AI APPLICATION PLAYBOOK Building a large AI application company is not easy but it’s becoming very clear what to do. Specifically, you need to: a. Automate a high value workflow b. For a large end market c. While showing strong market pull d. With a repeatable GTM motion e. And ample expansion opportunities If you can systematically prove that you can do each of the above, the world is yours. Inception / pre-seed: a, b Seed: a, b, c Series A: a-d + a bit of e Series B+: a-e
Perfectly outlined with first principles thinking Gokul Rajaram thank you for sharing! A company that shows the potential to run a through c well, would have gained enough strategic data advantages to play a strong expansion game, building upon the growth flywheel from its usage and micro-network effects. I imagine they make d and e relatively targetable. Doing a and b with high quality results seems to be the greyarea that’s oversold unvalidated puzzle, which if solved holds the potential for the greatest value unlock.
Great framework Gokul! One additional dimension I see emerging—especially as AI matures—is the need for unique data moats or defensibility, plus a clear story on responsible AI. Curious how you see these factors weighing in at Series A and beyond. Curious: Where do you see the biggest drop-off for founders—transitioning from workflow automation (a) to repeatable GTM (d), or in actually unlocking meaningful expansion (e)?
Gokul Rajaram if you replace the AI specific point (a), is the rest consistent with starting any company?
Gokul Rajaram, This seems to be same as SaaS?! But we know it is not. Can you highlight the differences from your perspective? I am also curious how the evaluation process has changed from investor’s perspective. Both combined would be a great article!
Gokul, great algorithm. In short, "Find a Need. Fill the Need. Learn & Scale" :-). That applies at every stage.
This framework is intriguing, Gokul. What insights do you have on identifying those "high value workflows" that can really drive sustainable growth in AI applications?
OBSERVATION: In just 24 months, the expert advice on how to build a successful new AI business has shifted from teaching people to use chatbots… to horizontal AI… to vertical AI… and now to process automations that really only exist in one company. Something tells me that building a successful AI automation company isn’t about building repeatable patterns yet ;-) The integrators will have the lead until the new patterns congeal. Best to follow their lead for now - don’t you think?
if not (automate_high_value_workflow and target_large_market): raise ValueError("Come back with a real use case.") if openai_or_big_tech_enters(): print("Pivoting...") return "Find niche, go deeper, avoid direct sunlight (aka general-purpose models)." if not market_pull: raise RuntimeError("No one wants it. Yet.")
Helping Tech Teams Hire the Top 1% of Global Engineering Talent — Vetted, Reliable, and Cost-Effective
1wDoes C (showing strong market pull) mean a repeatable way of getting paying customers?