hi bud buds,
last week, a new mit report went viral on ai twitter: 95% of gen ai pilots at companies fail to deliver meaningful impact.
here are the key takeaways you need to know:
buy > build: mit found most enterprises (especially in finance, healthcare, and other highly regulated sectors) are attempting to build their own gen ai tools. but the data says they shouldn’t: buying from specialized vendors had a 67% success rate vs. just 33% for internal builds
most roi in back-office: over 50% of gen ai budgets go to sales + marketing tools, but the biggest roi showed up in back-office automation, especially where companies replaced agencies or outsourcing. additionally, instead of mass layoffs, many companies are backfilling fewer roles when people leave
workflow integration is the real roi bottleneck: the limiting factor isn’t model capabilities; it’s employees not knowing how to integrate ai tools into their day-to-day work
what does this mean for pms?
copilot design > agentic design
ai’s impact on productivity will be slow + steady, not overnight
buy > build for internal ai tooling
when building ai features:
integrate into existing workflows
focus on automating tedious tasks > moonshot use cases
have a great week, y’all. i love you.
- dj
last week’s biggest product releases
personal faves
google announced an ai fitness coach for fitbit, with sleep tracking, custom routines, and remembered preferences
notion launched offline mode 🎉 been waiting for this one
meta added ai voice translations to facebook + instagram, letting creators easily translate their content into other languages. i remember mr beast mentioning one of his biggest growth hacks was hiring teams to translate his videos; cool to see meta productize this
nice job guys
youtube music launched tour alerts, user comments on albums + playlists, and fan loyalty badges. this is sick
tiktok added campus verification so college students can connect with classmates using their academic emails
google launched pixel journal, an ai journaling app for pixel phones
google launched a shortcut in google drive to vids, its ai video editing + creation tool
apple launched chatgpt config tools, letting enterprises better control how ai is used on their mac devices
other llm news
deepseek released v3.1, optimized to run on chinese-made chips instead of nvidia chips
openai launched chatgpt go in india, a $5/month premium plan
openai updated gpt-5 to be warmer + friendlier
idk tbh
google added natural language photo editing to google photos. imo, this is a weak use case for natural language. photo editing is extremely iterative, and it feels 10x faster to just use sliders / selectors than describe every tweak in words
grammarly launched a document-based interface with ai proofreading and ai citations. i’ve seen startups in this space popping off absolutely popping off, but i feel like google docs / notion will eventually kill all of them?
that’s it for this week; thanks for reading. if you enjoyed this issue, please consider sharing it with a friend so i can buy a country home in japan.
love, dj