Found this note from October 2017 Amy for follow

Found this note from October 2017: “Amy for following up and introductions, personal assistant with invisible AI.” At the time, Amy (from a different company called x.ai) lived in your inbox as a seemingly human assistant who quietly handled meeting scheduling. What I really meant was: what if you took that same “invisible AI” pattern and pointed it at where the money actually is in email, which is automated follow‑ups and intros, not just picking a time on a calendar 💌👇 In 2017, “AI assistant” mostly meant email scheduling and basic productivity. Amy would copy‑paste time slots, resolve conflicts, and keep the thread going, all while pretending to be a human assistant. The magic was that you never “used a tool,” you just CC’d a person and things happened. It felt like the closest thing to a real assistant you could get without hiring someone, even if it broke in all the predictable ways when context got messy or people replied in creative ways. Sales tools in 2017 lived in a different world. You had CRMs, outreach sequencers, template libraries, open‑tracking, and all that, but they were obviously “software.” You logged in, clicked buttons, exported CSVs. None of them felt like an assistant you could just CC. My note was basically asking: why isn’t there an Amy that lives in my inbox and says, “Great, I’ll follow up with them in 3 days,” or “I’ll introduce you to three more people like this,” and then just does it? Fast forward to 2025 and the landscape looks different. Invisible AI is still the right instinct, but it shows up in more opinionated ways. Superhuman is a good example. It is not pretending to be a human, but it quietly owns the follow‑up layer of your life. You hit a couple of keyboard shortcuts and suddenly every important thread boomerangs back at exactly the right time. You become the person who “never drops the ball,” even though it is really your email client running a very strict follow‑up schedule behind the scenes. With FirstQuadrant, we took that 2017 idea and leaned into the sales side even more. Instead of just “remind me,” it is closer to “own the pipeline.” The assistant figures out who to follow up with, when, with what kind of message, and in which channel. You still stay in the loop on the high‑judgment parts, like crafting a sensitive intro or handling a tricky negotiation, but the machine owns the boring repetition. That is the invisible part. It is not constantly asking you what to do. It is proposing concrete next actions and, in many cases, just doing them. The economics became clearer over time. Scheduling is valuable, but it is mostly about saving cognitive load and avoiding back‑and‑forth. Great, but you are not directly printing money. Follow‑ups and introductions, on the other hand, turn into revenue. If you close even one extra deal a month because your assistant never forgets that one critical email, the ROI is orders of magnitude higher. That is why the gravity shifted away from “AI for your calendar” to “AI for your pipeline.” Comparing Amy to Superhuman and FirstQuadrant, a few patterns stand out: • Amy tried to look like a human and hide the AI entirely. Super impressive for the time, but also fragile. • Superhuman leans into being a power tool. It is transparent, fast, and honest about what it does, while still feeling invisible in day‑to‑day use. You do not think “I am using AI” every time it saves you. • FirstQuadrant pushes toward an agent that makes decisions inside your workflow, but with enough structure, guardrails, and context from CRM and email that it does not need hand‑holding on every step. If I zoom out, a few general lessons about where AI assistants actually win: • They win where the work is repetitive but the stakes are high, like outbound sales or investor updates. • They win when they can sit on top of existing behavior, like email, instead of asking you to adopt a brand new ritual. • They lose when they over‑promise “general intelligence” without access to the right data, structure, and feedback loops. • They lose when they try to replace human judgment in delicate social situations, instead of narrowing the scope to things that can be templated, scheduled, or ranked. So the 2017 note was not just about “Amy, but for something else.” It was quietly pointing at a shift from productivity AI to revenue AI. From “help me find a meeting time” to “help me not lose revenue because I forgot to follow up.” And looking back from 2025, that evolution feels pretty natural. The assistant that finds you 3 extra hours a week is nice. The one that finds you 3 extra deals a month is a no‑brainer. Here’s the original note from 2017: https://github.com/AnandChowdhary/notes/blob/main/notes/2017/amy-for-following.md