Do you agree that there is a growing insurer frustration over multi-year, £50m–£500m core system projects that frequently run over budget, often underperform and and lock insurers into rigid architectures? Not just core systems but many a claims transformation project has fallen into a similar trap of delivering more than just digital FNOL but less than hoped for.
It's not that big pictures don't work- where there is a commitent from the C-Suite and shareholders, clear project strategy (what problems are we solving and why?) proper implementation resources and the partners to work with a committed workforce to achive thhe measurable success sought for insurers and customers.
It's rather that there is reaisation insurers can still think big to anticipate unmet needs potential and improve customer satisfaction and retention whilst tackling the compelling issues faster with smaller projects. These cumulatively can still add up to big picture transformation whilst delvering proff of value faster.
In October EY and Instanda announced a new partnership that prompted the title question in my mind. Particularly as the partnership announcements discussed the pair's abilitiy to address both transformation, unlock the full potential of AI and support insurers' people deliver value across the business (See sources). EY has a good track record delivering major projects whilst Instanda positions itself as delivering proof of value in short 12-week periods and with full success production deployment rapidly.I'll come back to this later and wanted to contrast this with an article published by BCG during October.BCG discuss the big topic of AI as do EY and Instanda.
BCG, see sources, published an article argueing for a unified agentic core to leverage AI, LLMs and agents which raised the immediare question in my mind “is this another expensive multi-year program?".
I believe an analysis of the BCG article raises more questions than answers before moving onto the enticing idea that EY and Instanda might offer a new business and commercial model that is viable.
BCG describes three archetypes for bringing agentic AI into the core: keep the existing core and add an agentic platform; a "skinny core" with orchestration, rules and decisioning moved to an agentic platform; or a unified agentic core, where deterministic and agentic capabilities are designed as one AI-native stack. It argues that only the third solves the underlying integration problem, advises against waiting for vendors such as Guidewire and Duck Creek, and says the other decision leaders cannot outsource is building the data ontology. It concludes that the unified core is the best long-term option for most insurers.
What is strong
- The ontology section is the best part. The point that agents need a shared, governed, machine-readable vocabulary, or meaning is lost at every handoff, is correct and practical.
- The orchestration model is sensible. An orchestrator, bounded specialist agents and subagents, tied to a process blueprint that separates fixed-rule stages from agent autonomy, is a clean way to think about deterministic versus probabilistic work.
- It is candid about cost. It states plainly that the unified core carries the highest effort, risk and engineering demand.
Where it is weak
1. It does not engage with appetite, the binding constraint. BCG does not ignore risk: it concedes the longer path and higher risk. But it treats risk as a price worth paying and never asks whether the buyer will pay it. For Tier 1 and Tier 2 carriers and global brokers with multiple policy and claims platforms, years of sunk cost in packaged suites and board-level memory of failed programmes, appetite is the binding constraint. An architecture recommendation that is correct on paper but unbuyable in practice is incomplete advice. The precise criticism is that BCG discounts appetite, not that it ignores it.
2. The central claim is asserted, not demonstrated. "Only one of three archetypes solves the integration problem" is the headline, but the article gives no case, benchmark or reference implementation. It also does not eliminate integration: the unified core is still "integrated with the existing data platform", so the problem is relocated rather than removed.
3. There is no transition path. The hardest question for a carrier running several legacy platforms is how it gets from here to a unified core without a big-bang cutover. The article is silent on coexistence, data migration, run-off of legacy books and parallel running. That silence is precisely where Tier 1 reluctance lives.
4. The alternatives are partly straw men. The skinny-core option, which shifts workflow, rules, decisioning and document handling to an agentic layer and keeps ledger, reinsurance accounting, regulatory reporting and audit in the core, is close to what many carriers are actually doing. Dismissing it as a temporary productivity boost that adds layers is a claim, not an analysis.
5. The case against waiting for vendors is thin. Three bullets (rollout may be slow, early builders gain advantage, vendor platforms lack your ontology) are assertions. The third is actually an argument for owning the ontology, which applies equally to a wrap strategy and does not require replacing the core.
6. Unquantified benefits, unattributed numbers. The claims of underwriting cycles falling from 10 to 20 days to 1 to 2 days and straight-through processing of 70 to 90 percent are "based on BCG experience". More importantly, nothing links those gains to the choice of core archetype. They could be achieved with an agentic layer over an existing core. The "lower run costs" and "faster iteration" benefits are likewise not quantified.
