Zapier: the basics for Optimists
Zapier connects apps so that an action in one triggers an action in another. You don't write code. You just define a trigger ("when this happens") and an action ("do this"). That's a Zap.
Example Zaps we use
- New lead form submission → enriched in HubSpot + notification in Slack
- Client check-in complete → follow-up task created in Teamwork
- CRM stage change → email draft triggered
- New transcript in Google Drive → AI AE Assistant kicks off
The goal for every Optimist: build or propose at least one automation. You don't have to know how to set it up yourself. You just have to notice a repetitive process and ask, "Could Zapier handle this?"
When you have an idea, post it in #ai-and-automations on Slack. Include: what the automation would do, what time or effort it would save, and what apps are involved. That's all we need to evaluate it.
The AI AE Assistant: a case study in "AI as an assistant"
The AI AE Assistant is the most advanced AI workflow at OO. It's a great example of how to think about AI at scale: not as something that replaces human review, but as something that handles first drafts of everything so the human can focus on judgment.
Here's how it works, step by step:
-
A Google Meet call is recorded.
Gemini auto-generates a transcript directly from the Google Meet recording.
-
A Google Apps Script detects the new transcript in Drive.
The script runs automatically when a new transcript file appears in the designated Drive folder.
-
The transcript is sent to the Claude API (Sonnet model).
Claude processes the full transcript using a custom system prompt that defines exactly what to generate.
-
Claude generates seven outputs simultaneously.
A Slack summary, individual task assignments for each team member, a follow-up email draft for the client, Teamwork tasks, appreciation messages for the team, Codex update recommendations based on client feedback, and a client health score.
-
The AE reviews everything and approves or edits before sending.
About 1 to 3 errors per run are typical. The AE catches them and fixes them. Then everything goes out. Total AE time: under 10 minutes for what used to take 45.
The cost
About 1 to 5 cents per meeting. That's the API cost. The time savings per meeting is somewhere between 30 and 60 minutes per AE. For a team that runs 20+ client calls a week, that math gets compelling fast.
This is what "AI as an assistant" looks like in practice. The AI drafts seven different outputs. The human reads, corrects, and approves. Nobody is replaced. Everyone saves time. That's the model.
Claude Code and the AI Creator Platform
Claude Code is a command-line tool that lets developers build full applications with AI assistance. Think of it as having an extremely capable coding partner who can write, debug, test, and iterate on entire apps.
At OO, we've used Claude Code to build:
- The AI AE Assistant
- The Agency Prep Checklist at tools.onlineoptimism.com/agencyprep
- Internal tools for the team
- Client-facing tools on the AI Creator Platform at create.onlineoptimism.com
💡 For non-developers: You don't need to know how to use Claude Code yourself. But you should know it exists and what it can build. If you have an idea for a tool that could help the team or a client, pitch it in #ai-and-automations. That's how the best tools get built.
Try it yourself
Pick one repetitive task you do at least once a week. It could be copying data between two tools, sending a standard follow-up email, or compiling a weekly report. Ask yourself:
- What's the trigger? (What causes you to do this task?)
- What's the action? (What do you actually do?)
- What apps are involved?
If you can answer those questions, you have an automation idea worth posting in #ai-and-automations. Eliza would love to hear it.
Common automation mistakes
- Automating without reviewing. The AI AE Assistant still requires human review. Every automation that sends external communication should have a human approval step.
- Not posting automation ideas in Slack. If you notice a repetitive task, tell someone. You're probably not the only one doing it.
- Thinking automation is only for developers. The best automation ideas come from the people doing the repetitive work. That might be you.
- Using Zapier to send client-facing communications without a review step. Automation speeds things up. It doesn't replace judgment. Keep humans in the loop.