COMPARE · INSURANCE UNDERWRITING AI
MPL and specialty insurance
The best AI for insurance underwriting depends on what is actually slowing your underwriters down. For specialty carriers, MGAs and medical professional liability (MPL) insurers whose bottleneck is submission intake, clearance and triage against their own policy admin, claims and loss history, OutcomeCatalyst is the strongest fit because it connects those systems and delivers the agents for you instead of handing your team a platform to configure. If you want a self-serve underwriting workbench, a new core system, a pricing engine or a third-party risk model, Sixfold, Kalepa, Federato, Cytora, Hyperexponential, Gradient AI and FurtherAI are the serious options, and each is covered fairly below.
This guide is written for heads of underwriting and COOs at small and mid-sized carriers, MGAs and MPL writers. Every vendor fact below was checked on the vendor's own site or in trade press on October 6, 2026. Where we sell against a vendor, we say so.
At a glance
Tool | Best for | Not a fit for | Needs a data team? |
|---|---|---|---|
OutcomeCatalyst | Specialty, MGA and MPL books that need intake, clearance and triage run against their own policy admin, claims and loss history, delivered and operated for them | Teams that want a self-serve tool to configure in-house, or a replacement policy admin system | No. OC builds and runs it; your underwriters approve the work |
Sixfold | Commercial P&C and life and health carriers that want AI risk summaries against their written guidelines | Buyers who need their own claims and policy history wired in without IT involvement | Light for setup; IT support for integrations |
Kalepa | Carriers and MGAs that want one vendor for ingestion through quote and bind | Teams that only want one narrow step automated | Light; it is a software platform your team runs |
Federato | Carriers and MGAs ready to adopt an AI-native core across policy, billing and claims | Lean MGAs that want to keep their current policy admin system | Yes for a core rollout; it is a platform program |
Cytora (Applied Systems) | Commercial insurers, MGAs and wholesale brokers digitizing intake across lines and countries | Buyers who want decision support on their own loss history more than intake digitization | Moderate; configurable platform |
Hyperexponential (hx) | Commercial P&C and specialty pricing teams with actuaries who build their own models | Books without actuarial staff to own models | Yes, actuarial and modeling staff |
Gradient AI | Workers comp, group health and P&C carriers that want predictive scores from a pooled industry dataset | Niche specialty lines where pooled data says little about your risks | Light to moderate; scores feed your workflow |
FurtherAI | Carriers, MGAs and wholesalers automating document-heavy tasks like submissions, SOVs and audits | Teams wanting a system that reasons over their whole book | Light; task-focused agents |
How we judged these tools
Most underwriting AI demos look alike: a PDF goes in, extracted fields come out. The differences show up after the demo, so we looked at four things.
Whose data drives the decision. Third-party data and pooled industry models help in high-volume lines. In MPL and specialty, the signal that matters most is usually your own: prior submissions on the same insured, open claims, loss runs you already hold, and the referral notes your senior underwriters left.
Which step it actually removes. Intake, clearance, triage, rating, quote and bind are different jobs. A tool that is excellent at one can be absent from the next.
Who runs it. Self-serve platforms assume someone on your side configures rules, maps fields and maintains integrations. A 40-person MGA rarely has that person.
What stays human. Every tool here keeps an underwriter on the bind decision. We looked at how clearly each one shows its reasoning so that approval is quick.
Why intake first? Accenture's P&C underwriting survey with The Institutes (2021) found the average underwriter spends about 40% of their time on administrative tasks and only about 30% on actual underwriting. That admin time is where most of the near-term return sits.
1. OutcomeCatalyst: best for specialty, MGA and MPL books that need intake and triage on their own data
OutcomeCatalyst (OC) connects the systems an insurer already runs (policy admin, claims, the submissions inbox, document stores, rating spreadsheets, underwriting guidelines) into one governed AI context layer built on a knowledge graph. We call it the brain. On top of it, OC deploys agents trained to make the calls your veteran underwriters make: is this a clearance conflict, is it in appetite, what is missing, who should see it first.
The submission and triage agent reads each incoming submission, checks it against your existing policies and claims history, flags gaps in the application, and ranks it for the underwriting desk with the reasons written out. An underwriter approves, edits or rejects. Nothing binds without a person.
What makes OC different is the delivery model. OC builds the connections, trains the agents on how your team decides, and runs them after go-live. Your underwriters do not configure rules or map fields. For a specialty book where nobody on staff has "data engineer" in their title, that is usually the deciding factor.
