Herald

Graded on the same reasoning applied to other connectivity and workflow platforms in this index. Strip the models and a substantial working product remains, because the carrier connectivity interface came first and is still what the company is bought for: 35 carriers, 80 products and a data standard, none of which requires a model.

The artificial intelligence line arrived by acquisition in July 2025, when Herald bought Babbix and turned it into an AI division, and it is genuine rather than announced: an agent operating in the broker's inbox extracts client data, resolves gaps at intake and assembles submissions. That places models in the operating path for brokers who use the agent workflow and nowhere at all for those consuming the raw interface. Revisit if the agent becomes the reason brokerages buy rather than the connectivity underneath it.

An agent reads a broker's inbox, extracts client information and sends submissions to carriers. Sending a submission is a market facing act performed on behalf of a client, and it is the point at which a broker's professional duty attaches. Nothing public describes an approval step before dispatch, a confidence threshold on the extracted fields, a review queue, or any account of what happens when the agent has read a figure wrongly and the submission has already gone.

This is the same shape the index recorded where an agent conducted outreach to counterparties on live transactions with no described sign off: the automation is real, the control around it is undescribed, and the grade reflects the absence of a stated mechanism rather than any evidence that the automation is unsafe.

Nothing was located on validation, accuracy, monitoring or governance of the extraction and agent capability. There is no error rate for reading a portable document format file or a spreadsheet into a submission, no benchmark, no certification and no published programme. The company's own description of the capability, that it removes the burden of human input, states the objective without any account of how the output is checked once the human is removed from it. That gap is more consequential than for a vendor whose model output is a recommendation, because here the output leaves the building.

Herald states that it works with eight of the 25 largest brokerages, which if accurate is significant penetration of the top of the market, and it names none of them. One named customer voice exists: a national cyber practice leader at the wholesaler Wholesure, quoted on the Babbix acquisition. Counterparty scale is specific and checkable in a way the customer list is not, at more than 35 carrier partners and more than 80 products.

Investors include Lightspeed Venture Partners and Brewer Lane Ventures, the latter also leading the Series B of another vendor in this pocket. What is entirely absent is outcome: no quote turnaround figure, no bind rate, no extraction accuracy, no time saved, no case study with numbers. A named executive quoted and no quantified result is the definition of the middle grade on this axis.

Herald occupies an unusually powerful data position and does not describe how it treats it. Sitting between eight of the largest 25 brokerages and more than 35 carriers means visibility of submission flow across both sides of a market, including which risks are being shopped where, at what point in the cycle, and by whom.

Nothing states whether that aggregate view is used, whether one brokerage's flow informs anything served to a competing brokerage, or whether carriers can see what reaches rivals. The stated ambition to set an open data standard makes the question larger rather than smaller, because a standard setter is also the party best placed to read everything that flows through it.

No data processing agreement, subprocessor list, retention schedule or lawful basis statement was located. The privacy surface here is wider than the product description suggests, because the agent operates inside a broker's email inbox rather than a bounded application.

An inbox contains the client correspondence the agent is there to read and also everything else the broker receives, and nothing describes what is accessed, what is retained, or how material outside the scope of a placement is handled. Commercial insurance data concerns businesses rather than consumers for the most part, which lowers the exposure, but the inbox boundary question is not answered by that.

The site states that the platform carries enterprise security certified for the most sensitive and valuable data, and names no certification. This is the pattern the index has recorded before and it deserves the same treatment: asserting a certification without enumerating it is weaker than saying nothing, because it invites a buyer to assume a report exists that they cannot check, request or verify.

No service organisation control report, no ISO standard, no trust centre, no penetration testing statement and no subprocessor list was located. Service level objectives are offered as tailored to each customer, which is a commercial commitment rather than a security one. Recorded as the third distinct shape of credential inflation observed in this pocket, after a badge row mixing supervisors with certifications and a review site inventing certifications a vendor never claimed.

Herald is a technology supplier and holds no insurance licence, which is the correct posture for its shape and carries no penalty. No statement of regulatory position was located in any form, and none of the usual routes to a higher grade is present: no supervised test, no programme admission, no supervisory engagement.

Worth noting for a later pass rather than graded here: placement is a licensed activity, the platform reaches surplus lines products where filing and diligence duties attach to the producer, and an agent that assembles and dispatches submissions operates inside that regulated act without the vendor addressing it.

No fairness testing, governance programme or impact assessment was located. One product capability deserves recording because it sits close to a live conduct question rather than a theoretical one: the placement system is marketed on the ability to move business toward a brokerage's top carrier partners.

Steering placement by carrier relationship rather than by client fit is precisely the conflict that contingent commission rules and broker duty of care exist to police, and a model that ranks or routes submissions with that objective encoded is making a distribution decision on the client's behalf. Nothing public addresses how client interest and carrier preference are balanced, or whether the broker sees which is driving a recommendation.

No error rate, remediation commitment, liability position or correction path is published, and the consequence chain here is longer than for most vendors in this index. An agent extracts figures from client material and submits them to a carrier. If an exposure, a payroll or a revenue figure is read wrongly and the policy is bound on it, the party who discovers the problem is the insured at claim time, when a misstatement in the application is exactly the ground on which cover is disputed.

The broker carries the professional liability, the carrier carries the underwriting consequence, and the insured carries the loss. Nothing states where responsibility sits or how a misread field is caught before bind.

No model provider, family, version or country of processing is disclosed for the extraction and agent capability. The artificial intelligence line was acquired rather than built, which normally makes the supply chain easier to describe, and no description is offered. A broker cannot determine whether client information pulled from an inbox and assembled into a submission passes through a third party model provider on the way to the carrier.

Integration is the company rather than a feature of it. More than 35 carrier connections carry more than 80 products across six named lines of business, and the interface is designed to drop into whatever the brokerage already runs, whether an agency management system, a customer relationship system or a platform built in house.

A public application programming interface reference is maintained at its own documentation subdomain, which is a level of openness most of this index does not offer. The stated ambition goes further than integration into standard setting: Herald describes itself as building an open data standard for broker to carrier transactions, addressing the duplicative entry and inconsistent formats that make commercial placement slow. An email inbox is treated as a first class integration surface alongside the systems, which is an accurate read of where brokers actually work.

Delivered as hosted infrastructure with service level objectives described as tailored to each customer and tested at production scale. No region list, residency commitment, single tenant option or self hosted path was located. Exposure is narrower than for a global vendor because the footprint appears to be United States only, so a buyer can reasonably infer domestic processing, but inference is not disclosure and nothing states it.

No pricing, tier structure, billing basis or indicative range is published for either the connectivity interface or the agent, and the route to a number is a contact request. This is the index norm and is measured against Sumsub, which publishes per verification rates on a public page.

The absence is more noticeable for an infrastructure product sold to developers, where per call or per transaction pricing is the ordinary convention and a public reference already exists for everything except the cost.

Both sides of the distribution relationship are served, which is unusual: brokerages, wholesalers and managing general agents on one side, and carriers seeking digital distribution reach on the other, plus software platforms embedding the connection. Six lines are named and they describe small and mid commercial business rather than the whole market: cyber, management liability, professional liability, business owners policy, general liability and workers compensation.

Held at the middle grade because the footprint is United States only, the line coverage stops short of property, marine, casualty towers and the specialty classes, and there is no life or health exposure at all.