Creative Assistant
The world's first AI design tool — Sawa, was acquired by Mailchimp in 2018. I was part of the core team. After a streamlined integration, it became the Creative Assistant.
AI-powered design assistant on a major online marketing platform – Mailchimp. The Creative Assistant scans your website, collects brand assets, and instantly generates designs for any of your marketing channels. It is a collection of cross-integrated services & APIs.
Role
As the design automation lead, I oversaw the development of the logic behind all the magic under the hood which enabled the automation of human design decisions.
The design engine is a very dynamic service, that “cooks from ingredients”:
1. content data (text & images)
2. brand data (logos, colors, fonts, buttons, personality)
3. channel & marketing objective data
Design engine is capable of generating up to 1500-2000 design options for each campaign request. The number of options depends on the amount of variables submitted. To ensure the fastest experience, the team has developed a prediction method that determines the best and most unique design option in advance.
Magic web-page scraping, creates your Brand Kit automatically in a few seconds from the URL provided
Design Engine, 32 designs instantly, beautifully & on-brand
Brand Kit as a control panel
Magic instant resizing
🎾 Teamwork is magic! Building a great team is key to success, when with the feeling of fearlessness and confidence in each other we can get incredible stuff done. (And the teamwork won’t ever be replaced by AI 😉). I love people.
💎 Aim to build great staff, not just the MVP. The first release is the most important one, try to include as much “WOW+quality” as possible. The second release may be easily reprioritized or propounded, you never know.
⚖️ Big ideas should be balanced with usability improvements, and this is a very important balance.
🎙 Storytelling is a superpower I definitely want to master! Excellent communication will take you further than anything else, but it is also important to stay true to yourself.
🖇 I did my best to work as a “Glue” connecting pieces of the puzzle together and making it work, being a bridge between design, data science and development. But the “glue culture” is even better, when everyone is encouraged to act as a glue, learn from each other, connect with whoever, start channels, arrange quick calls, and share updates, keeping the team on one page and feeling ownership.
👂 Talking to users should be as natural as a monthly/weekly team sync. “Superfans” and “Superhaters” user clubs is a really really good idea.
🍜 The research showed that there is a gap between designs after scrape and the capabilities of the design engine. The design engine needs good ingredients and nice recipes to cool great stuff. Scrape often doesn’t provide great ingredients and after switching to the best and the most different pick through prediction the service became x10 fasted but lost in the quality of recipes.
