How much does it cost to build an insurance platform
How much does it cost to build an insurance platform
Building an insurance platform in 2026 typically costs $80,000 to $2,000,000+ depending on what "insurance platform" means in your case. A broker portal that sells third-party policies sits at $80k to $200k. A full carrier system with a rate-quote engine, claims, and underwriting runs $500k to $2M. An embedded insurance API (Cover Genius or Branch style) lands at $300k to $1M, plus 18 to 24 months of enterprise sales cycle before revenue compounds.
The biggest cost drivers are not engineering. They are state-by-state licensing (NAIC filings in all 50 states cost $250k+ alone), compliance audits ($50k to $150k per cycle), and the reinsurance or carrier relationships you need before you can write a single policy. Code is the easy part.
Three platforms, three completely different price tags
"Insurance platform" is ambiguous on purpose, because vendors price the ambiguity. Before you scope anything, decide which of these you are actually building.
1. Broker portal: $80,000 to $200,000
This is the cheapest path. You are not the carrier. You are a digital agent reselling policies from Progressive, Hippo, Lemonade, or a Lloyd's syndicate via their API. The platform does quote intake, policy display, payment collection, and commission tracking. The carrier handles underwriting, claims, and regulatory filings.
Engineering scope is modest: a React or Next.js frontend, a Node or Django backend, Stripe or Adyen for payments, Plaid for bank verification, Twilio for SMS, and one or two carrier API integrations (Bold Penguin, Tarmika, or direct). You can ship the first version in 12 to 16 weeks with a mid plus a senior engineer.
The licensing piece is the real cost. You need a producer license in every state you sell, and most aggregators want $5k to $15k per state to handle the paperwork. Or you white-label under an existing MGA and skip the licensing, in exchange for 30 to 50 percent of commission.
2. Full carrier system: $500,000 to $2,000,000+
You are issuing policies on your own paper, which means you own underwriting, claims, reinsurance, statutory accounting, and 50-state regulatory filings. This is what Lemonade, Root, and Hippo built. It is a 24-to-36-month project before you can take your first policy live.
The engineering surface area is large:
- Rate-quote engine. Pulls 50 to 300 variables per quote, runs an actuarial model, returns a price in under 800ms. Typically built on Python or Go with Postgres, often with a feature store like Tecton.
- Policy administration system (PAS). Issues, endorses, cancels, and renews policies. Vendors like Duck Creek and Guidewire charge $2M+ per year. Building in-house takes 12 to 18 months.
- Claims management. First Notice of Loss intake, adjuster workflow, payment disbursement, fraud signals. Integrations with carriers like Snapsheet for photo damage estimation.
- Underwriting workbench. Pulls credit (LexisNexis), MVR (driving records), prior-loss data (CLUE, A-PLUS), property data (Verisk). Each data feed is $0.50 to $5 per query.
- Compliance and reporting. NAIC quarterly filings, annual statements, market conduct exams, state-specific endorsements.
A small team (3 to 5 engineers plus a licensed actuary and a compliance officer) takes 18 to 24 months. Cost: $500k to $1.2M in engineering, another $300k to $800k in compliance, audits, and reinsurance setup.
3. Embedded insurance API: $300,000 to $1,000,000
This is the Cover Genius, Branch, Boost, or Sure model. You build a developer-first API that other platforms (Shopify, Airbnb, Turo, Tesla) embed to sell insurance at the point of transaction. Your customers are product teams, not policyholders.
Engineering investment is moderate ($300k to $700k for v1) but the go-to-market is brutal. Embedded insurance is a 18-to-24-month enterprise sales cycle. You need partnerships with carriers, a developer portal that rivals Stripe's, SOC 2 Type II, ISO 27001, and signed BAAs with every distribution partner. The platforms that win here are the ones that already had carrier relationships before they wrote a line of code.
Cost breakdown by approach
| Approach | Cost | Timeline | Pros | Cons |
|---|---|---|---|---|
| US full-time team (4 engineers + actuary) | $1.6M to $2.4M / yr | 18-24 months to v1 | Domain depth, tenure | Slow to hire (6-9 months), highest fixed cost |
| Insurtech-focused agency (Tintash, EPAM Insurance) | $400k to $900k | 9-15 months | Pre-built modules, NAIC knowledge | Premium pricing, vendor lock-in on modules |
| Freelancers (Upwork, Toptal) | $150k to $450k | 12-20 months | Cheaper hourly | High variance on insurance domain knowledge, you carry compliance risk yourself |
| White-label PAS (Duck Creek, Guidewire, BriteCore) | $300k to $2M / yr in license | 4-9 months to deploy | Pre-certified, carrier-grade | $2M+ annual TCO, rigid templates |
| Cadence | $500 to $2,000 / wk per engineer | 48-hour trial, ship in weeks | AI-native, weekly billing, swap any week, no recruiter loop | Less suited to multi-year enterprise procurement (you bring the compliance officer) |
The right answer depends on which platform you are building. For a broker portal, a 2-engineer Cadence team for 4 to 5 months is the cheapest viable path: roughly $40k to $65k all-in. For a full carrier, you are looking at a hybrid: agency or Duck Creek for the PAS, and a Cadence team handling integrations, claims UI, and underwriting workbench.
