Friends+Family DirectAI Monthly Update (April 2024)
Hey all!
Hey all!
Keeping you up to date with what’s new at DirectAI. To reiterate, we’re enabling anyone to build computer vision models that classify, detect, and track objects in images & videos, without training data. If you missed last month’s update, check it out here!
The Good News
We signed a new recurring revenue customer. We’re enabling them to release their first set of revenue-generating features through customized content moderation.
We completed agreed-upon work for the LOI we signed with a sports streaming company (and earned a one-time payment).
We supported a golf services company during an event they ran for The Masters! We automated components of their video production pipeline, enabling them to get the same work done with fewer hired editors.
We built out functionality that enables us to natively broadcast video to WHIP endpoints with WebRTC as well as swap classifiers/trackers without terminating a stream.
The Bad News
We didn’t convert any of the teams that went through a trial period last month. In one such case, we attempted to outperform an internal model that was trained on a proprietary dataset. It's hard to outperform internal tooling when category names are ambiguous and we aren't given any validation data! We had a call with them where we clarified expectations as to how the model should perform in different circumstances and we're trying again.
While we scoped out product/feature specifications, we weren’t able to agree on live event pricing in a contract with the sports streaming company.
What We Learned
All of our existing recurring revenue customers were inbound leads.
This has been somewhat puzzling given our effort over the past couple of months to ramp up our outbound email volume. This reinforces the fact that we need to distill where our inbound leads are hearing about us, what language sparked interest in getting in touch with us, and how we can repeatedly find leads like our existing ones.
Keep qualifying leads.
The sports streaming brand clearly didn’t have the budget for the work we offered, which wasn’t apparent until months after our first conversation with them. In the future, we should internally develop conviction on an acceptable price for an enterprise project and validate as quickly as possible that a lead has a willingness to pay in that ballpark.
What’s Up Next
We’re building out new features for our existing customers, like adding an integration with LLM providers to automatically double check our model when it's less confident. We’re also making model improvements to support the requests of leads we have in our pipeline!
Asks
Nothing this month! Thanks for your continued support!
Kindly,
Ben