# How do insurance AI vendors sell to carriers and MGAs?

> **The short answer:** Insurance AI vendors selling underwriting, submission intake and claims tools to carriers and MGAs often win pilots through innovation teams, then stall before production. The vendors that convert qualify the line-of-business budget owner and agree success metrics and a production price before any pilot. Converting vendors also bring security and AI governance into the deal early.

- Page: https://panelhop.com/industries/insurance/insurance-ai-underwriting-claims
- Section: Home › Industries › Insurance › Underwriting, submission intake and claims AI
- Updated 5 October 2026 · Based on Panelhop research, October 2026
- Written for: Vendors of underwriting, submission intake and claims AI selling to carriers, MGAs and brokers
- Publisher: Panelhop (https://panelhop.com/)

**Your AI pilot worked. Nobody funded production.** You sell AI for submission intake, underwriting, claims triage or fraud to carriers, MGAs and brokers. Innovation teams say yes to pilots, and line-of-business owners fund production. Most of your stalled pipeline sits in the gap between them.

- Typical deal: €40–500k a year (Illustrative)
- Sales cycle: 4–14 months (Illustrative)
- Proof-of-concept cap at leading insurers: About 60 days

## Underwriting, submission intake and claims AI · How a deal really moves

**AI deals are won or lost before the pilot. The scout has no budget, and governance arrives after the yes.** (Illustrative)

With Panelhop: The same pilot, gated before it opens. Owner, metric, price and governance agreed before the pilot.

One carrier, 10–16 people (model risk included) and an estimated 4–14 months, most of it decided at the pilot gate.

What opens a deal:

- **Board AI mandate** (Technology)
- **MGA or insurtech funding round** (Funding)
- **NAIC AI bulletin adopted** (Regulation · US)
- **EU AI Act high-risk rules** (Regulation · EU)
- **Catastrophe claims surge** (Budget)

The buyer: **A P&C carrier or MGA**. Also brokers and Lloyd’s syndicates.

Who decides:

| Seat | What worries them | Can veto |
|---|---|---|
| Line owner (CUO or CCO) | Regulatory exposure from a new model in their line. | Yes |
| Innovation team | An AI pilot that nobody funds once it ends. | No |
| AI governance | A model the insurer can’t explain to its supervisor. | Yes |
| CIO and data | A thin AI product the carrier’s own team could build. | Yes |
| CISO, third-party risk | A breach through an AI vendor’s integration. | Yes |
| Underwriters, adjusters | A tool that adds clicks or overrides their judgement. | No |
| Works council (DE, AT) | An AI pilot that starts before a test agreement exists. | Yes |
| Capacity provider (MGAs) | An MGA pricing risk on a model the provider never saw. | Yes |

How the deal moves:

| Stage | Typical time | Where it stalls today | With Panelhop | Service |
|---|---|---|---|---|
| Targeting | – | – | – | – |
| First meeting | – | Your pitch sounds like every AI vendor’s | Weekly briefs: the trigger, the line and who owns its budget | Signal Desk (In-market accounts, weekly): https://panelhop.com/services#signal |
| Discovery | – | The innovation scout can’t say who signs | Budget-owner and governance seats mapped and tracked per account | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| Proof of concept | 1–3 months | A free pilot can then wait 6–12 months | Owner, metric, data agreement and price set before the pilot opens | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| Production case | – | The core integration was never scoped | Installed core and connector on the deal before the pilot opens | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| AI governance review | 1–3 months | Legal reopens over model audit rights | CISO and model governance required at discovery, with a task if missing | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| Contract | 1–3 months | – | – | – |
| Rollout and expansion | – | – | – | – |

With Panelhop across the deal: Signal Desk · weekly on what opens a deal (in-market accounts, scored and mapped); Panel Check on the buyer (coverage baselined); Panel Ops · monthly from the last stage back to the next trigger (scores and plays tuned against the baseline).

*Source: Stages, seats, triggers and stalls from Panelhop research, October 2026; [Alvarez & Marsal and InsurTech NY](https://www.insurtechny.com/wp-content/uploads/2026/04/AM_480341_FSIG_Joint-InsurTech-NY-research-Paper.pdf); the services as described on the Services page.*
*Note: Durations, panel sizes and cycle lengths are Panelhop estimates from our research, not measurements.*

## At a glance

| Fact | Value |
|---|---|
| Typical deal | €40–500k a year (Illustrative) |
| Sales cycle | 4–14 months (Illustrative) |
| Proof-of-concept cap at leading insurers | About 60 days |
| Motion | Pilot first, then a production business case |

*Source: Panelhop research, October 2026; [Alvarez & Marsal and InsurTech NY](https://www.insurtechny.com/wp-content/uploads/2026/04/AM_480341_FSIG_Joint-InsurTech-NY-research-Paper.pdf).*
*Note: Values marked Illustrative are Panelhop estimates from our research, not measurements.*

## Why do insurance AI pilots fail to convert?

**The pilot is where AI pipelines leak.**

Insurance AI pilots fail to convert mainly because they open without a budget owner who will fund production. Panelhop research flagged pilot-to-production conversion as a likely bottleneck for 5 of the 6 insurance AI vendors it studied. The same research flagged integration with the buyer’s existing core for all 6.

