Four scenarios where an automated underwriting decision goes wrong

Picture a screen that says "we are unable to offer coverage at this time" eleven minutes after you pressed submit. No underwriter called, no nurse visited, and the message gives no reason. That experience is becoming common as insurers use software to approve, rate or decline simple applications on their own. Software is fast, and most of the time it is right. When it is wrong, the cause is usually not a mysterious algorithm but ordinary bad data: an old record, a mixed-up identity, a drug read the wrong way, or a rule that fired for a reason nobody explained to you.

This post walks through four hypothetical households. Each one hits a different kind of error. The names, ages and prices are made up for illustration, and none of them is a real person or a real insurer's price. What is real is the sequence of steps: find out what the decision rested on, correct the data at its source, and ask a human to look again. The laws involved differ by state, so treat the legal points as things to confirm with your state insurance department.

How an automated decision gets made

An "automated" or "instant" decision is still underwriting. The insurer is doing the same job described in how life insurance underwriting works: deciding whether it can offer coverage and at what price. What changes is who does the first pass. The application answers are checked against outside databases, and a set of rules or a statistical model sorts the case into approve, refer to a person, or decline. AI and data in life insurance underwriting and accelerated underwriting: approval without an exam explain the general approach.

The data sources are usually a short list:

• Your application answers, including every yes or no you gave about health, travel, hazardous activities and finances.
• The MIB file. A coded summary of conditions reported by insurers you applied to before (The MIB report: what insurers know about your history).
• A prescription database check. A history of the drugs pharmacies dispensed to you (how insurers check your prescription history).
• A motor vehicle record, and in some programs other outside data.

Any one of these can be wrong, and a rule that is sound in general can still fire wrongly on your case. Insurers write rules to be cautious: a flag that sends a case to a person costs the insurer a little time, while a flag that is missed can cost money. So the system is tuned to send borderline cases away from the fast lane. The other post on this subject, when an online instant decision says no: what to try next, looks at that first moment from the applicant's side, and What the MIB, prescription databases and driving records reveal lists what the databases hold. That is acceptable if the case then reaches a human who can sort it out. It becomes a problem when the output is a flat "no" and the applicant does not know it can be questioned.

The toolkit used in every scenario

Each scenario below uses the same short set of moves. Keep this list in mind as you read.

1. Ask what the decision was. Was it a decline, an offer at a worse rate class, or a referral to full underwriting? The wording of the notice matters.
2. Ask which reports were used. If a consumer report contributed, the adverse-action notice under the Fair Credit Reporting Act (FCRA) generally has to name the reporting company and tell you that you can get a copy and dispute it. If you received no notice, ask the insurer or agent.
3. Ask for the specific reasons in writing. Many states have an insurance information and privacy law based on the NAIC's Model 670 that lets an applicant request the specific reasons for an adverse underwriting decision. Not every state has one, so confirm with your state insurance department. See privacy and your life insurance data: who sees it and your rights.
4. Get your own copies. The MIB file and the prescription report are the first two to request. The FCRA generally entitles you to a free copy after an adverse action based on a report, if you ask within the window the notice gives, which is commonly 60 days.
5. Correct the source, not just the copy. If the error is in a doctor's chart, the doctor fixes it. If it is in a database, you dispute it with that database.
6. Ask for a person. Ask in writing for a human underwriter to review the corrected file, or ask the agent to resubmit the case under full underwriting.
7. Keep a log. Dates, names, what you sent, and what you were told.

Now to the households.

Scenario one: the record that is out of date

The household. Priya, 41, is applying for a $500,000, 20-year term policy so that her husband could keep the house if she died. This is a hypothetical. The application is the instant-decision type, with no exam. Years ago, during a stressful period, Priya had a few high blood pressure readings and was given a short course of medication. Her doctor later documented that her pressure had returned to normal and she stopped the medication.

What goes wrong. An older insurer application of hers left a code in her MIB file for elevated blood pressure. The prescription history shows fills from years back. The rules engine sees the code and the drug history and routes her out of the instant lane. Rather than a decline, she receives an offer in a lower rate class than she expected.

The arithmetic. Say the insurer's standard-class price for Priya would have been an illustrative $40 a month and the offer she got is $64 a month. These are illustrations, not market prices.

• Difference per month: $64 - $40 = $24.
• Per year: $24 x 12 = $288.
• Over the 20-year term: $288 x 20 = $5,760.

