Life insurers can shorten underwriting cycles and price risk more accurately by modernizing data collection, workflow management, and their use of AI and machine learning (ML). The payoff is lower “not taken” rates and life expectancy projections that keep pace with shifting mortality.
Who This Is For
Underwriting, new business, actuarial, and operations leaders at life insurance and annuity carriers who want to cut cycle times, modernize legacy processes, and bring AI into risk assessment responsibly
In Brief
- Faster underwriting directly improves “not taken” rates and customer satisfaction. Generative AI is now the biggest lever for speed.
- Pairing ML and robotic process automation (RPA) with AI agents lets carriers automate exception-heavy underwriting work on top of existing systems without costly replacements.
- Generative AI turns long, static underwriting journeys into personalized experiences that adapt in real time to each applicant’s behavior, risk level, and channel preferences.
In 2022, I described the long-standing inefficiencies faced by insurance underwriters as a “recurring dream”: familiar, frustrating, and stubbornly resistant to change. I wrote about how business process automation, digital data collection, process improvement, and business process management (BPM) could address the persistent and costly problem of slow turnaround times — and how AI and ML could help.
The pandemic exposed a second problem. Life expectancy assumptions that held steady for decades suddenly moved, and underwriters and actuaries needed better tools to account for it.
Some insurers have started waking up from their recurring dreams. Global Growth Insights says 48 percent of insurers are using RPA for claims verification and premium renewals alone. Accelerated underwriting has also gone mainstream. In Gen Re’s 2025 survey of 30 carriers, an average of 59 percent of individual life applications qualified for an accelerated path — reducing time to issue was the top goal.
But none of those advances can match the impact of generative AI. While the bedrock principles of underwriting remain, generative AI has even greater potential to make core tasks like case intake, risk analysis, premium setting, and payout determinations faster and more accurate. It has also added many new terms and concepts underwriters must master.
Below, I’ll walk through five ways life insurers can evolve their underwriting process and where generative AI changes the picture. As you’ll see, these tools make it easier to cut into the amount of time traditional underwriting takes, plus it sharpens your mortality and longevity assumptions and drives higher customer and agent satisfaction scores.
5 Ways Life Insurers Can Use AI for Underwriting Process Improvement
Let’s look at five ways AI can improve life insurance and annuity underwriting.
1. Collect life insurance underwriting information easier and faster.
When insurers decide whether to issue a life, disability, or health policy, they need data to make the best determination. We commonly refer to this information as pending requirements.
Some of this information comes from the applicant, and some comes from outside sources like healthcare providers, prescription histories, and claims data. Many carriers also check MIB Group’s database of coded information from past insurance applications, which helps underwriters catch conditions an applicant didn’t disclose. The type of information (and level of detail needed) can vary significantly across applications.
Some factors life insurers use to decide what information they need to collect include:
- How much and what type of insurance the applicant wants
- The applicant’s age and sex
- How the applicant answers key questions
- Answers from service providers
The leading business problem insurers face is the time required to collect this data. As the days drag on, more applicants walk away from policies they applied for, driving up your “not taken” rate. The longer it takes to gather information, the more likely your company and agent will waste their efforts while you incur the expense of collecting it.
Historically, technology improvements have sped up underwriting, lowered costs, and improved both customer satisfaction and the rate of policies taken. Snail mail and phone calls gave way to email and enterprise application integration, which automated much of the information gathering but didn’t really help with the process itself.
More successful companies modernized their legacy technologies, honed their focus on customer experience, and used data to identify process kinks and document the value of modernization. Now, large language models (LLMs) are giving companies access to solutions that can supercharge all these processes.
For example, conversational AI interfaces can guide applicants through the process in real time, reducing errors and incomplete submissions. Using natural language processing, LLMs can convert handwritten documents (even from doctors and pharmacists!) into usable, AI-ready data. Generative AI can also predict when you’ll need additional data, making it easier to collect everything up front.
2. Rethink underwriting workflows with business process management.
BPM tools aren’t new. Their roots go back to the 1980s, when companies like FileNet built systems to route scanned documents. Over the following decades, BPM platforms added process modeling and rules engines that businesses could maintain themselves, giving carriers a central hub for managing complex underwriting workflows. Robotic process automation then extended that reach into legacy environments, letting bots work with systems that have no APIs.
What’s changed is how much of the underwriting process you can automate. Traditional BPM and RPA follow fixed rules, which works well for structured, repeatable work like data entry and status updates. But underwriting is full of the opposite: attending physician statements, free-text disclosures, and exceptions that don’t fit a rule.
That’s where AI agents come in. They can interpret unstructured documents and work through exceptions, while RPA keeps the predictable steps consistent and auditable. The strongest programs treat RPA and AI agents as a connected system rather than competing options, with smart orchestration layers coordinating on top of existing systems instead of replacing them.
Cost matters when you decide what tool handles which step. RPA runs at roughly a tenth of a cent per task, while AI agents cost 1 to 10 cents per decision. Use RPA for high-volume, rules-based steps and save agents for the steps that need judgment.
AI also raises the temptation to automate first and ask questions later. Resist it. Automating a broken underwriting workflow just produces bad results faster. The understand, simplify, automate (USA) framework keeps the process ahead of the technology, and the math backs it up: A 20-step process running at 95 percent accuracy per step succeeds end to end only 36 percent of the time, while a five-step version succeeds more than 75 percent of the time. AI can even speed up the “understand” step by analyzing process data to surface bottlenecks and pain points.
Finally, modern BPM gives you visibility. Journey analytics show, in real time, where and why applicants drop off during underwriting, so you can adjust before conversion rates suffer. And as agents take on more underwriting decisions, monitoring and governance stop being optional. You need audit trails, clear ownership, and human oversight for any decision that affects compliance or the customer.
