6/ Lever 5: One agent across every customer touchpoint.
Most enterprises in 2024 had a separate vendor for each channel:
→ SMS: Twilio
→ Call center: Five9 or Genesys
→ Chat: Intercom or Zendesk
→ Cases: Salesforce Service Cloud
→ WhatsApp: a partner BSP
→ Email: Marketing Cloud
So 6 contracts, pricing models, data silos, teams etc.
They collapsed all of it into one agent.
They launched voice in October 2024 and within 12 months voice surpassed text as the largest interaction surface on the platform.
The GTM consequence is that one Sierra contract absorbs budget from 4-6 separate vendor line items.
4/ Lever 3: They only got paid when the AI worked.
They published a pricing manifesto in Dec-24:
AI agent resolves the case → Sierra gets paid a pre-negotiated rate.
Case escalates to a human → Sierra gets nothing.
To match Sierra's pricing incumbents would have to cannibalize the revenue base their public market cap is priced on.
"Business model transitions are harder than technology transitions. The revenue dips for a period as they come back out. Any public company CEO will tell you that's easier said than done." — Bret
1/ The same week ChatGPT launched, @btaylor walked out of the Salesforce co-CEO office for the last time.
A few months later he called @claybavor. They'd known each other 15 years (a "monthly poker game that happens roughly twice a year").
Most AI startups in early 2023 were going horizontal while Bret and Clay went vertical and picked customer service.
I put together a fresh list of open jobs across some of the most interesting teams hiring in Greece
Check it out and if you’re hiring but not on the list, let me know
https://t.co/wP59Cj2ygZ
10 levers they used:
1. Cut 50% of features before launch.
2. Wedged through VCs to win the founders behind them.
3. Walked away from the meeting bot to win the meetings that mattered.
4. Ran frontier models even when the math didn't work.
5. Priced the team plan below the personal plan so users brought their team.
6. Shipped the team workspace once individuals were already hooked.
7. Anchored to a habit that already existed on the calendar.
8. Built from London for Silicon Valley.
9. Shipped MCP to become the memory layer for AI agents.
10. Opened APIs and repriced around the context layer.
Other bits I enjoyed:
→ Beautiful design = mind share = cheapest CAC.
→ 50% of triers still active at week 10 is the real moat.
→ Built an angel cap table like a launch list
→ Easy product trade-offs, hard life. Hard trade-offs, easy life.
7. Act II: The Ladder (October 2024 → March 2026)
The market got brutal: @OtterAI at $100M ARR, @Fireflies_ai hit $1B in a tender, @ReadAI_ raised $50M, Plaud at $250M annualized on AI hardware pendants etc.
And so the question shifted from "how do we get users" to "how do we keep them when notes commoditize?"