Finish What You Start: Doing 98% of the Work, 98% Right Isn’t Good Enough
Productionalizing B2B software isn’t easy. It needs to be trustworthy, consistent, and useful. When you add trying to wrangle the output of an LLM(s) into the mix… it increases the degree of difficulty.
We recently launched our AI commissions tool and codified some obvious yet noteworthy rules for launching an AI MVP. We wanted to share them with those building, so they can accelerate adoption, increase usage, and ensure retention:
1️⃣ Start discrete work
2️⃣ Complete the work
3️⃣ Be correct
4️⃣ Don’t create more work
Want adoption? Start the work and finish it.
"Don’t tell me how hard you work. Tell me how much you get done." – James Ling
Managing commission payments from carriers is like herding cats—technically possible, but mostly chaos. Before building our solution, we explored the market, aiming to be good partners and understand existing tools. What we found:
✅ Some strong commission tools exist, but…
❌ None handled the entire process
❌ Very few initiated the process (e.g., data access & retrieval)
❌ Those that did only covered a fraction of carriers (i.e., direct feeds)
❌ The best solutions were purpose-built for commercial insurance (kudos to those teams!)
So, we focused on building an MVP that could scale to 100% of carriers, cover 100% of retrieval needs, and handle the nuances of the Life & Health insurance sector.
Want Trust and Retention? You Have to Be Correct.
"Be right, a lot." – Amazon Leadership Principle
Once we set our scope, we got to work. This problem was a perfect testing ground for a system that could retrieve and parse data from carrier web portals–especially those without API feeds.
However, parsing PDFs accurately was a major hurdle. Unlike AI-generated content, commission workflows require exact answers—every time. If our system wasn’t always right, it would never be adopted.
🔑 Our solution: A robust review system ensuring all values align across multiple reference sources. 100% accuracy, 100% of the time.
Don’t Build Another AI Tool—Build a System Powered by AI
"Simplicity is the ultimate sophistication." – Leonardo da Vinci
The last thing we wanted was another SaaS portal requiring constant management. Instead, we built a system that works autonomously:
✅ Starts the work – Automatically retrieves data
✅ Completes the work – Seamlessly integrates into downstream systems
✅ Is always correct – Built-in review processes ensure accuracy
✅ Reduces work – Automates many manual tasks in parallel
The Results
We’ve reduced costs by over 45%. Beyond cost savings, automated processing and better-structured data have drastically decreased lag time in identifying clients with payment issues.
Depending on the carrier, we can detect a lapsing policy or missed payment 7–90 days sooner—we’ll measure the impact on net income per policy with more time in the field.