3. Speed up underwriting decisions.
All these advances aim to accelerate underwriting. Traditional ML models sift through large amounts of data and predict what information is needed next. Then, they either recommend approval/rejection or provide a summarized recommendation. That way, underwriters don’t have to manually review every file.
Those models have limits, though. Each one typically handles a single task, and they struggle with nuanced reasoning across unstructured data. Generative AI can work with many types of data — including handwritten notes, images, and audio — within a single model, and it understands the relationships in that data with more sophistication.
The bigger shift now is AI agents that act on that reasoning. Instead of answering one question, an agent can work a case from start to finish. It can order and track requirements, read attending physician statements and lab results, and summarize an applicant’s medical history. It can also flag inconsistencies against prescription data, then use its reasoning to draft a recommendation for the underwriter’s review.
Agents also help when your understanding of risk changes. An example: Suppose new research links moderate drinking to higher cancer risk. An applicant reports two or three drinks a week, which your rules have always classified as standard. With traditional ML, you’d analyze the study, retrain the model, and validate it before it reflected the new evidence.
But if you add the research to guidelines an AI agent draws upon, it can flag the case as complex and explain why much sooner. Your team still decides whether the new evidence should change your underwriting rules.
That human role matters. Agents prepare the case and recommend a decision, but underwriters make the call — especially on complex or high-value applications.
Regulators expect it, too. About half of the U.S. have adopted NAIC’s model bulletin on insurers’ use of AI, which makes clear that carriers remain responsible for AI-driven decisions, even when a vendor built the model.
Colorado goes further. The state’s Regulation 10-1-1 rule requires life insurers that leverage external consumer data and predictive models to run a board-overseen governance framework and report on it annually. Build governance into your underwriting AI from the start, with clear documentation and human review of any decision that affects an applicant.
4. Model mortality and longevity risk more accurately.
Speed is only half the equation. The pandemic showed how quickly the mortality assumptions can move behind underwriting and pricing. Life expectancy in the U.S. fell to just under 76.5 years in 2021, then rebounded to 79 years in 2024 — the highest on record, according to final data from the Centers for Disease Control and Prevention. Swings like that make static actuarial tables a liability.
ML can help underwriters, actuaries, and data scientists project life expectancy and healthcare costs more accurately than traditional statistical methods alone. A model trained on a large dataset of relevant factors and outcomes can learn patterns that are hard to see otherwise. Inputs might include:
- Demographic information, like age and sex
- Personal and family medical history
- Lifestyle factors, like diet, exercise, alcohol use, weight, and foreign travel
- Environmental factors, like air quality and access to healthcare
- External sources, like government and scientific statistics, MIB records, submission data, and claims data
Many of the most useful inputs don’t come neatly packaged. They exist in attending physician statements, lab reports, and other unstructured documents. Generative AI can read those documents and cull the details an ML model needs, meaning it isn’t limited to what made it into a database field.
Generative AI can also help test how well your assumptions hold up. Actuaries can use it to build plausible what-if scenarios, such as another pandemic or a faster decline in drinking rates. You can then run those scenarios through the model to see how pricing holds up.
Remember that life expectancy plays a part in annuity underwriting, too, with the risk running in the opposite direction. Because annuities pay income for as long as a contract holder lives, accurate longevity projections matter for them as much as mortality assumptions do for life insurance. As conditions change, ML gives actuaries a way to keep both aspects current. That way, carriers can price fairly and keep the promises they make to customers.
5. Personalize the customer journey.
Even with all these improvements, complex or high-value applications will still take time, especially as insurers and their customers absorb the financial and human effects of unforeseen natural disasters.
That makes communication between applicants and agents more important. A digital customer journey keeps those lines open, from the first time a customer encounters a product through purchase and, ultimately, filing a claim.
With life insurance and annuities especially, this journey is often long, so carriers need efficient, easy-to-access, accurate touchpoints. A digital journey offers multiple channels (apps, web, phone, email, text) with modern interfaces and smooth access, and generative AI makes all of them more powerful.
That’s because generative AI takes digital customer journeys beyond static mapping. Modern systems can adapt in real time to customer behavior, risk level, and channel preferences to create truly personalized experiences.
More dynamic mapping also helps carriers adapt as younger, more tech-savvy consumers start families and enter the life insurance market. These buyers expect tailored products, responsive and proactive interactions, and access from more devices. The behavioral insights generative AI delivers can inform agent outreach and automatically trigger the right communications based on where customers are in their journey.
What’s Next for Life Insurance and Annuity Underwriting
Life insurers are analyzing the causes of underwriting inefficiency and working to stop it. Despite its traditional resistance to change, the insurance industry is seeing real value in modernizing underwriting, especially with generative AI.
Successful insurers will blend these tools with foundational insurance principles and human insight to create a modern, responsive underwriting experience, and they’ll keep their mortality and longevity models current as conditions shift.
One challenge is the ethical dimension of generative AI. That requires greater focus on governance, risk management, and human-centered design. These protect customer data, ensure regulatory compliance, and augment rather than replace human underwriters.
The tools for transformation are here, and the change is already underway. Generative AI and other modern technologies aren’t simply automating old problems. They’re reimagining how carriers approach risk, speed, and service.
Want to hear how other carriers are approaching it? Join our webinar, Modernizing Legacy Insurance Systems With Agentic AI, on Nov. 12, where AI and carrier leaders share what’s working, what’s not, and where people still matter most.
Are you ready to explore how AI can fit into your business but aren’t sure where to start? Our AI experts can guide you through the entire process, from planning to implementation.
