Once an insurer accepts that the research burden described earlier in this series is a problem of infrastructure, a fair question follows: should we build it ourselves?
The case for building is real. Large insurers and syndicates have data science teams. They know their own workflows, risk appetite and systems better than any supplier. AI tools are more capable and more accessible than ever. Building in-house promises control, a precise fit and freedom from dependence on a vendor.
For many kinds of software, those arguments win. Intelligence infrastructure is different in three ways that matter.
The foundation takes years to accumulate
Understanding how risk has developed around a location needs years of history: not a pile of news articles, but structured records with precise locations, consistent categories, dates and sources, connected to the places and infrastructure around them.
A team starting today can connect to current feeds. The history is harder. It has to be gathered, structured and checked, and much of the value of a long record comes from consistency over time. History accumulates at the speed of the calendar, and money shortens the process only so far.
The same applies to the wider context. A useful foundation sets events against hundreds of contextual and reference datasets, each of which has to be found, licensed where necessary, connected and kept current. That is permanent work, not a project with an end date.
The skills sit at a rare intersection
Building this well needs three kinds of expertise working closely together:
- engineering and data science, to structure unstructured material reliably, with known confidence and traceable sources
- geopolitical and security analysis, to know what matters, what is noise and how situations develop
- specialty insurance knowledge, to understand coverage, accumulation and how underwriters actually work
Each of these exists in the market. People and teams who combine all three are rare, and insurers compete for them with technology firms and specialist intelligence providers. A combined team takes time to form, and the knowledge it builds lives as much in its working practices as in any document.
Every engineer on intelligence pipelines is not working on something else
Insurers have heavy demands on their engineering capacity: policy administration, broker distribution, pricing, data and regulatory change. Intelligence infrastructure is not a one-off build. It needs continuous maintenance, new sources, recalibration and support.
The question is not whether an insurer could build it. Given enough time and money, a large one could. The question is whether that is the best use of scarce engineering and analytical talent, compared with the work that sets the insurer apart.
Deploying keeps the advantage where it belongs
An insurer's advantage lies in risk selection, underwriting judgement, client and broker relationships, and how it shapes its portfolio. The intelligence foundation supports all of these, but owning the foundation is rarely what wins business.
Many industries settled this question long ago. Banks subscribe to market data rather than collect prices themselves, and compete on what they do with it.
Deploying CORTEX does not mean giving up internal capability. It changes where that capability goes. The foundation provides structured, evidenced intelligence about places and events. The insurer's own teams build on top: their own views of risk, their underwriting guidelines, their integrations and their portfolio analysis. Their analysts start from an evidenced baseline and spend their time on judgement.
Deploying still takes work and should be planned with care. Integration with existing systems needs budget. Dependence on a supplier needs managing in the normal way: clear terms, access to your own data and outputs, and a route out if needed. Those are familiar disciplines, and they cost far less than building and running the foundation yourself.
Four questions settle the decision
- Does our advantage come from owning an intelligence pipeline, or from how well our people use intelligence?
- How long can we wait for a home-grown foundation to reach useful depth, and what happens to our decisions in the meantime?
- Can we recruit and keep a team that combines engineering, geopolitical analysis and specialty insurance knowledge?
- Is this the best use of our engineering and capital, compared with the work that sets us apart?
For a small number of firms whose business is the intelligence itself, building may be right. For most specialty insurers, the answers point to deploying the foundation and investing in what they do with it.
The final article in this series looks at what changes when brokers and underwriters work from the same evidence.
First published in an earlier form in early 2026.