Best for:
MPL insurers, specialty carriers and MGAs whose underwriters lose hours to intake, clearance and chasing missing information
Books where your own policy, claims and loss history is a better signal than third-party data
Teams without a data or engineering group who want the work delivered and operated
Operators who also want the same context layer to serve policy file integrity and claims and subrogation later
Not a fit for:
Carriers that want a self-serve underwriting workbench their own IT team will configure and own
Anyone shopping for a replacement policy admin or billing system. OC works with the one you have.
Pricing teams that need an actuarial modeling environment. Hyperexponential is built for that.
For a step-by-step look at where automation pays off in an underwriting workflow, read our insurance underwriting automation guide.
2. Sixfold
What it is. Sixfold is an AI underwriting assistant. It reads submissions, including loss runs and statements of values in P&C and prescription histories, lab results and driving records in life and health, then scores risk against the carrier's underwriting guidelines with an explanation of its reasoning. Its site names Zurich North America, Guardian, AXIS, Skyward Specialty and Generali GC&C among its users.
Best for. Commercial P&C and life and health carriers that already have written appetite and guidelines and want underwriters reading a summarized, guideline-checked file instead of raw documents. The explainability is a genuine strength.
Where it falls short for a lean operator. Sixfold is a product your team adopts and tunes. Its public site focuses on submission documents and guidelines rather than on reasoning across your own policy admin and claims history, so connecting those still depends on your IT capacity.
3. Kalepa
What it is. Kalepa Copilot is underwriting software with seven modules: submission ingestion, clearance, triage, risk analysis, rating, quote and bind, and portfolio management. It aims to give underwriters one decision-ready view of each risk, combining exposures, loss runs, third-party data and appetite. Kalepa reports on its site (2026) a 58% reduction in quote time for customers.
Best for. Carriers, MGAs, mutuals and brokers that want a single vendor covering the commercial underwriting workflow end to end, sold directly to CUOs and COOs.
Where it falls short for a lean operator. Breadth is the trade-off. Seven modules is a platform decision, and getting value from all of them means your team adopts Kalepa's way of working. If your real pain is one step (clearance against an old policy admin system, say), it can be more than you need.
4. Federato
What it is. Federato started as a RiskOps underwriting platform and now describes itself as an AI-native insurance core covering policy administration, billing, claims and product development. Named users on its site include QBE, Palomar, Ryan Specialty and Accelerant. It raised a $100 million Series D led by Goldman Sachs Asset Management in November 2025 (FinTech Futures, 2025).
Best for. Carriers and MGAs that are ready to replace or consolidate core systems and want appetite-driven underwriting built into that core. Its pitch centers on ranking submissions by appetite and portfolio fit, and its site reports a 50% increase in hit ratio for customers (Federato, 2026).
Where it falls short for a lean operator. A core platform is a multi-system program. An MGA happy with its current policy admin system, or one tied to a carrier partner's system, may not want to move its core to get better triage.
5. Cytora (now part of Applied Systems)
What it is. Cytora is a risk digitization platform that turns submissions, ACORD forms, emails and documents into structured data for underwriting, claims, renewals and mid-term changes. Applied Systems announced its acquisition of Cytora on September 9, 2025, with terms not disclosed (Business Insurance, 2025). Cytora's site still lists commercial insurers, wholesale brokers, MGAs and reinsurers as its customers.
Best for. Insurers and MGAs with high submission volume across several lines or countries, especially those already using Applied software.
Where it falls short for a lean operator. Cytora's center of gravity is intake digitization. Turning a submission into clean data is necessary, but the triage judgment that follows (does this insured have an open claim, did we decline them two years ago) depends on how well it is wired into your own records.
6. Hyperexponential (hx)
What it is. Hyperexponential's hx platform is a pricing and underwriting platform for commercial P&C. Actuaries build and own code-based pricing models, run portfolio batches and publish new model versions, and the company now adds agents for submission triage, renewals and portfolio monitoring. It says insurers writing more than $75 billion in annual commercial P&C premium use it.
Best for. Commercial P&C carriers, specialty MGAs and reinsurers with actuarial teams that want control over complex pricing models.
Where it falls short for a lean operator. hx assumes someone owns the models. A small MGA that rates off a carrier's filed rates or a spreadsheet will not use most of what it offers.
7. Gradient AI
What it is. Gradient AI sells predictive models for underwriting and claims, built on what it calls the industry's only proprietary data repository of pooled insurance data. It serves group health and P&C, with workers' compensation as its most visible P&C line, and works with carriers, MGAs and PEOs. It cites more than 300 enterprise deployments on its site (2026).