**Exhibit 1: Where the pipeline leaks: 5 points across 8 stages.**

1. **Carriers can’t tell you from other AI vendors** (stage: First meeting). What you see: Meeting-to-opportunity conversion falls as carriers say they already talk to several similar AI vendors. Why it happens: The AI vendor’s messaging opens with the technology instead of the line’s loss or expense ratio.
2. **Pilots start with nobody who can fund production** (stage: Proof of concept). What you see: AI opportunities sit in a pilot stage for quarters, and the carrier asks for a second pilot in another line. Why it happens: The pilot was agreed with an innovation team that holds no production budget.
3. **Free pilots eat your engineers and runway** (stage: Proof of concept). What you see: AI pilot scope grows without a fee, and data access takes longer than the pilot itself. Why it happens: No paid-pilot policy, price or data agreement exists before the work starts.
4. **Carrier IT offers to build the tool in-house** (stage: Production case). What you see: AI deals are lost to an in-house build, or the carrier asks for shorter terms and API access to build around the product. Why it happens: The product is positioned as a feature rather than a line outcome, and AI-assisted development makes building cheaper.
5. **Model risk asks for audit rights after the yes** (stage: AI governance review). What you see: Model risk and legal ask for documentation and audit rights after the business has said yes. Why it happens: AI governance and the CISO were never mapped as seats in the carrier’s buying group during discovery.

*Source: Panelhop research, October 2026.*

## What triggers a carrier or MGA to buy underwriting or claims AI?

**Mandates and regulation set the AI buying calendar.**

Board AI mandates, MGA and insurtech funding rounds, catastrophe claims surges and AI regulation open budgets for underwriting and claims AI. Regulation also adds governance seats to every insurance AI deal.

**Exhibit 2: The 5 events that open or close the window for a deal.** (Illustrative)

- **Board AI mandate** (Technology). What happens: An insurer’s board asks for AI in production, not more experiments. Where to spot it: Insurer results calls, annual reports and AI strategy announcements. Window: Within the current budget year, while the mandate has a sponsor.
- **New MGA or insurtech funding round** (Funding). What happens: A newly funded MGA or insurtech carrier buys underwriting and data tools to launch programmes. Where to spot it: Quarterly insurtech funding reports and trade press. Window: Typically 0–6 months after the round closes (illustrative estimate).
- **NAIC AI bulletin adoption** (Regulation). What happens: A US state adopts the NAIC AI model bulletin, making insurers answer for third-party AI and data. Where to spot it: The NAIC adoption map and state insurance department bulletins. Window: Adopted in 25 states plus DC by 31 August 2026, with vendor reviews following each adoption.
- **EU AI Act high-risk rules** (Regulation). What happens: AI used to assess risk and price individuals in life and health insurance falls under the AI Act’s high-risk obligations. Where to spot it: The EU AI Act implementation timeline and supervisor guidance. Window: Obligations apply from 2 December 2027, so buying decisions are likely in late 2026 and 2027 (illustrative estimate).
- **Catastrophe claims surge** (Budget cycle). What happens: A catastrophe floods claims teams and exposes capacity limits in FNOL, triage and fraud. Where to spot it: State regulator claims trackers and carrier loss announcements. Window: System purchases typically follow 3–12 months after the event (illustrative estimate).

*Source: Panelhop research, October 2026; [NAIC](https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf); [EUR-Lex](https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ%3AL_202601744).*
*Note: Timings are Panelhop estimates from our research, not measurements.*

## Who buys underwriting or claims AI at an insurance carrier?

**The scout takes the meeting and the budget owner signs.**

At a carrier, the Chief Underwriting or Chief Claims Officer owns the budget for underwriting or claims AI and answers for the loss or expense ratio. Innovation teams take first meetings but rarely fund production. Model risk and AI governance teams review more deployments as US states adopt the NAIC AI bulletin.

