Intact Insurance's AI Lead: Business Priorities, Not Data, Should Drive Transformation

Why Intact Insurance Is Reworking Its Data and AI Strategy

Luca Piccolo, data and ML products delivery lead at Intact Insurance, argues that the data function exists to serve business priorities — not to set its own agenda. In an interview with Insurance Business, he pointed to Intact UK's stated goals: improving broker experience, expanding broker distribution, sharpening underwriting and claims performance, and targeting a low-90s combined operating ratio. In his view, any data or AI initiative that cannot be traced to one of those strategic objectives should not survive scrutiny.

His sharpest criticism is reserved for the conventional transformation roadmap, which builds data foundations first, then reporting, dashboards and finally AI over several years. Piccolo says that sequence asks too much institutional patience: by the time value appears, leadership trust has often already eroded. He advocates a 'vertical stripe' model instead — choose a single domain, build every layer in it from foundations to AI, and demonstrate an end-to-end return before expanding to the next area.

Piccolo also separates mature AI from emerging AI. Fraud detection, pricing and actuarial models have already delivered measurable value, so the job is scaling and embedding them. Generative AI, by contrast, is promising but still sometimes 'a solution looking for a problem.' He points to causal AI — models designed to capture cause and effect rather than mere correlation — as a quieter development with significant implications for pricing, underwriting and claims.

For consumers, he warns that automation has not always struck the right balance, recalling an automated call centre journey that took an hour before a human agent resolved the issue in minutes. He also notes that insurers still do not know how consumer use of AI chatbots instead of search engines will reshape distribution and engagement.

Inside Piccolo's Case Against Layered AI Roadmaps

Why Intact's Data Function Starts in 'Follower Service Mode'

Piccolo's stance that business priorities should lead the data agenda is not abstract. Intact UK's published objectives — broker experience, broker distribution, underwriting and claims outperformance, and a low-90s combined operating ratio — become the filter for investment. Early in a transformation, he argues, the data team should operate in a responsive mode, solving business problems rather than imposing a pre-formed technical vision. That approach sacrifices initial proactivity but builds the credibility needed for more ambitious work later.

The Flaw Piccolo Sees in Horizontal AI Roadmaps

He rejects the layer-by-layer sequence of foundations, KPIs, dashboards and AI because it delays visible returns and erodes organisational goodwill. In his framing, many senior leaders do not fully understand data work, so a four-year roadmap asks them to fund years of preparation without proof. The vertical stripe model compresses the wait: it builds an entire data-to-AI stack inside one domain, producing a tangible internal success story that makes change management easier and gives the organisation something repeatable to point to.

Generative AI vs the Quiet Rise of Causal AI

Piccolo draws a practical distinction between correlation and causation. Most models in commercial use identify that A and B tend to occur together, not that A produces B. In insurance, where pricing and claims decisions carry financial consequences, conflating the two can lead to poor underwriting or customer outcomes. Causal AI embeds cause-and-effect logic directly, and Piccolo believes it is approaching an inflection point. Generative AI is treated more cautiously: useful in some cases, but the deeper task is to question whether an existing process should be automated or reinvented.

What Legacy Thinking Actually Costs Insurers

For Piccolo, the industry's greatest threat is not a single technology but the way transformation is managed. Resistance, he says, is usually not wilful obstruction; it reflects objectives and incentives that the change team has not understood. The remedy is co-creation, where business users are present, consulted and influential rather than handed a finished system. Adoption improves because the solution is done with the business, not to it — reducing the risk that a completed platform is quietly shelved.

How Insurance Data Teams Can Use the Vertical Stripe Model

For insurance data and technology leaders, Piccolo's framework points toward several practical shifts:

• Map every data and AI initiative to a named Intact-style strategic objective — broker experience, broker distribution, underwriting, claims or the low-90s combined ratio — and stop projects that cannot be traced to one of those goals.
• Pilot the vertical stripe model in a single claims or underwriting domain, building from data foundation to AI inside that domain and measuring an end-to-end return before scaling horizontally.
• Separate mature AI from emerging AI in planning: budget for scaling and embedding fraud detection, pricing and actuarial models; for generative AI, require a genuine process problem before approving a use case.
• Add causal AI to the evaluation pipeline for pricing and underwriting, where cause-and-effect modelling could matter more than correlation-based tools.
• Replace supporter-versus-detractor change management with structured motivation interviews; when a project stalls, identify the leadership pressure on the other side and redesign the initiative to serve both goals.
• Audit customer-facing automation with Piccolo's consumer test: if an automated journey takes substantially longer than a competent human agent, treat it as a process problem rather than an AI problem.