For Insurance Underwriters & Risk Teams

Run AI on loss runs and policy data. Without it ever leaving your control.

Loss history and policyholder data are exactly what AI is good at analyzing, and exactly what you've been told to keep away from it. Nera closes that gap without asking you to trust a policy.

Pro is $299/mo for one seat, month-to-month, no contract. Need SSO, role-based access, or audit logs? See Business pricing below.

What this is actually worth to you

Numbers from independent research on AI-assisted analytics adoption, not a Nera-specific claim, but a real starting point for the math on your own team.

60–80%
less time on manual data prep, review, and routine reporting for teams using AI-assisted analytics
9.1 hrs/week
lost per analyst to inefficient manual workflows, on average (2025 benchmark)
$299/mo
what Pro costs. A fraction of one underwriter's time on a single manual loss-run review.
Faster loss-run review means faster quotes. Faster quotes win more business, and every account reviewed with less manual effort is capacity for another one.
Source: dbt Labs & The Harris Poll, 2025; enterprise AI analytics adoption research, 2025.

What every underwriting team already knows

The data that makes underwriting decisions better is the data you're least able to run through AI.

Loss run review is still manual

Five years of claims history, read line by line, for every account. There's been no safe way to point AI at it, so nobody has.

Policyholder data is regulated

State privacy and insurance regulations don't leave room for "we were careful." The data needs to stay structurally protected, not just handled carefully.

Patterns get missed at scale

Anomalies across a book of business are hard to catch by hand, and the tools that could catch them automatically aren't ones you can safely use today.

How it actually works
Policy and claims data is encrypted before it leaves your environment. It stays that way, even during analysis.
See how the architecture works →
Ready to see it on a real book of business?Start now, month-to-month, no long-term contract →
Claims dataEncrypted, your keys
The AI modelQuery + structure only
Your answerDecrypted with your key

"Why not just use ChatGPT Enterprise?"

Fair question. Here's the answer, no dodge.

Enterprise tiers protect you with a policy

Admin controls, a zero-retention agreement, a contract that says your data won't be used to train the model. Real protections. But the model still receives policy and claims data in a readable form, every safeguard is a promise about what happens after that.

Nera removes the question entirely

The model never receives policy and claims data in a readable form, not under a policy, by architecture. There's no promise to trust, because there's nothing exposed to begin with.

Already just pasting it into ChatGPT directly?

You're not alone, plenty of teams skip the enterprise conversation and use the tools they already have. Nera runs in the same kind of chat interface, nothing new to learn. Just add the one thing that was missing.

What it actually costs to protect a team your size

Per-seat sticker prices don't matter if you can't buy at that seat count. Here's the real annual cost to cover an 8-person team doing this actual job, sustained analytical work, not casual chat, control plane included.

ChatGPT Enterprise
$108,000/yr
150-seat minimum, forced, whether you need 8 people or 150. You're paying for 142 seats you'll never use.
Claude Enterprise
$75,000–$147,000/yr
Base: 20-seat minimum (you need 8) at $20/mo, $4,800/yr before anyone sends a query. Usage: sustained analytical work, the kind this job actually is, runs $750–$1,500/user/month per published heavy-workload benchmarks. For 8 active users, that's $72,000–$144,000/yr on top. You won't know where in that range you land until the bill arrives.
Nera Business
$48,000/yr
$4,000/mo, flat, for up to 8 users, control plane included. No usage meter, no surprise, no guessing what next month costs.

Figures based on an 8-person team doing sustained analytical work. Usage estimates drawn from published heavy-workload benchmarks for comparable AI usage patterns, actual costs vary by vendor and usage pattern, confirm current rates directly with each vendor.

A premium capability, not a premium price

Even at the top of any Enterprise contract, ChatGPT and Claude still receive your data in a readable form, that's true whether you're paying $8,000 a year or $150,000. More spend buys better controls around the data. It doesn't change whether the model sees it. Full data isolation, residency, and sovereignty, the actual requirement across regulated industries and regulated markets, isn't a pricing tier anyone else sells at any price. Nera includes it by default, and doesn't charge a premium to do it.

Nera's own pricing, plainly

Pro
$299/mo
One seat. $250/mo if billed annually. No contract.
Business
$4,000/mo
Includes SSO, role-based access control, and audit logging, plus up to 8 users. Additional seats are $449/mo each.
Enterprise
Custom
For organizations that need dedicated deployment and custom integration with their own systems. Contact us to talk through your requirements.

See the full pricing page for what's included at each tier.

Ask it like you'd brief a colleague

Summarize the loss history for this account over the last five years.
Flag anomalies in this book of business compared to the prior period.
Compare loss ratios across these segments and highlight outliers.
What changed in this account's claims pattern since the last renewal?

Built for regulated risk data

Nera is engaging with underwriting and risk teams evaluating exactly this use case today.

Live
On Google Cloud, official partnership
$299/mo
Pro tier, per user
GDPR & HIPAA
Compliant today
FAQ

Questions underwriting teams ask us

No. The model receives your question and the structure of the data, never the underlying claims values themselves. That's the core of how Nera works, not a configuration option.
Nera works with structured data, spreadsheets, exports, and similar formats already common in underwriting workflows. Bring us a sample and we'll walk through the fit directly.
Because the model never receives readable policyholder data, the architecture itself reduces exposure regardless of which state's requirements apply. Worth a direct conversation on your specific regulatory footprint.
No. Not ours, not anyone's. It can't be, the model never sees it in a form worth training on.

Bring us a real book of business

Fifteen minutes, no obligation. See what it actually looks like on your data.

Book a call