Friends+Family DirectAI Monthly Update (July 2023)
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!
Last Month’s Rundown
We signed our second recurring revenue contract! We’re helping an interior design software company detect and classify furniture in user-uploaded images. Most excitingly, they switched over to us directly from Google Vision.
We publicly released our image classification and detection API! Anyone can generate credentials and make up to 1000 calls per month for free (no credit card required)! Check out our docs and quickstart repo to learn more.
In a switch up from last month’s projected announcement, we’ve partnered with Ant Media to release a plugin that natively integrates computer vision into livestream video. Ant Media is the streaming engine for thousands of companies in over one hundred countries and we’re excited to work more closely with them in the coming weeks.
What We Learned
We lost a deal on price this past week. Pricing (and the sales pipeline in general) is one of our weaknesses.
Learning: Lean into value-based pricing and rigorously evaluate Willingness To Pay before quoting a “cost-plus” amount. Monetizing Innovation and the author’s First Round Capital blog post on the same topic have been recommended. We’re open to alternate suggestions!
Learning #2: We’re very much in the “do things that don’t scale”/design partner recruitment phase of the company’s growth. Top of mind question in the next couple months: are we committing to a bottom-up or top-down sales motion?
We got a bunch of useful feedback from users/leads this month after opening up the API. For one, our customers reported several “Internal Server Error” responses after making requests.
Learning: We should be running unit tests. After a testing hackathon last week, we now have infrastructure to support them! (Code coverage is very low but rising)
Learning #2: We should know about server errors before our clients. The aforementioned testing hackathon also yielded a monitoring system that will alert us when things go down!
Several users reported confusion with terms (e.g. NMS threshold) in our demo and API documentation.
Learning: Our marketing and documentation needs to be well-aligned with our ideal customer profile. If we’re trying to make computer vision accessible, we should provide explanations for some of the more esoteric domain-specific phrases and abstain from using them in our marketing copy.
What’s Up Next
We’re doubling down on our partnership with Ant Media. This will include a marketing push (blog post incoming) as well as some development work on the plugin (ingesting images more efficiently). We’re highly motivated to bring object tracking costs down by speeding up inference.
We’re rethinking our top of funnel and aiming for meetings with 15 qualified leads this month. Hoping to do some guerilla marketing in a few tech communities, Google Ads experiments, and another public feature release.
Asks
Please star our quickstart repo on Github! I will help you make an account if you don’t have one!
Do you or anyone you know use Google Vision, AWS Rekognition, or Ant Media? Please help me set up time to talk with them! I have a forwardable blurb here and our shared calendly is here!
We’re considering hiring a fall intern. Does anyone have strong opinions on important tradeoffs and factors to consider in that process?
Kindly,
Ben
P.S. Check out this demo that Isaac made to count jumping jacks!