The previous chapters established what intelligence infrastructure is, why it matters, and what it enables. A natural question follows: Should we build this ourselves?
The question is reasonable. Large insurers have data science teams. They understand their own workflows. AI tools are increasingly accessible. Building internally offers control, customisation, and avoidance of vendor dependency.
These are legitimate considerations. But they require honest assessment of three structural barriers that make internal builds theoretically possible but practically disadvantaged: the intelligence foundation, the expertise intersection, and the opportunity cost.
This chapter examines each barrier directly, then considers what “buy” actually means in the context of intelligence infrastructure.
The Build Temptation
Before examining barriers, it’s worth acknowledging why building internally is appealing.
“We have data scientists.”
Large insurers have invested in data science capabilities. Teams exist. Tools are available. Analytical capacity isn’t zero. The question “why can’t our people do this?” is natural.
“We understand our workflows.”
No external provider understands a firm’s specific workflows, systems, and requirements as well as internal teams. Custom-built solutions can optimise for exactly how the firm operates rather than accepting generic approaches.
“AI tools are increasingly accessible.”
Large language models, machine learning frameworks, and analytical tools are more capable and accessible than ever. What required specialised teams five years ago can now be accomplished with smaller teams using better tools.
“We could own it completely.”
Internal builds avoid vendor dependency, pricing negotiations, feature roadmap uncertainty, and the risk of vendor failure or acquisition. The firm controls its own destiny.
These arguments have merit. For some capabilities, internal building is correct. The question is whether intelligence infrastructure is one of those capabilities—or whether the structural barriers make external adoption more practical despite the apparent appeal of building.
Barrier 1: The Intelligence Foundation
Intelligence infrastructure requires a foundation: validated historical data, structured for automated processing, linked to authoritative sources, continuously updated and refined. This foundation isn’t built overnight.
Historical depth requires time.
Understanding how risk has evolved at a location requires years of historical incident data. Not just “incidents occurred” but structured data: precise locations, accurate categorisation, severity assessment, contextual linkage, outcome tracking.
An insurer starting today might access current incident feeds. But the historical foundation—understanding what happened at Location 47 over the past decade, how patterns evolved, what preceded major events—requires either building that history over time or acquiring it from providers who have already done so.
Building historical depth is a multi-year project. Each year that passes adds one more year of history. The firms that started earlier have more history. This gap doesn’t close—it widens.
Expert validation requires networks.
Intelligence quality depends on expert validation. Not just algorithmic processing but human expertise confirming that categorisations are accurate, assessments are calibrated, and edge cases are handled appropriately.
Building an expert validation network requires identifying experts, establishing relationships, developing validation workflows, and iterating based on experience. This takes years, not months. The network must span geographies and domains—a terrorism expert in Southeast Asia validates differently than a political risk expert in Latin America.
Existing providers have established these networks over years. A new entrant starting today must build comparable networks from scratch while existing providers continue expanding and refining theirs.
Source relationships require trust.
Authoritative sources—government agencies, security services, on-the-ground observers—share information selectively. Relationships develop over time through demonstrated reliability, appropriate handling, and mutual value creation.
An insurer starting to build intelligence capabilities must establish these relationships independently. Sources that have worked with established providers for years won’t automatically extend the same access to new entrants. Trust develops over time through demonstrated performance.
The time-based moat.
These three elements—historical depth, expert networks, source relationships—create a time-based moat. Money can accelerate many things, but it cannot compress the time required to build historical depth, develop expert relationships, or establish source trust.
A firm deciding to build intelligence infrastructure internally today will require 2-3 years minimum to develop a foundation comparable to existing providers—and during those 2-3 years, existing providers continue advancing. The gap doesn’t close; it extends.
This isn’t a criticism of internal capabilities. It’s recognition that some assets require time to develop, and starting late means remaining behind.
Barrier 2: The Expertise Intersection
Intelligence infrastructure requires a specific intersection of expertise that is genuinely rare.
AI and machine learning expertise.
Processing unstructured intelligence into validated, structured outputs requires sophisticated AI capabilities: natural language processing, entity extraction, classification, temporal reasoning, uncertainty quantification. These capabilities require engineers who understand not just how to use AI tools but how to build reliable systems that operate at scale with appropriate confidence calibration.
This expertise is in high demand across industries. Technology companies, financial services, healthcare, and defence all compete for the same talent. Insurers compete against firms that can offer technology-focused cultures, equity compensation, and mission alignment with engineering priorities.
Geopolitical expertise.
Understanding what intelligence means—distinguishing significant incidents from noise, recognising escalation patterns, interpreting political dynamics—requires geopolitical expertise. This isn’t engineering expertise; it’s domain expertise in how political violence, terrorism, civil unrest, and related risks actually develop.
Geopolitical experts typically work in government, academia, consulting, or specialised intelligence firms. Few have chosen insurance careers. Attracting them requires demonstrating that insurance applications of their expertise are valuable and interesting—not obvious to experts who have built careers elsewhere.
Insurance domain expertise.
Understanding how intelligence applies to insurance—policy structures, coverage triggers, aggregation concerns, regulatory requirements, workflow integration—requires insurance domain expertise. What matters to an underwriter differs from what matters to an academic analyst. How intelligence connects to policy terms requires deep insurance knowledge.
This expertise is more available within insurers, but it rarely combines with AI engineering or geopolitical analysis. Insurance experts know insurance; they don’t typically know how to build AI systems or assess geopolitical risk.
The intersection.
The challenge isn’t finding AI expertise, geopolitical expertise, or insurance expertise individually. Each exists. The challenge is finding the intersection: people or teams who understand all three domains well enough to build systems that actually work for insurance applications.