**Exhibit 3: At a carrier buying underwriting or claims AI, 10–16 people sit on the panel and 6 seats can stop the deal.** (Illustrative)

At a carrier buying underwriting or claims AI: 10–16 people.

| Seat | Typical titles | Cares about | Worries about | Can veto |
|---|---|---|---|---|
| Chief Underwriting or Chief Claims Officer | Chief Underwriting Officer, Chief Claims Officer, SIU Director | A lower loss ratio or handling time, with payback inside the budget year. | Regulatory exposure from a new model or data source in their line. | Yes |
| Innovation or digital team | Head of Innovation, Head of Digital Transformation | Proving a new AI capability fast and visibly. | An AI pilot that nobody funds once it ends. | No |
| Model risk and AI governance | Model Risk Lead, AI Governance Lead, Chief Risk Officer | Explainable models, testing for unfair discrimination and documented human oversight. | An AI model the insurer cannot explain to its supervisor. | Yes |
| CIO and data architecture | CIO, Head of Data, Enterprise Architect | Clean integration with the core and the data platforms IT already runs. | A thin AI product the carrier’s own team could build. | Yes |
| CISO and third-party risk | CISO, Head of Third-Party Risk | Where policyholder data goes when an AI vendor processes it. | A breach through an AI vendor’s integration. | Yes |
| Underwriters, adjusters and SIU investigators | Underwriter, Claims Adjuster, SIU Investigator | Fewer manual steps and fewer false positives in their daily queue. | An AI tool that adds clicks or overrides their judgement. | No |
| Works council (Germany and Austria) | Betriebsrat | Whether the AI tool can assess individual staff performance. | An AI pilot that starts before a test agreement exists. | Yes |
| Capacity provider (MGA deals) | Fronting carrier, Reinsurer, Lloyd’s syndicate | Underwriting discipline and auditable decisions under delegated authority. | An MGA pricing risk on a model the capacity provider never reviewed. | Yes |

*Source: Panelhop research, October 2026.*
*Note: The panel size is a Panelhop estimate from our research, not a measurement.*

## What do insurance AI vendors sell, and to whom?

**Your buyer answers for one line’s loss ratio.**

Insurance AI vendors sell underwriting and claims tools that cut handling time, loss costs or leakage in one line of business. The buyers are the line’s underwriting and claims leaders at carriers, MGAs, brokers and syndicates. Deals then expand across the book, and security and AI governance hold vetoes.

**What vendors of this type sell**

- Submission intake and triage for commercial and specialty lines
- Underwriting workbenches and decision support
- Claims triage, FNOL automation and estimating
- Fraud detection and SIU case management
- Automation for policy servicing and operations

**Which carriers, MGAs and brokers buy it**

- P&C, specialty and excess and surplus (E&S) carriers
- MGAs and MGUs, with the capacity provider in the background
- Wholesale and specialty brokers
- Lloyd’s managing agents and their syndicates
- Life and health carriers using AI in underwriting

## How long does it take to sell AI to an insurance carrier?

**AI deals stall between the pilot and production.**

Selling AI to a carrier takes 4–14 months, an illustrative range from Panelhop research, and what is agreed before the pilot decides most of it. Insurance AI deals stall when a pilot starts before anyone agrees who will fund production.

**Exhibit 4: Stage by stage: what you do, what the insurer does, and what changes at the 5 stages where deals stall.** (Illustrative)

