Build vs buy claims AI gets argued on licence cost, and licence cost is the one number that stops mattering the moment you want out. The question that decides the outcome is who holds the evidence record behind each claim decision, in what format, and for how long after the contract ends.
The prompt for this is an Insurance Journal piece published on 21 September 2026, covering an Insurance Nerds webinar that made the case for insurers building their own software instead of renting it. Their diagnosis is correct. The disagreement is about where control is won: in the contract, not in the architecture diagram.
We covered the components-versus-platform version of this question in open-source fraud detection versus commercial platforms. This one is narrower: it is about an artifact rather than an architecture.
The webinar is right about the problem
Answer
Why are insurers talking about building their own software again?
Because run costs crowd out everything else. A 2025 West Monroe survey of 300 US insurance executives found more than half spend 51% to 75% of IT budget just keeping existing systems running, and 94% had at least one strategic technology program delayed or cancelled for budget reasons in the past year.
Those figures come from a West Monroe survey of 300 US insurance executives. When maintenance eats that much of the line, the next renewal notice does not read as a price. It reads as a claim on the only money left.
Nick Lamparelli, who runs Insurance Nerds, offered the synthesis: "now, it can be buy and build, not buy or build."
Combined Ratio Solutions, co-founded by panelist Luke Magnan, publishes CRS OSPolicy, an open-source policy administration system with "no upfront licensing fees," and sells the implementation, hosting and support around it. The licence is free and the work around it is the business. That is a legitimate model, and it means the panel had a direct commercial interest in the conclusion it reached. Universal Shield, the carrier on the panel, is the vendor's customer.
The scope difference matters more. CRS OSPolicy is a policy administration system, and it writes structured records into a database the carrier already owns. A claims AI system produces something else, and where that something else lives is the whole question.
What the build case costs when you price it honestly
Answer
How often do insurance core system builds and AI projects fail?
BCG puts failure for large-scale transformations at 74% and prices core IT modernisation at 2% to 4% of non-life gross written premium in capex. S&P Global Market Intelligence found 42% of companies abandoned most of their AI initiatives, up from 17% the year before, with 46% of proofs-of-concept scrapped before production.
BCG published both numbers on 12 January 2026. The 2% to 4% of non-life gross written premium figure converts a build from a line item into a capital programme with a board sponsor and a multi-year burn. The 74% failure rate is the base rate the build case has to beat.
CIO Dive reported S&P Global Market Intelligence's survey of more than 1,000 respondents in North America and Europe on 14 March 2025. Its figures are abandonment rates for projects that had already been funded.
Then there is the unit cost of the team. The US Bureau of Labor Statistics puts the median annual wage for data scientists at $120,230 as of May 2025, with employment projected to grow 35% from 2025 to 2035. A build is a standing hiring commitment in a tightening market, not a one-time delivery.
The build case starts against a 74% failure base rate
Published base rates for the programme, for the AI projects inside it, and for the team that has to keep running it.
A build is a capital programme and a standing hiring commitment, not a one-time delivery.
The build case has real wins. Building is the right answer when the capability is a genuine differentiator you intend to maintain for a decade, when you already run a data science function, and when the artifact is a record you would own anyway in your own database.
Universal Shield's policy admin decision sits squarely inside those conditions. Claims AI usually does not.
Why licence cost is the wrong thing to compare
Answer
Why is licence cost the wrong way to decide build vs buy for claims AI?
Because a claims AI system produces a regulated artifact. A policy administration system writes records into a database you already own. A claims investigation engine produces a reasoning trail that can exist only inside the vendor's platform, and that trail is part of the claim file a regulator can ask you to reconstruct.
Retention rules are where that becomes concrete. Rhode Island regulation 230-RICR-20-60-4 states at section 4.6(A) that "a claim file and accompanying records shall be maintained for the calendar year in which the claim is closed plus four (4) years." Investigation records and correspondence are on its list of what a property and casualty claim file has to contain.
The arithmetic, stated as arithmetic
This is a derivation from two published inputs, not a regulator's finding. A Rhode Island claim closed in 2026 must be producible through the end of 2030. A three-year SaaS term signed in 2026 expires in 2029. The retention obligation outlives the contract by a year in that example, and by more in states with longer floors. Check your own states before signing a term.
Hesper runs 15+ investigation phases in parallel on a claim, and every phase leaves evidence behind: what was pulled, from where, when, and what came back clean. What that record has to contain is worked through in what the claim-file rules asked for before AI. This post is the other half of the same question: who gets to keep it.
The retention duty outlives the contract that created the record
Rhode Island's closed-file rule against a three-year subscription, both starting from a claim closed in 2026.
- Claim closed2026A Rhode Island property and casualty claim, closed and filed
- Contract ends2029Three-year SaaS term signed in 2026
- Retention ends2030Closing year plus four years, accompanying records included
Build vs buy claims AI, scored on the dimensions that bite
Answer
Is building or buying claims AI better?
Scored on exit rather than on price, the three options rank differently. Building wins custody of the record and avoids contractual lock-in. Buying wins time to first production decision, maintenance ownership and model refresh. Hybrid, meaning buy the engine and own the record by contract, wins where the artifact outlives the arrangement.
