Our first version (which failed miserably) was a prediction-only product with no workflow and a minimal interface. Here's what happened:
- Initially our hypothesis was we could build data models that automatically suggested a specification or tasks based on certain criteria the user identified (for e.g. iOS app for video based calling).
- The system would suggest programming languages, frameworks and a set of tasks based on the criteria
- User would immediately reject the suggestions/predictions outright (even if they were based collectively on data from stack overflow and other platforms). This would be due to a number of reasons: 1) Didn't trust the ML models to make the right predictions 2) The predictions were their first interaction with the platform.
As a result- we went back to the drawing board- and realized that to build a true "smart" project management system, we would need to start with intuitive workflows, and pepper in predictions over time based on usage and a team's actual stats. The AI tld stuck, plus we're planning on bringing the ML models to our public version soon. TBD.
We also plan on using NLP for automatic task connect to Git repos/PRs/commits and ML for predictions around effort estimates. But - we're still some time away from the open beta on making that a reality.
P.S. We recently acquired the tara.com domain which took 2 years of negotiations. A story for another time.
- Initially our hypothesis was we could build data models that automatically suggested a specification or tasks based on certain criteria the user identified (for e.g. iOS app for video based calling). - The system would suggest programming languages, frameworks and a set of tasks based on the criteria - User would immediately reject the suggestions/predictions outright (even if they were based collectively on data from stack overflow and other platforms). This would be due to a number of reasons: 1) Didn't trust the ML models to make the right predictions 2) The predictions were their first interaction with the platform.
As a result- we went back to the drawing board- and realized that to build a true "smart" project management system, we would need to start with intuitive workflows, and pepper in predictions over time based on usage and a team's actual stats. The AI tld stuck, plus we're planning on bringing the ML models to our public version soon. TBD.
We also plan on using NLP for automatic task connect to Git repos/PRs/commits and ML for predictions around effort estimates. But - we're still some time away from the open beta on making that a reality.
P.S. We recently acquired the tara.com domain which took 2 years of negotiations. A story for another time.