After working on Artificial Intelligence in the highly regulated payments industry founding deeprisk.ai along the way, I feel we have learned a lot and would like to share lessons to fellow AI founders especially in FinTech or broadly, RegTech.
Avoid Starting with Technology
The number one mistake to avoid is thinking like this:
What business can we build around our artificial intelligence expertise?
Start with People
What area or domain are you an expert in? Are you intimately familiar with a “day in the life” of a person or a group in this domain? Even better, are you that person? Build a solution to fully solve that problem. And then use AI to bring the solution to the next level. Your MVP may not need AI.
Real Data
Too often startup teams come up with pitches based on AI models trained on sample data. This is not as valuable as real data and not a compelling pitch because you have no differentiation from 100 other startup teams who can also access the same data. Also, you have no clue how your model would perform on real world data. It might totally change your assumptions regarding your model performance.
Access
You have to start with deep relationships in your domain. Relationships create trust and that’s the only way to get access to real data for your models. Trust goes both ways. Treat real data as the sacred asset it is. Use privacy and security focused practices to ensure you only use the data for what you said you would.
Modular
Use a modular approach. Don’t try to solve an entire industry in one go. Bite off chunks and solve one specific problem and prove your model out. Then go solve the next one.
Change is hard
As a new technology startup in a highly regulated industry, the only way to succeed is with deep relationships and changing one thing at a time. Too much change at once with the latest and greatest but unproven technology can make potential customers reluctant to pull the trigger on a purchase.
In conclusion, the only way forward in a highly regulated industry is to cultivate deep relationships, empathize with people’s real world problems, divide them up into modules, solve them one at a time and use AI as a way to superpower the solution.