Feature-by-feature cost breakdown
These are 2026 prices for the commodity pieces. Build everything else.
| Component | Build cost | SaaS option | Verdict |
|---|---|---|---|
| Auth + RBAC | $8k to $20k | Clerk (free to 10k MAU, then $0.02/MAU), Auth0 ($240/mo+) | Buy. Insurance has 4+ user roles (insured, broker, adjuster, underwriter) and Clerk handles all of them. |
| Payments + ACH | $4k to $12k | Stripe (2.9% + 30¢), Adyen | Buy. Stripe Issuing also covers premium-finance disbursement. |
| Document signing | $3k to $10k | DocuSign ($45/user/mo), Dropbox Sign | Buy. State regulators want auditable e-signature, which DocuSign certifies. |
| KYC / identity | $5k to $15k | Persona ($1.50/check), Onfido | Buy. |
| Address + property data | $0 to $5k | Smarty ($0.005/lookup), Verisk ($0.50 to $3/lookup) | Buy. Property carriers cannot quote without Verisk peril scores. |
| Driving records (MVR) | NA | LexisNexis ($1 to $4/pull), TransUnion DriverRisk | Buy. Direct DMV access requires state-by-state contracts. |
| Rate-quote engine | $80k to $400k | Earnix, Akur8 | Build for differentiation. SaaS rate engines exist but the model is your moat. |
| Policy administration (PAS) | $200k to $800k | Duck Creek, Guidewire, BriteCore | Build if you are flexible. Buy if you need to issue policies in 6 months. |
| Claims FNOL intake | $20k to $80k | Snapsheet, CCC | Build the workflow, buy the photo estimation. |
| Reinsurance ledger | $40k to $150k | Sapiens, Genius Avenue | Build. Reinsurance treaties are too bespoke for SaaS. |
| Hero infra (Postgres, hosting) | $0 to $2k/mo | Supabase, Neon, Vercel | Buy. Same calculus as our cost to migrate MySQL to Postgres guide shows: Postgres is the right default for insurance schemas. |
Insurance is the rare domain where commodity SaaS works for 70 percent of the platform. The custom 30 percent is the rate engine, the underwriting rules, and the claims workflow. That is where engineering dollars belong.
The hidden cost: licensing, compliance, and the 18-month sales cycle
The buyers we talk to who under-budget always under-budget the same three things.
State-by-state licensing. Producer or carrier licensing in all 50 states costs $250k to $400k if you do it yourself (NAIC filings, surplus lines stamps, certificate fees, surety bonds). Aggregators like AgentSync or Agentero cut the per-state work to $1k to $3k but they still take 4 to 9 months for a full 50-state rollout. Plan for a phased launch: start with 5 to 10 states (often TX, CA, FL, NY, IL) and expand quarterly.
Compliance audits. SOC 2 Type II is $40k to $90k for the first year and $25k to $50k annually after. NAIC Model Audit Rule (MAR) audits run $50k to $150k. HIPAA (if you touch health-adjacent data like life or disability) adds another $20k to $60k. If you are taking premiums in California, the CDI conducts a market conduct exam every 5 years that costs $100k+ in legal time alone.
The 18-to-24-month enterprise sales cycle. This is the warning we give every founder. If you are selling to carriers, MGAs, or large distribution partners, expect: 3 months to first meeting, 6 to 9 months in technical evaluation (security questionnaires, penetration testing, sandbox integration), 6 to 12 months in legal (BAA, MSA, data processing agreement, indemnity carve-outs). Companies that ran out of runway in this category usually ran out before the second renewal.
Budget 24 months of runway minimum. If you cannot, the broker portal or the embedded API for SMB platforms (not enterprise carriers) is the faster path to revenue.
How to reduce cost without cutting corners
Five things that consistently save 30 to 50 percent without sacrificing the platform.
