Vertafore

Remove the models and the business is barely disturbed, which places this with its co incumbent rather than with the specialists. What remains is a management system estate spanning three agency size tiers, native carrier download connectivity, benefits brokerage software, insurance reference content and, most durably, a producer licensing and compliance network the company states carries half of all industry compliance transactions. None of that depends on a learned component.

The artificial intelligence platform is genuine, recent and clearly incremental: introduced in April 2026 with agents arriving through June and July, layered onto systems whose architecture dates back decades, and scoped to email handling, statement matching, document extraction and knowledge retrieval. Independent analysis of this market reaches the same reading, treating the agent surface as incremental to a long established management system rather than as the thing the platform is.

The oversight formulation is more specific than the lane norm and is stated consistently across the agent line rather than for a single product: professionals stay in control for review, approval and exception handling. Naming exception handling alongside review and approval is the detail worth crediting, because it acknowledges the failure path rather than describing only the happy one, and exceptions are where an automated matching or extraction process actually needs a person.

The wider framing is consistent, with capability positioned to enhance rather than replace expertise, and the submission agent's output described as organised data for underwriting review rather than as an underwriting outcome. What keeps it below the top band is that all of this appears in announcement language rather than in product documentation defining the control, with nothing published naming which action classes require approval, what confidence threshold produces an exception, or what a user sees before accepting an agent's output.

One genuine transparency claim exists and the measurements behind it do not. The coverage research agent is described as returning clear and explainable responses grounded in authoritative insurance content, which is a grounding claim rather than a fluency claim and is the right architecture for a domain where a wrong answer about coverage becomes advice.

Against that, two passes located no accuracy or error rate for any agent, no validation methodology, no benchmark, no drift monitoring and no model documentation. Independent analysis of this vendor reaches the same conclusion and lists the missing figures explicitly, naming extraction accuracy, reconciliation match rate and straight through processing rate as open questions a reviewer should put to the company.

The reconciliation agent is the clearest instance, because statement matching has a verifiable right answer and a published match rate with an exception rate beside it would be straightforward to produce.

Market position is evidenced at a scale few vendors in any lane reach. The company states support for more than 95 percent of the top agencies and insurers and half of all industry compliance transactions, and independent analysis puts the footprint at more than 20,000 agencies and 1,000 carriers, with 98 of the top 100 agencies and 96 of the top 100 carriers. Reaching both sides of the distribution relationship at that concentration is a structural fact rather than a marketing claim.

Two artificial intelligence outcomes are published with figures attached: benefit plan setup reduced from 20 to 30 minutes to under five, and manual reconciliation effort cut by 90 percent. Three reservations apply and none of them displaces the grade. The outcome figures carry no measurement method or customer attestation. The agents are very recent, so adoption counts do not yet exist. And independent analysis of this vendor lists the absence of accuracy and straight through processing figures as an open question rather than a settled one.

The published position is a paragraph of intent rather than a set of commitments. The company states that it builds artificial intelligence to be trusted with ethical and responsible use at the centre, applies what it calls agile governance, and designs models to be reliable and accurate and to enhance rather than replace professional expertise.

Two passes then located nothing operational behind that: no statement on whether agency, insured, producer or carrier data is used to train or improve any model, no data handling policy for the agents, no description of what leaves the platform during inference, and no retention position for the documents and email content the agents process.

The exposure spans an unusually wide range of parties, because the agents read agency email, extract from carrier documents, process managing general agent submissions and match carrier commission statements, while a separate network holds licensing records for individual producers.

Two passes located no privacy policy content covering agency, insured, producer or carrier data, no data processing agreement, no retention schedule and no subprocessor list. The published statement closest to the subject appeals to the parent rather than to a document, describing security and transparency as backed by a committed owner, which is a reassurance about corporate seriousness rather than a commitment a buyer's counsel can read.

Two data categories deserve specific attention here. Agencies hold personal and commercial information about insureds who are not the customer, and the artificial intelligence layer now reads their email and extracts from carrier documents. Separately, the compliance network processes producer licensing data, which is personal information about individual insurance professionals held at industry scale, and nothing published describes its handling, retention or use.

Two passes across the company's site, its press archive, its artificial intelligence pages and third party coverage located no trust centre, no service organisation control report, no information security certification and no dedicated security page. The one published statement on the subject appeals to corporate parentage rather than to evidence, describing security and transparency as backed by a committed owner, and a parent company's reputation is not a control assessment.

The contrast with the other incumbent in this market is stark and worth recording, since that competitor publishes its audit cadence, the criteria assessed and the products in scope. The grade rests on inference rather than evidence: a vendor supporting more than 95 percent of the top agencies and insurers and carrying half of industry compliance transactions has been examined repeatedly by parties equipped to demand proof, so the controls very probably exist unpublished.

This is where the record separates from its co incumbent, and the reason is infrastructure rather than intent. The company operates the producer licensing and compliance network through which appointments and licensing transactions move between producers, agencies, carriers and state regulators, and it states that half of all industry compliance transactions run through it.