If the rate was wrong because of an out-of-date record, that is $5,760 of premiums over the term that should not exist.

What she does.

1. She reads the offer letter and finds that it says a consumer report contributed, and names MIB and a prescription reporting company.
2. She requests her MIB file and her prescription report. Both are free once she has an adverse-action notice or, in MIB's case, once a year.
3. The MIB entry shows the old code. It is accurate as a historical fact: a reading was reported years ago. The question is not whether the code is wrong but whether it still describes her.
4. She asks her doctor for a letter or a recent record showing normal readings and no treatment for several years. This is the kind of evidence underwriters accept; if an underwriter wants more, they may ask for an attending physician statement from the doctor.
5. She writes to the insurer through her agent: "Please reconsider my rating using the attached records and assign a human underwriter."

What often happens. The code does not disappear, since it is a true entry about a past report. But a current record that shows control and time off medication can change how the underwriter weighs it. The insurer may re-rate her upward, or it may decline to change. Either way she has learned that the first answer came from a stale-looking file, and the agent can place the case with another insurer whose guidelines treat resolved blood pressure more favorably. Reconsideration is covered in asking for a better rate class later.

The lesson. A past record that is true can still be misleading if nothing in the file shows what happened next. Supply the missing "next chapter."

Scenario two: the wrong person

The household. Marcus, 36, has a common name, and so does someone else living two states away. Marcus applies for $250,000 of term life and gets an instant decline.

What goes wrong. Identity matching is imperfect. Databases link records using name, birth date and address history, and a loose match can attach another person's prescription fills to the wrong file. The report seen by the rules engine lists drugs Marcus has never taken, including medications associated with serious conditions. The system declines.

How he spots it. Marcus does not know any of this at first. The notice says only that the decision relied in part on a prescription report, and it names the vendor. He requests his copy.

When it arrives he reads it line by line, which is the most useful thing any applicant can do. The pharmacy names are in a state he has never lived in. The prescribing doctors are strangers. The birth year is off by two.

What he does.

1. He disputes the report in writing with the reporting company. Under the FCRA, the company must investigate and correct or delete information it cannot verify. The usual period for a reinvestigation is about 30 days, though it can run longer in some cases.
2. He sends the company proof of his own identity and address history, as the company's instructions require.
3. He tells the insurer, through the agent, that the report was disputed, and asks that the decision be revisited when the corrected report is issued.
4. He checks his MIB file as well, in case the mix-up leaked into other records. It did not.
5. In the meantime he does not apply to five other insurers. Each application leaves a record, and if the same mismatched report is sold to each of them he could collect several declines for one error. He asks the agent to wait until the corrected report exists.

What often happens. Once the report is corrected, the insurer can re-run the case. Insurers generally do not owe a re-run, so the agent's help matters. If the insurer refuses, the corrected report is still valuable, because it is the document every other insurer will see.

The lesson. A decline is not always about you. Read the report as if it belonged to a stranger, because sometimes it does. If the mix-up looks like it could be identity theft, the steps in privacy and your life insurance data: who sees it and your rights about credit freezes and breach notices apply too.

Scenario three: the prescription that means something else

The household. Elena, 52, applies for $300,000 of term coverage. She takes a medication prescribed for nerve pain after an injury. The same drug is also used, at times, for seizure disorders. This is a hypothetical, and the example is deliberately generic: many drugs have more than one use.

What goes wrong. The prescription report lists the drug, the dose and the fill dates, but not the diagnosis. The insurer's rule infers a condition from the drug. In Elena's case the rule assumes the more serious of the two uses and refers her for a full review, or declines depending on the program.

Elena had answered the application truthfully, listing her injury and the nerve pain. The answers and the drug history point in different directions, and a system that cannot read the context treats a mismatch as a risk. This is why, as how insurers check your prescription history explains, a medication is a clue and not a decision.

What she does.

1. She asks the insurer, in writing, which information caused the decline or referral.
2. She asks her doctor for a short letter stating the diagnosis for which the drug was prescribed, the start date, the expected duration and whether there are any seizures or neurological findings (there are none in this hypothetical).
3. She asks the agent to attach the letter and request an underwriter review. She notes that her application disclosed the injury.
4. She keeps a copy of the pharmacy's printout of her fills, so that if the prescription report has an error in dates or doses she can compare.

