Majesco

The company describes itself as artificial intelligence native more insistently than any vendor graded in this index, and the removal test does not support the claim. What the business does is administer insurance: policy, billing, claims, underwriting, distribution and loss control across property and casualty, life, accident and health, and pension and retirement, with customers processing more than 100 billion dollars of direct written premium through it.

Take the models out and every dollar of that still processes. The company's own phrasing is accurate about the relationship even while the branding overstates it, describing artificial intelligence and agentic artificial intelligence as embedded throughout the product portfolio, which is a description of something added into products that exist. Cloud native is the better supported half of the claim, since the suites were rebuilt for cloud delivery. Artificial intelligence native is a repositioning of a policy administration company, and a buyer comparing vendors should test it against the same removal question rather than the marketing.

Agents at consequential points with no published boundary, and one capability that warrants a direct question. The estate combines an assistant present on every screen, which is inherently advisory since a person is looking at the screen, with thirteen specialised agents spanning the property and casualty core suite, loss control and the life, accident and health core suite, and agentic automation described across quoting, servicing, billing and claims.

The Spring 2026 additions include agents for bill validation, payment reconciliation and claim reopening. Claim reopening is the one to ask about: reopening or declining to reopen a claim determines whether a policyholder gets a further hearing on a loss they believe was wrongly settled, and it is a decision with a clear affected party.

Across two passes nothing published states what any agent may complete unattended, what a person must approve, what confidence threshold applies, or whether an insurer can configure the boundary, and no governance layer is described.

Mechanisms described, measurements absent. The disclosed mechanisms are better than most: sentence level citation on every document processing finding, telemetry and analytics described as present throughout for proactive insight, quality and reduced risk, a data lakehouse providing real time operational data access, and a named third party model service behind the assistant so the underlying provider is identifiable. Those give a risk function things to inspect.

What none of them supplies is a number. Across two passes no accuracy figure, error rate, precision or recall measure, validation methodology, benchmark, sample or observation period was located for the assistant, any of the thirteen agents or the document product, and no model documentation of the kind an insurer must hold on a vendor model was found. The phrase provable accuracy is used in the document processing material as a claim rather than as a result, with nothing published demonstrating it.

The best operational disclosure in this lane, and it measures delivery rather than sales. Most vendors publish customer counts and logos. This one publishes 176 customer go lives in a single fiscal year, which counts systems actually in production rather than contracts signed, and states that 80 percent of its customers run current generally available versions, which is the honest test of whether a platform can be kept current.

Scale is stated in an externally meaningful unit: more than 100 billion dollars of direct written premium processed on its core platforms as of January 2026 against 36 billion in 2024, alongside 375 or more customers and roughly 500 million dollars of revenue. Growth detail includes eleven new customers and 25 cross sells.

Independent recognition is plural and named, spanning a quadrant leader placement for property and casualty core platforms, top rankings for both property and casualty and life core platforms from a second analyst house, a life policy administration technology award, and repeated industry list placements. No individual customer is named anywhere, and the marketing register is unusually assertive throughout.

One control here is the best anti hallucination mechanism located in this lane and deserves to be described precisely. The document processing product ingests documents of 200 pages and beyond and attaches sentence level citations to every finding it produces, meaning each extracted fact points back to the specific sentence it came from rather than arriving as an assertion.

For document heavy insurance work, where a wrong figure pulled from a policy wording or a medical report propagates into a decision, traceability at sentence granularity is the difference between a reviewer being able to check a finding in seconds and having to re read the document. The company attaches the phrase provable accuracy to it, which overstates what citations deliver, since a correctly cited sentence can still be misread, but the mechanism itself is real.

Supporting it: the model provider is a named enterprise hosted service, and telemetry and analytics are described as present throughout for quality and risk. Absent: model card, evaluation results, red team, incident disclosure and acceptable use boundary.

Nothing published, across a data estate that widened considerably in January 2026. Across two passes no privacy policy content, data processing description, retention schedule, subprocessor list or named privacy regime was located.

The acquisition added group and benefits and pension and retirement administration, which brings health related benefits data, absence management records and retirement balances alongside the existing property, casualty and life books, so the sensitivity of what the platform holds increased at the same time as its breadth.

The generative assistant compounds the question in a specific way: it is described as present on every screen of every solution and powered by a third party hosted model service, which means the assistant sits alongside whatever a user is viewing in policy, claims, benefits or retirement records. Nothing published describes what data reaches that model service, what is retained there, or what boundary separates a customer's records from the assistant's context.

A disclosure gap rather than a security one, with the usual inference available. Across two passes no trust centre, named certification, attestation report, penetration test summary or subprocessor list was located on any public surface.

An independent request for proposal assessment lists security certifications explicitly among the criteria buyers should evaluate for this vendor, naming the relevant frameworks as a category, without confirming which the company holds, which indicates the information sits inside procurement rather than in public.

The controls are very likely substantial, since more than 375 insurers have each completed third party risk assessment before entrusting policy, claims and now retirement administration to the platform, and a major software private equity owner conducted technical diligence. That places this alongside the established vendors here whose empty public surface can be discounted. A direct competitor in the same lane publishes obtainable certificates and audit reports, which shows what is achievable.

An unregulated supplier with compliance offered as product function and no regulatory position of its own. Compliance capability appears in the product descriptions, including artificial intelligence powered compliance support in absence and claims operations, and a core platform administering pension, retirement and group benefits necessarily encodes a large volume of jurisdictional rules. That is capability delivered to customers rather than standing held by the vendor.