Lamparelli's "buy and build, not buy or build" is the column that wins most often for claims, and the reason is in row two rather than row one. You are not building the investigation engine. You are building, or negotiating, the custody arrangement around it.
The cost categories that make a full build expensive are itemised in the hidden integration costs of legacy claims AI, and none of them are the licence.
One honest note on row eight: some lock-in is structural and no clause removes it. The industry's contributory cross-carrier claims database gets its value from being shared. You cannot port a network out of a network.
Lock-in is almost never a clause saying you cannot leave. It is the four clauses nobody wrote: export format, export window, custody of the audit trail, and what survives termination.
The exit and portability clauses to negotiate before you sign
Answer
What has to be in the contract for you to still hold the record in 2030?
Nine clauses cover it: ownership of the evidence record, export format and completeness, export SLA and API access during the term, transition assistance, survival of access after termination, rights to models trained on your data, deletion and certified return, audit rights and regulatory cooperation, and sub-processors plus escrow.
Every clause below costs a vendor close to nothing to agree before signature and is close to unobtainable afterwards. That asymmetry is the entire negotiating position.
Clause four has a drafting benchmark, and it is the only place these terms are written into law anywhere. The European Commission's Data Act fact page and Regulation (EU) 2023/2854 set, in Chapter VI:
- A mandatory maximum transitional period of 30 calendar days to complete a switch, extendable to a maximum of seven months where it is technically unfeasible
- A minimum 30-day window to retrieve data afterwards
- Erasure of exportable data once retrieved
- The removal of all switching charges, including data egress fees, from 12 January 2027
A US-only buyer gets none of this by operation of law and all of it by asking.
Who runs the review that tests all nine clauses is a separate job, mapped in the procurement and vendor-risk review for AI claims investigation.
What US regulators have and have not said about your vendor's data
Answer
Does any US insurance regulator give you a right to your vendor's data?
Not directly. US insurance regulators have written vendor-oversight and audit-rights expectations and almost no data-portability right. The NAIC model bulletin tells insurers to put audit rights and a regulatory-cooperation duty into third-party contracts. The obligation runs to the insurer, not the vendor, and the contract is where it gets discharged.
The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers is the governing text, and the cleanest readable primary is a state adoption of it: the Wisconsin OCI bulletin of 18 March 2025.
It asks for "the inclusion of terms in contracts with third parties that: a) Provide audit rights and/or entitle the Insurer to receive audit reports by qualified auditing entities. b) Require the third party to cooperate with the Insurer with regard to regulatory inquiries and investigations related to the Insurer's use of the third-party's product or services." Lift that into clause eight.
NYDFS Insurance Circular Letter No. 7, issued 11 July 2024, says much the same on audit rights. It also states its own limit: "this Circular Letter also is not intended to address phases of the insurance product lifecycle other than underwriting and pricing." Useful as a drafting source, not as claims authority.
Put the two bodies of law side by side and the gap is the finding. The US tells you to supervise your vendor. The EU tells your vendor it has to let you leave. Neither one tells a US carrier how long it keeps reading its own claim reasoning after a subscription lapses. The buyer has to write that sentence into the order form, before signature.
The US says supervise your vendor. The EU says your vendor must let you leave.
Neither tells a US carrier how long it keeps reading its own claim reasoning after a subscription lapses.
The survival period on your own claim reasoning gets written by the buyer, into the order form, before signature.
Where Hesper lands on this
Answer
What is Hesper's position on data ownership and exit?
Hesper AI is an AI claims resolution platform: agents take a claim from first notice to final recovery with investigation-grade evidence behind every decision. On exit, the posture is that the evidence record belongs to the customer and is exportable, and that Hesper does not train models on customer data.
We would rather be portable than sticky, and the reason is commercial as much as principled: a record a carrier cannot take with it is a record its compliance function cannot rely on. Hesper integrates into the claims system of record - Guidewire, Duck Creek, Majesco and others - rather than replacing it.
Manual SIU investigation runs 14+ days per case against 200+ cases per investigator, and the file arrives in minutes, not weeks. Where each stage sits on the claim path is mapped in the guide to automating the claims lifecycle.
The honest way to test any vendor is to ask for the export schema and a sample file before you ask for a price. See how the platform works.
Key takeaways
- The Insurance Nerds panel's control argument is legitimate, and the panel had a direct commercial interest in it: Combined Ratio Solutions sells the implementation, hosting and support around its no-licence-fee platform.
- Building wins on custody of the record and loses on time, maintenance and failure rate. BCG puts large-scale core transformation failure at 74%.
- For claims AI the deciding variable is who holds the evidence record. A Rhode Island claim closed in 2026 must be producible through the end of 2030, while a three-year contract signed in 2026 ends in 2029.
- US insurance regulators have written vendor-oversight and audit-rights expectations but almost no data-portability right, so portability has to be negotiated clause by clause.
- The negotiable version of the build-your-own argument is nine contract clauses, agreed before signature.