What often happens. Underwriters see this pattern often. Once the context is documented, the rate class depends on the actual condition, and the case is priced on that. The point of the letter is to give the human a reason to override the rule. If the insurer's program cannot take a referral, the agent can place the application elsewhere.

The lesson. Disclose the reason along with the medication. An application that explains the drug gives the underwriter something to match against the report. See what happens if you lie on a life insurance application for why leaving it out can turn a harmless mismatch into a serious one.

Scenario four: the decline nobody explains

The household. Sam and Jo, both 33, apply separately online for $400,000 of term coverage each. Jo is approved in minutes. Sam, who is a similar age and in similar health, sees a message that no coverage is available. There is no reason given and no phone number for an underwriter.

What goes wrong. Sam does not know. It might be a data error as in the earlier scenarios. It might be an application answer: Sam recently started a new job with some travel, answered a question about foreign travel in a way that triggers a rule, and does not realize it. It might be a build entry that was typed incorrectly. Or it may be a real underwriting reason, such as a recent medical event, that Sam would have to accept.

This scenario has the least information, so it needs the most systematic approach.

What he does.

1. Rereads the application. He downloads or requests a copy and checks every answer for typos: height, weight, birth date, a yes where he meant no. Entry mistakes are more common than anyone likes, and a wrong weight can push a case over a build-chart limit (life insurance, weight and build charts).
2. Looks for the notice. Even an instant online decline is often followed by an email or a letter that names a reporting agency. If it does, he requests those reports.
3. Writes to the insurer. He asks for the specific reasons for the decision and the sources of information used. In a state that has adopted a law based on Model 670, he cites it and asks for his recorded personal information; elsewhere he asks anyway and notes that he intends to complain to the state insurance department if there is no answer.
4. Asks for a human. He asks whether the case can be reviewed by an underwriter or resubmitted with full underwriting, including an exam if needed.
5. Requests his MIB file and checks for entries he does not recognize.
6. Does not give up on coverage. While he waits, he talks to an independent agent about other routes. Declined for life insurance? What to do next lays out the alternatives, from another fully underwritten insurer to simplified or guaranteed-issue products.

What often happens. Sometimes the insurer gives a general reason ("results of our review of the information you provided") and little else. Sometimes there is a specific one, like a database mismatch, that is fixable. If the insurer does not respond, a complaint to the state insurance department often produces an answer that a letter did not; how to file a complaint against a life insurer explains the process.

The lesson. An unexplained decline is a request for information, not the end. Whether a person is entitled to a detailed explanation depends on the state and on what data was used, but asking is free.

What regulators have started to say

The rules about software in underwriting are still being built. There is no single federal law on how life insurers may use artificial intelligence. Colorado has a regulation aimed specifically at life insurers' use of algorithms and outside data, the NAIC adopted a model bulletin in December 2023 that a number of states have issued, and New York's Department of Financial Services issued a circular letter in July 2024. As of this writing in November 2026 the picture keeps changing, so check your state insurance department. AI regulation in insurance: Colorado, the NAIC bulletin and New York has the details.

None of these gives you a right to inspect an insurer's model. What they do is push insurers to keep governance programs, test for unfair discrimination and, in New York's case, explain adverse decisions that rest on outside data. That shapes what you can reasonably ask for: reasons, data sources, and whether a person reviewed the decision.

When to get more help

Handle the simple cases yourself. Escalate when:

• the insurer will not tell you which report it used and you suspect an error;
• a reporting company does not fix a verified mistake after its investigation period;
• you suspect a decision was discriminatory, rather than just wrong;
• the decline is blocking something time-sensitive, such as a business loan or a divorce settlement that requires coverage.

The first stop is your state insurance department, which takes complaints at no cost. For disputes over a consumer report that is still wrong, a consumer attorney can advise on your FCRA rights. Neither this post nor the insurer's notice is a substitute for that advice.

Key takeaways

• Most automated mistakes trace to data: a stale record, a mismatched identity, a drug read without its context, or a typing error on the application.
• Ask the insurer for the decision type, the reports used and the specific reasons, in writing, and request the MIB and prescription reports promptly after an adverse notice.
• Correct errors at the source, then ask for a human underwriter to review the corrected file.
• Do not pile up applications while an error is unresolved; each leaves a record, and a corrected report is worth more than a fifth decline.
• A decline from one insurer is not the industry's verdict; use the time to look at other routes while the dispute runs.