Across two passes no financial services authorisation, supervisory examination outcome, named regulatory regime or operational resilience statement was located, and no position on artificial intelligence regulation was found despite the company placing agents into underwriting and claims workflows.

An independent request for proposal assessment treats compliance, security and regulatory support as an evaluation category for this vendor without confirming which certifications it actually holds, which indicates the question is open to buyers rather than answered publicly.

Recognition for artificial intelligence maturity, and no evidence about fairness. The company holds an analyst designation as a pioneer in artificial intelligence maturity, which assesses capability and adoption rather than whether outcomes fall evenly, and the two should not be confused.

The exposure runs across the estate because the agents and assistant reach underwriting, quoting, claims, claim reopening and loss control, all of which produce determinations affecting policyholders and applicants, and the acquisition extended that reach into group benefits and retirement administration where decisions affect people's healthcare access and pension entitlements.

Across two passes no bias testing, fairness metric, disparate impact analysis, model card, explainability documentation or artificial intelligence governance statement was located for the assistant, the thirteen agents or the document processing product. The sentence level citation mechanism supports checking an individual finding, which is a different and narrower thing than knowing whether the system treats groups of people consistently.

No commercial instrument is published. Across two passes no terms of service, master agreement, warranty, indemnity, liability cap, service level or uptime commitment was located, and nothing states what an insurer is owed when the assistant or any of the thirteen agents is wrong. The exposure widened materially in January 2026.

Beyond underwriting and claims decisions affecting policyholders, the estate now administers group benefits and pension and retirement arrangements, where an error reaches a person's healthcare entitlement or their retirement income rather than a premium. The claim reopening agent is the sharpest instance, since a wrong disposition there denies someone a further hearing on a loss.

Every one of those affected parties is a policyholder, member or claimant rather than the customer, is not told a model was involved, and has no published route to see the reasoning or contest it. Nothing published addresses either the insurer's remedy or the individual's.

The model provider is named openly, which remains uncommon enough to distinguish a record. The generative assistant is stated to have been developed in collaboration with Microsoft and to be powered by that company's hosted enterprise model service, so an insurer conducting third party risk assessment can identify exactly whose foundation models sit behind the assistant appearing on every screen, and can assess that provider's own terms, data handling and regional availability directly rather than inheriting an undisclosed dependency.

The data layer is separately identified as a lakehouse providing real time operational access, and the agent estate is enumerated at thirteen specialised agents with their suites named, so the shape of the deployment is visible.

What is not disclosed is depth: no model family or version is given, nothing states whether the thirteen agents use the same provider as the assistant, and the document processing product's underlying technique and provider are undescribed despite it being the component making the strongest accuracy claim.

The strongest published position in this lane on the problem that actually defines core platform ownership, which is not initial integration but staying current. The company states an upgrade cycle of ten days or less for both its property and casualty and life core suites, and backs it with the metric that matters, namely that 80 percent of its customers run current generally available versions.

That is the counterpoint to the recurring complaint recorded against a direct competitor in this index, where independent reviewers describe upgrades involving heavy customisation as taking many months. Around it sits a conventional but well rated integration surface: an independent request for proposal assessment scores ecosystem and integration at four out of five, crediting partner support for rating bureau and distribution integrations and an open posture suited to multi vendor estates, while noting that integration timelines still depend on partner and carrier maturity and that marketplace breadth trails the largest suite vendors. Across two passes no public interface documentation or connector catalogue was located.

Cloud native delivery is the company's central architectural claim and the placement detail behind it is unpublished. The suites are presented as built for cloud rather than migrated to it, and a Microsoft collaboration underpins the generative assistant on a named hosted model service, so the infrastructure lineage is at least partly visible.

Beyond that, across two passes no region list, tenancy description or data residency commitment was located, and no customer hosted option is described. The gap matters more after January 2026 than before it, because the acquisition brought pension, retirement and group benefits administration into the estate, and retirement and benefits records are held for decades and carry jurisdictional handling requirements that differ from general insurance. Nothing published describes whether the acquired platform runs on the same infrastructure as the existing suites or is being converged, which is a live question for a customer buying across both.

No price, unit or tier is published, and two passes across the company's site, its press archive and third party directories produced nothing on how the platform is charged. What is unusual for a privately held vendor is that revenue is disclosed, at roughly 500 million dollars following the January 2026 acquisition, which gives a buyer a sense of the company's scale and stability that most private competitors withhold entirely.

Two published operational figures also bear directly on total cost of ownership, which in core platforms usually exceeds licence cost: a stated upgrade cycle of ten days or less and 80 percent of customers running current versions. Those speak to the recurring cost of staying current, which is where core platform economics actually break.

What remains unknown is the commercial basis, and with a portfolio now spanning three market segments and separately sold core, underwriting, loss control, distribution and digital solutions, nothing indicates whether charging follows premium, policies, modules, seats or subscription.

The widest line coverage in this index, achieved deliberately and evidenced by customer counts on each side. Following the January 2026 acquisition the company covers property and casualty, life, accident and health, group and benefits, and pension and retirement including pension risk transfer, and claims to be the only core platform provider spanning those on one cloud architecture.

The customer base is split rather than asserted in aggregate: more than 275 property and casualty insurers and more than 100 life, accident and health and pension and retirement providers, out of more than 375 in total. The strategic logic is stated with a supporting figure, namely that over 40 percent of insurers operate in both property and casualty and life segments, which is the consolidation opportunity a single line competitor cannot address.

Functional depth spans core administration, underwriting, loss control, distribution management, digital portals, billing and claims. The footprint is described as global without country level detail, and appears primarily North American.