AI loops automate revenue workflows without constant human prompts
🎧 PodShort
29 min squeezed to 2
AI SprinklerAS AI / ML New
Sarah Nooravi
CEO at OpenClaw
Full episode from Lenny's Podcast
Quotable Moments
Yes, there are tons of great use cases for loops and we're going to talk about how you can use those and how they can be beneficial, especially with software engineering, but there are some reasons why you wouldn't want to use loops and honestly, I still do a little prompting.
I really like to think about loops as designing workflows and designing jobs to be done for people, it just happens to be that you can put this intelligent agent against the loop and then it's ready to go.
If you do not write that loop well, or your validation criteria is too thin, guess what, your agent is going to burn tokens.
Key Insights
- Prompts are out, and loops are in, meaning AI agents should be able to prompt themselves through automation rather than relying on constant human input.
- A loop is fundamentally an autonomous process where an AI agent can kick off a task with a prompt or set of prompts on a recurring schedule until the job is completed.
- Loop automation can be triggered by internal lifecycle events (e.g., a tool call, session start) or external hooks (e.g., a webhook from an external session or an incoming email).
- A 'goal' is a distinct type of loop where an agent continuously works towards a defined outcome until that outcome is achieved, measured, validated, or the agent becomes blocked.
- To write an effective loop, it's crucial to think like a manager delegating a job to an employee, defining what needs to be done, when, and what constitutes successful completion.
- Worktrees are essential foundational elements for loops, serving to isolate an agent's work within a sandbox, ensuring consistency in execution, clean workspaces, and resolution of conflicts.
- Loops can become very expensive by burning through AI tokens rapidly if they are not well-defined, or if their validation criteria are too thin, leading to excessive compute time without clear outcomes.
- For goal-based loops specifically, it is critical to be very precise when writing the prompt, explicitly outlining the evaluation and success criteria to avoid unintended outcomes and wasted resources.
Metrics Mentioned
- 40 PRs (The speaker mentions having around 40 pull requests that need to be reviewed or merged, highlighting a backlog scenario that a daily aging PR review loop could address.)
RevBots.ai View:
- SaaS Hoarders can use loops to reduce manual workflows but risk cost overruns without tight controls.
- AI Sprinkler teams should adopt goal-based loops for measurable outcomes like lead scoring or churn analysis.
- ARM organizations will orchestrate loops across their stack for autonomous revenue workflows.
- All stages must monitor loop efficiency to avoid becoming a token-burning money pit.
Join The RevBots ARMy
The insider daily for Autonomous Revenue Masters.