That places the software inside a statutory process, since producer licensing and carrier appointment are legal prerequisites for transacting insurance and are administered by state departments, and a failure in that flow has regulatory consequences rather than merely operational ones. Anomaly detection is applied to the same flow to flag data problems before they cause disruption.

The company holds no licence of its own, which is why this sits below the top band, and two passes located no description of how the network is examined, what accuracy commitment attaches to a filing, or where responsibility rests when a transaction fails.

The company publishes guidance telling its customers that artificial intelligence governance is no longer optional and that auditing models improves fairness, compliance and confidence in decisions, and two passes located no audit, fairness result, bias evaluation or model card covering its own models. That pattern, publishing the standard as thought leadership while evidencing none of it, appears elsewhere in this lane and is worth naming each time it occurs.

The governance language on its own surfaces is procedural and unquantified, describing ethical and responsible use and agile governance without naming a structure, a review body, a standard adopted or a testing regime. The distribution side exposure is about advice and attention rather than pricing, since a coverage research agent shapes what a producer tells a client and extraction agents determine which submissions reach an underwriter in usable form.

Recourse is unaddressed in both directions and the intermediary structure diffuses it further. The customer is a licensed professional carrying its own duty of care, so when a coverage research agent returns a wrong answer that becomes advice to a client, the exposure lands on the agency's errors and omissions cover, with nothing published allocating anything between the parties.

The submission agent adds a second affected party, since a business whose submission is extracted incompletely may reach an underwriter looking worse than it is, or not reach one at all, and it will never know why. The insured and the applicant have no relationship with this vendor, no notice that agents shaped what they were told or how their submission was presented, and no route to either party. Two passes located no liability language, no error handling policy and no artificial intelligence specific service commitment.

The platform is named, the agents are named, their functions are described in detail, and what powers any of them is never stated. Two passes located no provider, model family, version, hosting arrangement or fine tuning approach for the artificial intelligence platform or for any of the six agents announced during 2026, and no notification commitment when an underlying model changes.

The coverage research agent is the one where provenance would matter most to a buyer, because its answers are described as grounded in authoritative insurance content, and grounding is a property of a retrieval architecture whose components a buyer cannot assess without knowing what they are. The direct comparison in this market names its supplier for at least one capability, which establishes that disclosure at this level is commercially possible in this segment rather than unheard of.

The company owns rails on two separate dimensions, which is what lifts this to the top band. On the servicing side it is the management system for a large share of the independent agency market, with native carrier download connectivity built in rather than bolted on, so policy data flows from insurers into the agency's system of record as a property of the platform.

On the regulatory side the licensing and compliance network carries a stated half of all industry compliance transactions between producers, agencies, carriers and state regulators. Around those sit the client portal, client communications, certificates, benefits brokerage, reference content and managing general agent workflows, integrated as a suite with single sign on across the enterprise products.

The artificial intelligence follows the same principle, embedded into the systems professionals already work in rather than presented as a separate destination, including inside the email client where agency work actually happens.

Two passes located no deployment documentation: no hosting provider, no region list, no processing location statement, no tenancy description, no residency commitment and no private deployment path. The estate compounds the question rather than simplifying it, because it was assembled over two decades through acquisition and a merger of two separate software companies, and independent analysis notes that the older management systems carry legacy architecture with cloud migration still in progress.

That means products of different vintages plausibly run on different infrastructure, and nothing published describes which run where or how data moves between them. Operations appear concentrated in the United States, which limits the cross border weight of these omissions without answering the questions a buyer's information security function would ask.

Pricing is quote only across the estate, with no published band for any management system tier, and independent cohort analysis of this market explicitly penalises the vendor for that motion. The gap between customer types makes the silence more consequential than usual: a small agency on the entry tier and a multi location brokerage running the enterprise system with inter divisional accounting are buying products with different architectures and presumably very different economics, and neither can establish a starting point.

Nothing published indicates the unit of billing, contract term, or whether the Velocity agents released through 2026 are included with the underlying system or licensed separately. That last question is live because the agents shipped into specific products at specific dates rather than as a platform wide upgrade, and the compliance network raises its own undisclosed question about who pays on a two sided transaction flow.

Coverage spans the whole distribution channel and reaches the carrier side as well, which is what the top band requires. Agency size is addressed with distinct products rather than configurations, from smaller independent operations through the mid market flagship to an enterprise system built for agencies above 100 staff handling multi location structures, specialty lines and acquisition driven growth.

Beyond agencies the customer set includes managing general agents, with dedicated workflows and a submission agent built for them, employee benefits brokers through a separate platform, and carriers, whom the company reports supporting at more than 95 percent of the top 100. The compliance network extends further still, reaching individual producers and state regulators as parties to licensing transactions. Line coverage runs across property and casualty and benefits. The honest limit is the same as its co incumbent's: this is distribution technology, with no carrier core, underwriting or claims system anywhere in the estate.