- Pick one line of insurance for v1. Auto, homeowners, renters, pet, life: each has different rating variables, regulatory bodies, and reinsurance markets. Founders who build "multi-line from day one" burn 18 months of runway. Lemonade started with renters only. Root started with auto only.
- White-label your first license. Partner with a state-licensed MGA (MGA Insurtech, Accelerant) and sell under their paper for 12 to 24 months while you file your own. You give up 30 to 50 percent of commission but you compress time-to-revenue by 12 months.
- Use Duck Creek or BriteCore for the PAS, build only the UX. A full custom PAS is 14 months. Wrapping Duck Creek with a custom React frontend is 4 months. The carriers do not care what runs underneath.
- AI-native engineers compound on a domain this dense. Insurance has 60 years of accumulated regulatory language. Engineers who can drop NAIC model laws, state-specific endorsement language, and ISO form numbers into Cursor or Claude Code and get accurate code generation ship 2 to 3 times faster than engineers learning the domain by reading PDFs. Every engineer on Cadence is AI-native by default, vetted on Cursor and Claude Code fluency before they unlock bookings. Our internal data shows AI-native engineers ship the first commit in a median 27 hours after booking, which matters when each compliance milestone has a regulatory deadline attached.
- Stage your launch by state, not by feature. It is faster to ship the full product in 5 states than half the product in 50. Regulators love specificity, and your distribution partners can sell the launched states immediately.
The fastest path from idea to insurance platform
Three steps, in order. Do not skip step 1.
- Pick a line and a launch state. Auto in Texas. Renters in California. Pet in Florida. One line, one state, one carrier partnership, one quote-to-bind flow.
- Get the carrier paper or MGA agreement signed before you build. No matter how good the product is, you cannot sell a policy without underwriting capacity. Six months of paperwork happens before code matters. Talk to Hippo, Branch, Boost, Accelerant, or a Lloyd's broker like Westfield Specialty in the first 30 days.
- Book a small engineering team. Ship in weeks, not quarters. If your CTO is the actuary or the compliance officer (common at insurtechs), the engineering bottleneck is the constraint. The fastest founders we work with book a mid plus a senior on Cadence, get the broker portal live in 12 weeks, and reinvest the savings into the licensing and reinsurance work that actually moves the business forward.
For founders unsure whether to build the rate engine in-house or buy Earnix, the same logic from our Build vs Buy vs Hire decision framework applies: build only what differentiates you, buy commodity, hire for both.
Trying to scope a v1? Cadence shortlists 4 vetted engineers in 2 minutes. Every engineer is AI-native, weekly billing, and there is a 48-hour free trial so you can verify domain fit on real code before you commit. See what your insurance team costs on Cadence.
FAQ
How long does it take to build an insurance platform?
A broker portal: 12 to 16 weeks with a 2-engineer team. A full carrier system: 18 to 24 months including compliance and licensing. An embedded insurance API: 9 to 15 months for v1, plus 18 to 24 months of enterprise sales before meaningful revenue. The engineering timeline is rarely the constraint; carrier paper, licensing, and audits are.
Do I need to be a licensed insurance carrier to launch?
No. You can sell as a digital agent or broker under a partner's MGA, which skips most of the licensing burden in exchange for 30 to 50 percent of commission. This is how most insurtechs launch v1 (Hippo did this for 18 months before becoming a carrier). True carrier licensing in all 50 states costs $250k+ and takes 12 to 18 months.
Build a custom PAS or buy Duck Creek?
If you have under $500k in engineering budget or under 12 months of runway: buy. Duck Creek, Guidewire, and BriteCore are battle-tested, pre-certified with state filings, and run by carriers handling $100B+ in premium. If you have $1M+ and 18 to 24 months: building gives you the flexibility to launch new products in weeks instead of quarters, which is a structural advantage at scale.
What tech stack do most insurtechs use?
Postgres for the policy and claims schemas (rarely MongoDB; insurance is heavily relational). Python or Go for the rate engine. Node or Django for application layers. React or Next.js for portals. Stripe for payments. Auth0 or Clerk for auth. AWS or GCP, never Azure unless a carrier partner requires it. The exotic piece is usually the actuarial model framework, which often ends up in Python with NumPy plus a feature store.
Can a non-technical founder build an insurance platform?
Not without a technical co-founder or a fractional CTO. Insurance has too many moving parts (rating, underwriting, claims, compliance, reinsurance, distribution) for a no-code stack to handle. The closest path: white-label a partner's platform (Boost, Branch, Sure) and focus on distribution while a fractional engineering lead (a lead engineer on Cadence at $2,000/week is one option) builds the differentiated layer.