7. A single substrate has downsides the article does not discuss. One configuration point and one substrate for deterministic and probabilistic components concentrates change risk, blast radius and vendor or build dependency. Regulators focused on operational resilience (for example DORA in the EU and the PRA and FCA regime in the UK) will ask hard questions about that concentration and about auditability of agent actions. The article treats governance as a feature of the architecture rather than a risk to be evidenced.
8. The urgency is rhetorical. Telling leaders the window is open for "months, not years" while recommending the slowest, riskiest path is internally awkward, and no evidence is offered that the window closes.
9. Incentives. The authors include BCG Platinion, BCG's technology architecture and engineering arm. The recommended path is the one that needs the most architecture and engineering support, and the article says the ontology cannot be outsourced to vendors. That does not make the argument wrong, but a reader should weigh it as advice from a party that benefits from the larger programme.
10. It ignores the base rate. Roland Berger, in an August 2026 study, reports programme costs typically between EUR 20 million and more than EUR 100 million over two to five years, with around half of core replacement programmes missing time or budget targets. BCG's recommendation needs to explain why a still more ambitious build beats that base rate.
What the market is doing instead
Recent partnership and regulatory evidence points to bounded, non-exclusive deals rather than whole-business transformation.
- Vendors cannot staff a Tier 1 replacement alone, so integrators step in. EY announced an alliance with INSTANDA in October 2026 to pair consulting and delivery capacity with a modern platform. EY was also named a Leader in ISG's 2026 Guidewire services study, so it stays connected to the incumbent cores (EY on ISG).
- Partnerships are layered and case by case. INSTANDA joined ServiceNow's partner programme as a Build partner in February 2026 (INSTANDA), and ServiceNow already integrates with Guidewire InsuranceSuite on Guidewire Cloud for claims, policy servicing and underwriting (ServiceNow). Core vendors, workflow platforms and integrators are combining as needed, not converging on one stack.
- The control layer is contested. ServiceNow positions itself as an AI control tower. If orchestration and governance sit in a horizontal platform above the cores, the core becomes a system of record, the reverse of BCG's single substrate.
- Regulation rewards evidence, not architecture. The FCA's February 2026 insurance priorities extended oversight to delegated authority, remuneration and the use of AI in underwriting and claims, and its MGA and coverholder review is due to report in early 2027 (Insurance Business). Nothing there requires a unified stack; firms must show how oversight and data drive decisions. An evidence layer across existing systems answers that faster than a core replacement.
Will this EY Instanda agreement lead to a new business and commercial model?
Like you I only have the published articles and announcements to interpret. On the other hand there is a demand for shorter transformation projects that deliver value faster and Instanda sets its stall out to deliver that as do other modern coretech vendors like Genasys and ICE. The difference could be the EY partnership. Delivering POV projects in a 3 month window and being able to move onto the next priority issue in the following quarter would be a compelling rolling solution. Cumulatorily delivering constant innovation and improvement without the eye-watering costs of traditional projects. Both Instanda and EY have the hands-on imlementation skills needed to leverage AI and both have ample reference use cases of successful implementation.
Proof is in the eating so I will keep a tight eye on developments as they are announced and whether this heralds a new planning, proving and deploying lifecycle model with viable commercials for mid-sized carriers, MGAs and those large carriers wanting to introduce new products and enter new markets faster.
Sources
- BCG, The Case for the Unified Agentic Core in Insurance, 21 October 2026.
- EY, EY announces alliance with INSTANDA to help insurers modernize core architecture for the AI era, 14 October 2026.
- EY, EY named a Leader in the 2026 ISG Provider Lens Guidewire services study, 15 October 2026.
- Roland Berger, How insurers can maximize value from core insurance system transformation, 17 August 2026.
- INSTANDA, Integrated solution with ServiceNow, February 2026.
- ServiceNow, ServiceNow and Guidewire collaborate to improve insurance experiences, 2023.
- Sabine VanderLinden The Age of Agency: Why EY & Microsoft during Ignite 2026 Ask a Different Question

/Passle/5936d4a43d94760a145c92a7/SearchServiceImages/2026-11-12-17-09-38-539-6af5f2d2f52a2c5f2e030365.jpg)
/Passle/5936d4a43d94760a145c92a7/SearchServiceImages/2026-10-07-15-47-10-738-6ac6697ea94b9e84e53c38fc.jpg)
/Passle/5936d4a43d94760a145c92a7/SearchServiceImages/2026-10-05-10-49-57-927-6ac380d5147a7b8e1ace985b.jpg)
/Passle/5936d4a43d94760a145c92a7/SearchServiceImages/2026-09-23-14-13-13-321-6ab3de7902670829c849ef1a.jpg)