Best for. Workers' comp and group health writers that want risk scores informed by far more claims history than they hold alone.
Where it falls short for a lean operator. Pooled data is strongest in high-volume lines. In MPL or thin specialty classes, your own claims and the judgment of your senior underwriters often say more than an industry score, and Gradient is a model provider, not an intake and clearance workflow.
8. FurtherAI
What it is. FurtherAI builds AI agents for insurance tasks: submission intake, policy comparison, underwriting audits, SOV mapping, guideline and authority checks, and claims intake. It raised a $25 million Series A led by Andreessen Horowitz, per its site, and serves carriers, MGAs, wholesalers and reinsurers.
Best for. Teams with a clear list of document-heavy tasks they want off their desks quickly, such as SOV cleanup or policy checking.
Where it falls short for a lean operator. Task agents work task by task. If you want one layer that knows your whole book, so triage, file integrity and claims draw on the same history, you will be stitching that together yourself.
Best AI for MGAs
MGAs are a special case. They are lean, they often run underwriting on a carrier partner's systems or a mid-market policy admin product rather than Guidewire, and they almost never have a data team. The AI that works for a 5,000-person carrier often stalls at an MGA because nobody owns it after go-live.
For an MGA, ask three questions before any demo:
Does it work with the policy admin and claims systems we already use, including our carrier partner's? If the answer starts with "once you migrate," you are buying a core project.
Who maintains it? Appetite shifts, carrier guidelines change and new programs launch. Someone has to update the logic.
Can it clear against our own history? Duplicate submissions from competing brokers and prior declines are where an MGA wastes the most underwriter time.
If the honest answers are "we will need to switch systems" and "your team will maintain it," look at OC's delivered model. If you have an operations lead who wants to own a tool, Kalepa, Cytora and FurtherAI all sell to MGAs and are worth a demo. Federato fits MGAs ready to move to a new core.
Which one should you pick?
Your underwriters drown in intake and clearance, and you have no data team: OutcomeCatalyst.
You have IT capacity and want a guideline-checked risk summary: Sixfold or Kalepa.
You are replacing core systems anyway: Federato.
Your volume problem is digitizing submissions across many lines: Cytora.
Your edge is pricing and you employ actuaries: Hyperexponential.
You write workers' comp or group health and want pooled data: Gradient AI.
You want specific document tasks gone this quarter: FurtherAI.
One opinion we will stand behind: for MPL and most specialty lines, every vendor on this list reads documents well. The hard part is getting the AI to know what your company already knows about the insured, and keeping it right after the vendor's implementation team leaves.
Frequently asked questions
Do we need a data team to use AI for underwriting?
With most platforms, someone on your side configures rules, maps fields and maintains integrations. With OutcomeCatalyst, OC builds and runs the connections and agents, and your underwriters approve the output. If you have no data staff, ask every vendor who maintains the system six months after launch.
Will AI make underwriting decisions without an underwriter?
Not with any of the tools here as typically deployed, and not with OC. Agents prepare, check and rank submissions. An underwriter approves, edits or declines, and the agent's reasoning is shown so that review is quick.
Do you copy our policy and claims data into a new warehouse?
OC connects to the systems you already run and builds a governed knowledge graph over them, so you do not need to stand up a data warehouse first. Ask other vendors the same question, since many require data to be loaded into their platform.
Can it connect to our policy admin system if we are not on Guidewire?
Yes. OC is built to connect to whatever policy admin, claims and document systems an insurer or MGA already uses, including spreadsheets and shared inboxes. We confirm the specifics on a scoping call rather than promising a connector list.
Is AI underwriting worth it for MPL, where volume is low?
Often more so. MPL files are long, prior claims history matters a great deal, and a single missed conflict is expensive. The return comes from underwriter hours recovered on intake and fewer gaps reaching the desk, not from raw volume.
Why not build this ourselves?
You can, if you have engineers who will own it for years. Many lean specialty teams do not, and an internal build that nobody owns tends to stall after the pilot. Our buy vs build guide lays out the trade-offs honestly.
What are your security and compliance standards?
OutcomeCatalyst is HIPAA-aligned and SOC 2 Type 2 aligned, with the formal audit underway and expected to complete before year-end.
Sources
FinTech Futures, "Goldman Sachs leads $100m Series D for insurtech Federato" (2025)
Business Insurance, "Applied Systems buys insurtech" (September 9, 2025)
Book a strategy call to see this on your own book.
Your systems, your documents, and what your people have been carrying around in their heads. Thirty minutes to see what an AI brain could look like in your company.
Book a strategy call
© 2026 OutcomeCatalyst