| Stage | Typical time | What you do | What the insurer does | Today | With Panelhop | Service |
|---|---|---|---|---|---|---|
| Targeting | – | Targets carriers and MGAs by segment and line, rarely by budget owner. | A board AI mandate, a claims surge or loss-ratio pressure opens a budget. | Carriers are targeted by segment, not by who owns the line’s budget. | A draft ICP and tiers from your own closed-won data, with budget-owner coverage baselined per account. | Panel Check (GTM audit · 2–3 weeks): https://panelhop.com/services#audit |
| First meeting | – | Meets whoever books the demo, often an innovation scout. | An innovation or digital team vets AI vendors for the business. | The first meeting opens with the model, like every other AI pitch. Stalls: Every AI vendor sounds the same. Carriers hear the same purpose-built, live-in-weeks claims from many AI vendors, so a new AI vendor’s first meeting rarely stands out. | Weekly briefs on in-market carriers and MGAs: the signal that fired, the line of business in play and the seat that owns its budget. Your rep owns the first touch. | Signal Desk (In-market accounts, weekly): https://panelhop.com/services#signal |
| Discovery | – | Scopes the AI use case with the scout. | Underwriting or claims leaders define the problem; large carriers may run an RFP or ITN with scripted demos first, and DACH insurers inform the works council. | The AI deal runs through one enthusiastic contact. Stalls: No budget owner is named. When the scout cannot say who is measured on the outcome, the AI deal has a champion and no buyer. | A role map per tier, with missing budget-owner and governance seats enriched and coverage tracked per account. | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| Proof of concept | 1–3 months | Runs a pilot, often free, with no production price agreed. | Tests the AI tool on its own data against agreed metrics, while risk and legal often review the pilot as if it were production. | Pilots start free, with no end date and no owner. Stalls: A finished pilot waits for a decision. When pilot budgets sit apart from the business owner, a successful insurance AI pilot can wait 6–12 months for a decision. | Stage exit criteria: budget owner, success metric, data agreement and production price recorded before the pilot opens. | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| Production case | – | Builds an ROI case from the pilot results. | The line owner takes the case to finance against this year’s budget. | Production integration surfaces after the pilot ends. Stalls: Nobody scoped the core integration. Production integration with the carrier’s installed core was not scoped during the pilot, so carrier IT sizes it only after the results are in and the production order waits. | Each account’s installed core recorded in tiering, and the integration scope a pilot entry criterion next to the budget owner’s sign-off date. | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| AI governance review | 1–3 months | Answers security questionnaires and model questions late. | The CISO, third-party risk and model governance review data use, testing and audit rights. | Security and audit-rights questions reach legal after the yes. Stalls: Audit-rights terms reopen legal. The NAIC AI bulletin asks US insurers to seek audit rights or audit reports from AI vendors where appropriate, and standard SaaS paper often conflicts with those terms late in the deal. | CISO and model-governance seats required at discovery, with a task when one is missing. | Leak Fix (We build the fixes): https://panelhop.com/services#build |
| Contract | 1–3 months | Negotiates the production price and model audit and regulator-cooperation terms. | Procurement and legal agree terms, including cooperation with regulators’ inquiries. | Close dates move without a recorded reason. | Close-date moves and forecast accuracy reported monthly against the baseline. | Panel Ops (We run it monthly): https://panelhop.com/services#run |
| Rollout and expansion | – | Deploys on the pilot line and waits for the next. | Adds lines or books after measuring results on the first. | Each new line waits for a fresh business case. | Expansion triggers and renewal tasks on every live carrier account. | Leak Fix (We build the fixes): https://panelhop.com/services#build |

*Source: Panelhop research, October 2026; [Alvarez & Marsal and InsurTech NY](https://www.insurtechny.com/wp-content/uploads/2026/04/AM_480341_FSIG_Joint-InsurTech-NY-research-Paper.pdf).*
*Note: Typical times are Panelhop estimates from our research, not measurements.*

## How does Panelhop help insurance AI vendors convert pilots?

**Every pilot passes a gate before it opens.**

Panelhop puts the budget-owner, success-metric and governance gates in front of the pilot at named accounts, so the pipeline shows which pilots can convert. A Panel Check (GTM audit · 2–3 weeks) baselines pilot conversion first.

What we baseline and report:

1. Share of pilots opened with a named budget owner and a production price
2. Pilot-to-production conversion, against the baseline
3. Days in security and AI governance review, against the baseline

## What other vendors sell to carriers, MGAs and brokers?

**Other vendor types in insurance.**

The same carriers, MGAs and brokers buy from these vendor types too, through different panels and pipelines.

- [Core policy, claims and billing platforms](https://panelhop.com/industries/insurance/core-policy-claims-billing): Policy administration, claims and billing systems sold to carriers and Lloyd’s managing agents through formal, adviser-led selections.
- [Distribution and producer network platforms](https://panelhop.com/industries/insurance/distribution-producer-platforms): Producer management, comparative rating, carrier connectivity and agency workflow platforms sold to carriers, MGAs, brokers and agencies.
- [MGA and delegated-authority platforms](https://panelhop.com/industries/insurance/mga-delegated-authority-platforms): Policy administration, rating, bordereaux and data platforms sold to MGAs, coverholders and programme administrators.

[The whole insurance market: segments, panel and pipeline →](https://panelhop.com/industries/insurance/)

## What do terms like “Submission intake” and “FNOL” mean?

**The words your buyers use, defined.**

Plain definitions of the terms that come up when you sell underwriting, submission intake and claims AI to carriers, MGAs and brokers.

- **Submission intake**: Reading broker submissions, such as emails, schedules and loss runs, into structured data an underwriter can price. A common first AI use case in commercial and specialty lines.
- **FNOL**: First notice of loss: the first report of a claim to the insurer. FNOL automation captures and triages claims at that moment.
- **SIU**: Special investigations unit: the carrier team that investigates suspected fraud. The SIU judges fraud tools on their false positives.
- **Pilot purgatory**: Market slang for an insurance AI pilot that has finished without a production decision, usually because no budget owner, success metric or production price was agreed before it began.
- **NAIC AI model bulletin**: Guidance from US state insurance regulators on insurers’ use of AI. It makes insurers responsible for governing the third-party AI systems and data they use.
- **Loss ratio**: Claims paid and reserved divided by earned premium. Underwriting and claims leaders are measured on it, so carriers read AI proof through it.

