Use cases

Any decision a rule can get wrong.

We start in P&C insurance, where one buyer and one quote-site channel cover car, home and small-business pricing. The same method extends to lending, disputes and health: anywhere the right answer is written down in a filing, a policy or a regulation.

Where we start

P&C insurance pricing

Premium is lost at the hand-off between quote sites and insurers, and in rating systems that drift from what was filed. The buyer is the chief underwriting officer.

Personal auto

$344B of premium a year. Verisk estimates at least $29B a year in premium leakage.

Where money is lost

Defaults filled in when quote-site data is translated (+24.9%). Missed household drivers. Renewal price caps that hide most of an error.

Trigger: NAIC AI review supplement; Colorado's AI law from Jan 1, 2027

Homeowners

$172.7B of premium a year.

Where money is lost

Roof data lost on the way from the quote site (+21.2%). A wrong rebuild cost that flows into every part of the price. Vendor models checked only from summaries.

Trigger: Re-approvals of vendors' catastrophe models

Small business

Part of a $104B commercial multi-peril group.

Where money is lost

A business-type classifier's confidence cutoff, and an owner not found in vendor data: $647, or 26%, on one policy priced from a filed manual.

Trigger: Public filed rate manuals make the fastest proof

What it’s worth: a carrier writing $1B of car insurance that recovers 0.5% through pricing and hand-off fixes gains $5M a year. That 0.5% is far below Verisk’s leakage estimate.

For fronting carriers: one evidence format across your MGAs, so delegated-authority audits compare like with like.

Where it extends

The same method, beyond insurance

We have mapped each of these decisions step by step. In every one, a rule, a table or a hand-off sets the outcome, not the model.

Card and bank disputes

$9.8B in disputed charges and $5.9B in chargebacks a year

The table that auto-resolves claims and the formula for what the customer owes. Denials based only on a fraud filter.

Auto loans

$1.71T in balances

A car's trim typed wrong at the dealer pushes the loan across a pricing cutoff: about $1,086 under-priced on a $30,483 loan. Dealer platforms send one application to many lenders, the way quote sites do.

Credit cards

$1.26T in balances

Score cutoffs and the credit-limit table: $3,000 too much credit, or a rate 3 points too low.

Mortgages and home equity

$13.1T in mortgages, $459B in home-equity lines

An outdated credit report adds $2,175. A low automated home valuation cuts a home-equity line by $24,300.

Small-business loans

SBA alone guaranteed $44.8B across 84,400 loans in FY2025

A missed short-term cash advance makes a business look able to repay: $5,750 lost in year one on a loan that should have been declined.

Medicare prior approval

With a clinical partner

About 53M decisions a year; 80.7% of appeals overturned

Software reading faxed records feeds a length-of-stay estimate that gets rounded, ending coverage 12 days early.

Furthest reach

Certifying the loans behind a deal.

Investors who buy bundles of loans, and banks that lend against them, rely on a lender’s loan tape: one row per loan, with its balance, rate and the borrower’s income. A federal disclosure rule (Reg AB, Schedule AL) already fixes those columns, so each row can be checked against the lender’s own decision records.

Dollar examples are composites from our research and are illustrative. Market figures: NAIC 2024 market share; Verisk on auto premium leakage; CFPB card market report (2025); New York Fed Household Debt and Credit (Q2 2026); SBA FY2025 lending; KFF on Medicare Advantage prior authorization (2024).

Free audit

Send one export. Get a findings report.

Start where there’s no security review and no system access: one de-identified file of past decisions. A spreadsheet is fine. We check it against your own rulebook and send back what we found, with dollars attached.

By submitting, you agree to be contacted about Tensyl and to receive occasional updates. We won’t share your email. See our Privacy Policy.

Email only here. We’ll reply with a secure way to send the file, so please don’t include data in this form.

  1. 01

    What we need

    One de-identified export of past decisions, such as past quotes for one state and line with the inputs, rate version and premium charged. No access to your systems.

  2. 02

    What you get

    A findings report: every decision that doesn't match your filing or policy, the rule that caused it, and the dollars involved.

  3. 03

    If it's useful

    A 60 to 90 day pilot on one use case. Your team plants known errors, and your validator re-runs the evidence without us.