Are You Multi-Cloud by Strategy or by Accident?

Let me start with a confession.

When I speak with P&C CXOs behind closed doors or in the quieter corners of industry events, the conversation sounds very different from what you hear on keynote stages. It is less polished, far more candid, and, frankly, more revealing.

What I hear, again and again, are variations of the same underlying concern. Leaders tell me they have invested in, or inherited, two or three cloud environments, yet innovation still feels slower than it should. They suspect duplication across data, pipelines, and costs, but lack the visibility to prove it. Some question whether they even have a cloud strategy at all, or just a distribution of vendors that looks like one.

There is also a growing disconnect between perception and reality. Boards often equate multi-cloud with resilience, while CIOs experience it as unresolved complexity. And while many organizations have brought in Google Cloud for AI, that capability often remains loosely connected, struggling to integrate meaningfully with existing COTS environments.

The pattern is clear. Most insurers today are not multi-cloud by strategy. They are multi-cloud by accumulation and far from having a true multi-cloud strategy in insurance.

How We Got Here: The Accumulation Problem

 In P&C insurance, this did not happen overnight. It unfolded gradually, almost organically.

AWS typically arrived first, powering early modernization efforts, data lakes, analytics, and initial claims use cases. Azure followed, often through the Microsoft ecosystem, Microsoft 365, Teams, and Dynamics, where collaboration tools quietly evolved into infrastructure decisions.

Core platforms reinforced the split. Guidewire Cloud Platform (GWCP) took root on AWS, while Duck Creek aligned with Azure, reflecting Microsoft’s underlying architecture. Then Google Cloud entered the picture, sometimes through a data science team drawn to BigQuery, sometimes through Vertex AI, and sometimes via an InsurTech partner already operating on GCP.

At no point did most organizations make a single, deliberate architectural decision to become multi-cloud. And yet, over time, that is exactly what they became, without a cohesive cloud architecture for insurers guiding those choices.

The consequences showed up quickly. Compliance teams grew cautious. Finance teams began questioning egress costs. Data engineering teams found themselves maintaining multiple pipeline patterns across environments. And all of this collided with the reality of how P&C insurance actually operates.

It runs on unstructured documents, submissions, loss runs, policy forms, ACORD files. It depends on multimodal data such as satellite imagery, drone footage, weather feeds, and telematics. It requires real-time inference for fraud detection, FNOL triage, and dynamic pricing. It leans on catastrophe models that demand massive bursts of parallel compute.

These are not the workloads AWS or Azure were originally optimized for. And somewhere in the middle of this already complex landscape, the AI agenda accelerated. And immediately hit a wall.

Because the hardest problem in insurance AI is not the model, it is the data. And that data is everywhere, fragmented across three clouds, multiple vendors, legacy on-premises systems, and shared platforms like Snowflake or Databricks that multiple teams claim ownership of, but no one fully governs.

The Question No One Puts on the Agenda

With Google Cloud Next just weeks away, CIOs will be inundated with sessions. The keynotes will inspire. The demos will impress. The insurance case studies will feel tangible and relevant. But there is a quieter question that every P&C CIO carries into that room, one that never appears on the official agenda: Are we running multiple clouds… or multiple strategies? And do we actually understand the difference?

Because they are not the same. Operating across multiple clouds without a clear framework creates the illusion of flexibility, while limiting the ability to act on it. You cannot shift workloads to a more cost-effective environment if the underlying data is locked into incompatible formats. You cannot build cross-functional AI agents if underwriting and claims systems operate under entirely different security and identity models.

What looks like optionality often turns out to be constraint.

Why AI Stalls Between Pilot and Production

This is where many CIOs find themselves stuck, chasing AI use cases that never scale and the reasons are rarely ambition, it is their architecture.

If your environment cannot support consistent data movement, unified governance, and interoperable security models, even the most promising AI initiatives remain trapped in pilot mode. And if you cannot clearly articulate how workloads, data, and decisions flow across your cloud ecosystem, then you do not have a multi-cloud strategy in insurance. You have a multi-cloud situation.

A Way Forward: The Purpose-Driven Cloud Model

To move forward, what is needed is not another vendor decision, but a design principle.

Based on my experience, I describe this as a Purpose-Driven Cloud Model. It is a way to bring intentionality to an otherwise organic sprawl, and to finally establish a coherent cloud architecture for insurers. It is structured across three distinct, but connected, layers.

The most important thing about this model is what it does not say. It does not say 'use one cloud.' It does not say 'use Google for everything.' It says: know why each cloud is there. Know what job it is doing. And make sure your System of Intelligence, where your AI lives, is architecturally coherent, not an afterthought.

Clouds Are Not a Strategy. Clarity Is.

The multi-cloud debate in insurance has long been framed around resilience (“do not put all your eggs in one basket”) and best-of-breed selection (“use the best tool for each job”). Both arguments are valid; however, neither is sufficient.

What is missing is a framework that defines the role each cloud plays within the broader architecture. The insurers that will lead over the next decade will not be those with the most cloud providers. They will be the ones with the clearest intent behind each one.

A System of Record that manages truth with discipline.
A System of Intelligence that concentrates data and enables AI with coherence.
A System of Action that executes decisions with speed, coordination, and auditability.

That is what a Purpose-Driven Cloud Model looks like. And that is the conversation I will be continuing this week at Google Cloud Next 2026.