Sign up to save your podcasts
Or
Share The Skeptic’s Guide to Production-Ready AI: Building a High-Trust Engine
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
The Skeptic’s Guide to Production-Ready AI: Building a High-Trust Engine
Building a High-Trust AI Engine for Insurance
In this episode of the Benefits Broker Boost Podcast, Adam Smith, Plansight CMO, is joined by Brent Combs, the lead engineer behind the Plansight AI engine. With a background in underwriting and a deep passion for mathematical problem-solving, Brent shares the philosophy that drives the platform’s development. The conversation explores the shift from deterministic models to probabilistic AI, the importance of purpose-built AI constraints, and how Plansight achieves 91.09% accuracy on the industry’s most challenging documents.
The Skeptic’s Approach to Innovation
Brent explains why he was originally a detractor of general-purpose AI and how he had to prove its validity to himself before building it into the platform. This “trust but verify” approach led to a system that is “guilty until proven innocent,” ensuring that every extracted value is held to a high standard of accuracy and consistency. By choosing the right AI for specific tasks, Plansight avoids the common pitfalls of “drift” found in standard chatbots.
Defining “Shape Accuracy”
A core theme of the discussion is the difference between content accuracy and “shape” accuracy. Brent details why Plansight uses a punitive grading system: if the AI correctly identifies data but fails to organize it into the correct plans, classes, or networks, it receives a zero. This ensures that the smartsheeting AI output is production-ready and requires minimal touch from the user.
The Power of Annotations
One of the most exciting features discussed is Annotations. This “breadcrumb trail” allows users to click on any extracted data point and instantly see its source in the original document. This feature is designed to close the trust gap, allowing brokers to move from data entry to strategy without losing the ability to validate information with their own eyes.
Unlocking 40% More Time
By automating the grueling task of manual spreadsheeting, Plansight helps marketing teams save a conservative 40% of their time. Adam and Brent discuss how this reclaimed time allows brokers to lean into their passions, solve more interesting problems, and provide a better experience for their clients. The goal is to move the industry toward scaling smarter with AI while keeping the human expert at the center of the process.
Ready To Streamline Your Strategic Presentations?
Plansight is the only end-to-end benefits marketing platform that uses AI and automation to ensure every renewal and RFP is consistent, accurate, and fast. By taking the manual work out of benefit planning, Plansight helps you stay ahead of the curve and deliver faster, smarter results to your clients. Visit Plansight to learn more about making your brokerage more efficient.
View all episodes
By PlansightBuilding a High-Trust AI Engine for Insurance
In this episode of the Benefits Broker Boost Podcast, Adam Smith, Plansight CMO, is joined by Brent Combs, the lead engineer behind the Plansight AI engine. With a background in underwriting and a deep passion for mathematical problem-solving, Brent shares the philosophy that drives the platform’s development. The conversation explores the shift from deterministic models to probabilistic AI, the importance of purpose-built AI constraints, and how Plansight achieves 91.09% accuracy on the industry’s most challenging documents.
The Skeptic’s Approach to Innovation
Brent explains why he was originally a detractor of general-purpose AI and how he had to prove its validity to himself before building it into the platform. This “trust but verify” approach led to a system that is “guilty until proven innocent,” ensuring that every extracted value is held to a high standard of accuracy and consistency. By choosing the right AI for specific tasks, Plansight avoids the common pitfalls of “drift” found in standard chatbots.
Defining “Shape Accuracy”
A core theme of the discussion is the difference between content accuracy and “shape” accuracy. Brent details why Plansight uses a punitive grading system: if the AI correctly identifies data but fails to organize it into the correct plans, classes, or networks, it receives a zero. This ensures that the smartsheeting AI output is production-ready and requires minimal touch from the user.
The Power of Annotations
One of the most exciting features discussed is Annotations. This “breadcrumb trail” allows users to click on any extracted data point and instantly see its source in the original document. This feature is designed to close the trust gap, allowing brokers to move from data entry to strategy without losing the ability to validate information with their own eyes.
Unlocking 40% More Time
By automating the grueling task of manual spreadsheeting, Plansight helps marketing teams save a conservative 40% of their time. Adam and Brent discuss how this reclaimed time allows brokers to lean into their passions, solve more interesting problems, and provide a better experience for their clients. The goal is to move the industry toward scaling smarter with AI while keeping the human expert at the center of the process.
Ready To Streamline Your Strategic Presentations?
Plansight is the only end-to-end benefits marketing platform that uses AI and automation to ensure every renewal and RFP is consistent, accurate, and fast. By taking the manual work out of benefit planning, Plansight helps you stay ahead of the curve and deliver faster, smarter results to your clients. Visit Plansight to learn more about making your brokerage more efficient.