## What do vendors of underwriting, submission intake and claims AI ask about selling to carriers, MGAs and brokers?

**Answers before your next insurer deal.**

### How long should an insurance AI proof of concept last?

An insurance AI proof of concept at a carrier should last about 60 days at most. That is the cap leading insurers use, according to an Alvarez & Marsal and InsurTech NY study. Agree the business owner, the success metrics, the data access and the production price before the pilot starts. Open-ended pilots drift, and finished ones can wait 6–12 months for a decision.

### Who owns an insurance AI tool after the pilot ends?

After the pilot, an insurance AI tool should pass to the line-of-business owner who funds production, such as the Chief Claims or Chief Underwriting Officer. Carriers often leave this open: the innovation team that ran the pilot moves on, and launch budgets omit monitoring, retraining and support. Name the post-pilot owner and those costs in the opportunity before the pilot starts.

### Who is the budget owner for AI at an insurance carrier?

The budget owner for underwriting or claims AI at an insurance carrier is usually the executive measured on the loss or expense ratio. That is the Chief Underwriting Officer, Chief Claims Officer, SIU director or COO. Innovation and digital teams often take the first meeting but hold pilot money and no production budget. A contact who cannot name who signs the production contract is a scout.

### What will insurers ask AI vendors under the NAIC AI model bulletin?

Under the NAIC AI model bulletin, US insurers ask AI vendors for the evidence an insurer needs to govern third-party AI. Insurers ask how the model works, how it was tested for unfair discrimination, what data it uses and whether the insurer gets audit rights. Where appropriate, insurers also write vendor cooperation with regulators’ inquiries into the contract. By 31 August 2026, 25 states plus DC had adopted the bulletin.

### How do insurance AI vendors stand out when every vendor makes the same claims?

Insurance AI vendors stand out with proof in the buyer’s economics rather than claims about the model. Carriers, MGAs and agencies hear the same purpose-built and live-in-weeks messages from many vendors. Gallagher Re found 99.1% of Q2 2026 insurtech funding went to AI-focused companies. Lead with a measured change to one line’s loss or expense ratio, shown for a peer in that line.

### Should insurance AI vendors charge carriers for a pilot?

Yes: insurance AI vendors should agree a pilot price, success criteria, a data agreement and the production price before a carrier pilot starts. Free, open-ended pilots tie up engineers, and carrier risk and legal teams often apply production-scale reviews that stretch a pilot from weeks to months. A pilot fee also shows whether the carrier has a budget owner who will fund production.

## Sources

**Where the numbers come from.**

Sourced figures link to their source below. Figures marked Illustrative, and figures given as estimates, are inferred from Panelhop research. Vendors appear only as types, never by name.

1. [Alvarez & Marsal and InsurTech NY, From Experimentation to Execution Discipline: How Leading Insurers Are Converting Insurtech Partnerships into Measurable Returns (2026)](https://www.insurtechny.com/wp-content/uploads/2026/04/AM_480341_FSIG_Joint-InsurTech-NY-research-Paper.pdf)
2. [Gallagher Re, Gallagher and CB Insights, Global InsurTech Report Q2 2026: Artificial Intelligence, Risks and Opportunities (2026)](https://www.ajg.com/gallagherre/-/media/files/gallagher/gallagherre/news-and-insights/2026/aug/global-insurtech-report-ai-risks-opportunities-q2-2026.pdf)
3. [NAIC, Implementation of NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers (status as of 31 August 2026) (2026)](https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf)
4. [NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers (2023)](https://content.naic.org/sites/default/files/cmte-h-big-data-artificial-intelligence-wg-ai-model-bulletin.pdf.pdf)
5. [EUR-Lex, Regulation (EU) 2026/1744 amending Regulation (EU) 2024/1689 on artificial intelligence (Digital Omnibus on AI) (2026)](https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ%3AL_202601744)
- Panelhop research, October 2026: our analysis of the vendors, buying panels, pipelines and triggers for underwriting, submission intake and claims AI in insurance, from public sources. Vendor names are not published.

## Next step

Find where your pipeline to carriers, MGAs and brokers leaks.

[Book a GTM audit](https://panelhop.com/book) · [What the Panel Check covers](https://panelhop.com/services#audit)