This intersection is genuinely rare. It develops in firms that have worked at this intersection for years—building teams, learning through iteration, developing shared frameworks that bridge the domains.
An insurer building internally must either find this rare intersection in the labour market or develop it internally through cross-training and collaboration. Both paths require time and carry execution risk.
The competitive reality.
Firms that have built at this intersection have advantages that don’t transfer easily. The knowledge is embedded in systems, processes, and team dynamics—not in documentation that could be replicated.
Internal builds face this competitive reality: existing providers have developed the intersection over years. New builds must develop it while existing providers continue refining their approaches. The expertise gap, like the historical data gap, doesn’t close automatically.
Barrier 3: The Opportunity Cost
Even if time and expertise barriers could be overcome, internal building carries opportunity costs that deserve honest assessment.
Engineering talent allocation.
Every engineer working on intelligence infrastructure is not working on other priorities. Insurers have many demands on engineering capacity: core system modernisation, digital customer experience, regulatory compliance, data analytics, operational automation.
Building intelligence infrastructure requires substantial, sustained engineering investment—not a one-time project but ongoing development, maintenance, and enhancement. This investment competes with other priorities for the same limited engineering capacity.
The question: Is intelligence infrastructure the highest-value use of engineering investment? Or would the same investment in other areas generate greater returns while intelligence infrastructure is accessed externally?
Capital allocation.
Building requires capital: engineering salaries, infrastructure costs, data acquisition, expert networks, ongoing operations. Multi-year projects require multi-year commitments.
Insurers allocate capital to generate returns. Engineering investment in infrastructure generates returns through operational efficiency—but those same returns might be achievable through external adoption at lower capital cost.
The comparison: Build cost over 2-3 years versus subscription cost over the same period, adjusted for risk of build failure and time value of earlier access through external adoption.
Core competency focus.
Insurance firms excel at risk assessment, capital allocation, client relationships, and underwriting judgment. These are core competencies developed over decades.
Building AI infrastructure is a different competency. Technology firms excel at it; insurers have less experience and less cultural alignment. Attempting to build competencies outside core expertise carries execution risk and diverts attention from areas of genuine advantage.
The strategic question: Should insurance firms build technology infrastructure, or deploy technology infrastructure built by specialists?
Most industries have answered this question similarly. Banks don’t build their own market data infrastructure—they subscribe to Bloomberg. Retailers don’t build their own payment networks—they integrate with existing providers. Airlines don’t build their own reservation systems—they adopt industry platforms.
Insurance intelligence infrastructure may follow similar patterns: built by specialists, adopted by insurers, differentiation achieved through how infrastructure is used rather than whether it’s owned.
What “Buy” Actually Means
If internal building faces structural barriers, what does external adoption actually entail?
Not abandoning internal capability.
“Buy” doesn’t mean abandoning internal intelligence capability. It means using external infrastructure as a foundation while developing internal capability to apply that infrastructure effectively.
Internal teams focus on: customising infrastructure outputs for firm-specific workflows, developing proprietary analytical overlays, building integrations with internal systems, training users, and extracting maximum value from the infrastructure foundation.
The infrastructure provides the foundation. Internal teams provide the customisation, integration, and application that create firm-specific value.
Foundation plus differentiation.
External infrastructure provides the foundation layer: validated intelligence, workflow automation, proactive signalling. This foundation is common across adopters.
Differentiation happens on top of the foundation: How does the firm use intelligence in underwriting decisions? What proprietary risk views does the firm develop? How does intelligence inform client relationships and market positioning?
The best outcome combines external infrastructure efficiency with internal differentiation capability. Leverage the foundation that would require years to build. Differentiate through application that reflects firm-specific expertise and priorities.
Dependency management.
External adoption creates dependency, which requires management. Contractual protections against price exploitation. Service level agreements ensuring reliability. Exit provisions enabling transition if necessary. Evaluation of provider stability and strategic alignment.
These are normal vendor management considerations, similar to dependency on cloud providers, market data services, or other critical infrastructure. The dependency is real; it’s also manageable through standard practices.
Integration investment.
External infrastructure still requires internal investment: integration with existing systems, workflow redesign to leverage new capabilities, user training, change management. These investments are smaller than full internal building but not zero.
Firms should budget for integration investment when evaluating external adoption. The comparison isn’t “build cost versus subscription cost” but “build cost versus subscription cost plus integration investment.” The latter is typically still favourable, but honest assessment requires including all costs.
The Decision Framework
How should firms evaluate build versus buy for intelligence infrastructure?
If competitive advantage is in intelligence itself:
Some firms compete primarily on proprietary intelligence—unique data sources, distinctive analytical methodologies, insights unavailable elsewhere. For these firms, internal building may be appropriate despite the barriers.
The question: Does owning the intelligence infrastructure itself create competitive advantage? Or does competitive advantage come from how intelligence is used?
For most insurers, the answer is the latter. Underwriting judgment, client relationships, portfolio construction, and capital allocation create advantage—not ownership of underlying intelligence infrastructure.
If competitive advantage is in using intelligence:
Most insurers differentiate through application rather than ownership. Better underwriting judgment. Stronger client relationships. Smarter portfolio construction. More effective capital allocation.
For these firms, infrastructure is an enabler rather than a differentiator. External adoption provides the foundation more quickly and efficiently than internal building. Internal investment focuses on application—where the firm’s actual competitive advantage lies.
The honest assessment:
- Does the firm have 2-3 years to build while competitors who adopt externally operate at higher velocity?
- Can the firm attract and retain the rare intersection of AI, geopolitical, and insurance expertise?
- Is intelligence infrastructure the highest-value use of engineering and capital resources?
- Would competitive advantage come from owning infrastructure or from using it more effectively?
