraud detection, or policy adjustments. Only one in three say they trust AI-driven insurance decisions outright, with another 26% still undecided. The distinction consumers are drawing is subtle but consequential: AI used to make services faster, smarter, and clearer earns trust; AI used to cut costs or automate decisions without explanation loses it fast. BFSI brands that lead their AI communications with operational efficiency benefits are pushing in exactly the wrong direction.

An insurer that leads its customer communications with "AI-powered claims processing" as a selling point may find the message lands very differently than intended. The more durable framing, backed directly by Insurity's data, is transparency about how AI is being used and what human oversight exists. That framing also happens to align with what regulators are increasingly demanding — which makes it both the principled and the pragmatic choice.

The Governance Gap Is a Marketing Problem Too

Grant Thornton's cross-industry findings put the governance gap in starker terms: 78% of business executives surveyed lack strong confidence that their organisation could pass an independent AI governance audit within 90 days. More than half — 51% — identified strategy as the biggest driver of AI ROI, yet only 22% of operations leaders reported having a fully developed and implemented AI strategy. The variable that most determines whether AI delivers returns is also the variable most organisations have not built yet. That's a difficult gap to talk around.

What separates firms capturing AI returns from those that aren't? The Grant Thornton data is specific: organisations with fully integrated AI are nearly four times more likely to report AI-driven revenue growth than those still piloting — 58% versus 15%. The gap between those two cohorts has little to do with the quality of the AI models they use. It comes down to the depth of operational integration and the governance structures that let firms deploy AI confidently into higher-value workflows, rather than keeping it locked in sandboxed experiments indefinitely.

For BFSI marketers, this matters in ways that go well beyond compliance checklists. A marketing team running AI-generated email campaigns, AI-scored lead segments, and AI-optimised media spend — without documented governance around model performance, data provenance, and output review — is carrying risk that compounds as usage scales. Regulatory scrutiny across banking and insurance alike is tightening. The National Association of Insurance Commissioners' model bulletin on AI has been adopted as a baseline standard in the majority of US states, and similar frameworks are emerging across Asia and Europe. Retrofitting governance onto a live AI stack costs substantially more — in time, money, and credibility — than building it in from the start.

Where the Practical Opportunity Sits

Despite the governance warnings, none of this data argues for slowing down. It argues for sequencing correctly.

First-party data infrastructure comes before any AI marketing initiative worth running. BFSI firms that invested early in building owned data assets — transaction behaviour, app interactions, policy history, customer service records — are seeing compounding returns now that competitors cannot close quickly. Banks still relying on third-party data or fragmented CRM records face a structural disadvantage that no AI tool purchase will fix.

Personalisation is the highest-ROI entry point for most BFSI marketing teams. McKinsey's estimates of 10–20% revenue uplift from AI-driven personalisation in banking are consistent with what is showing up in insurance retention data too, where AI-powered renewal campaigns have demonstrated meaningful reductions in policy lapse rates by identifying risk signals well in advance of renewal windows. The mechanism in both cases is the same: replacing batch, demographic-cohort campaigns with individual-level interventions timed to actual customer intent signals.

Agentic AI is the next frontier, but most BFSI marketing teams are not there yet — and pretending otherwise tends to produce expensive experiments rather than results. BCG's 2026 CMO survey found that 42% of marketing leaders still use generative AI only to assist humans with discrete tasks, and just 8% run campaigns where multiple agents operate autonomously. The teams pulling ahead are redesigning workflows around AI, not plugging AI into the workflows they already have. For BFSI specifically, where customer journeys span multiple touchpoints and drop-off points are expensive, agentic systems that optimise messaging, timing, and channel selection based on live intent signals represent a genuine step change in conversion efficiency — when the underlying data and governance are ready to support them.

The governance foundation is not optional, but it does not have to be built all at once. Grant Thornton's own case data shows that a focused effort to classify AI use cases by risk and complexity, establish monitoring standards, and document accountability chains gives firms the confidence to move faster. Governance built as a performance system — rather than a compliance exercise bolted on after the fact — is what separates the 52% of insurers reporting revenue growth from the 44% reporting project failure. Those two groups are making different choices, not facing different circumstances.

The Strategic Implication

The BFSI sector leads overall AI adoption globally, holding 19.6% market share of all enterprise AI investment. That lead does not mean most firms are capturing proportionate returns. The surveys published this year make clear that the gap between firms with AI in production and firms with AI integrated at scale is enormous — and that gap maps directly onto the difference between marginal productivity gains and genuine revenue impact.

Marketing teams inside BFSI firms, and agencies supporting them, are sitting at a specific inflection point. The tools exist. The data mostly exists. The budget intent is there — 86% of insurance organisations plan to increase AI spending in 2026 regardless of size. What's missing, in most cases, is the combination of strategic sequencing, data readiness, and governance structure that makes AI marketing produce outcomes rather than activity.

At Transformics, the BFSI briefs we find most revealing aren't from clients asking "should we do AI?" — they're from teams that have already spent on AI tools and are now asking why the returns aren't showing up. The answer is almost always the same: the tools outran the infrastructure. We work with BFSI clients to close that gap specifically — building first-party data pipelines, AI-assisted lead qualification, and compliant campaign infrastructure across paid, email, WhatsApp, and telecalling — so that the tools that touch revenue are actually connected to the foundations that let you scale them.