I run an 8-person startup that uses AI agents. We don't let our agents run on vibes — here's the playbook system we created

This as-told-to composition is according to a conversation alongside Rachel Woods, CEO and cofounder of AMP. The following has been edited for dimension and clarity.

My archetypal authoritative role in AI was at Facebook. I started as an intern and afterward joined full-time in 2018. I worked on AI investigation and betterment for the ads scheme that's powered by AI.

I don't have an engineering flat or background; I studied administration data systems at The University of Texas at Austin. Part of why I think I got my function at Facebook was since I spent period construction my own side projects and instruction myself how to train AI models and do data analysis.

I remaining Facebook in 2020 to commencement my own endeavor in the vino industry, and following that business sold, I founded an AI company. I would need to twice the figure of group on my squad if we didn't use AI playbooks.

If you're being thoughtful concerning environment up your AI operations, it's no distinct than construction a squad that you can trust.

That's item a lot of group are missing: how to really create these systems things you can obtain your eyes off and not keep yourself up at night in the process.

I've been an entrepreneur since the age of 15

My archetypal endeavor was a photography business, and I was too young to equal run to encounter my clients. I would continually capture rides alongside people, and my parents hated this.

I got into tech since in college, I was exposed to colleagues doing data analytics and operating alongside coding and AI at summer internships. That led me to appearance online and try to obtain connected classes, and I started to see that if you interrupt a endeavor issue into small adequate pieces, you can get AI to grip parts of it: predicting, sorting, generating.

Later, whenever I was applying to Facebook, I used AI and data inspection to build a study of coffee shops in Austin that were small apt to be too occupied for me to study at. I think that project helped me district my role.

After I remaining Facebook, I started a vino endeavor called Vinebase in 2020; we built a tech phase that used AI to assistance winemakers market their vino online and oversee the digital flank of their business.

Because this was before AI became additional mainstream, we didn't market it as AI. We called it our "auto magic" system. I think that if we had stated it was AI, it could've been a turnoff for our customers. Vinebase was acquired by Corksy in autumn 2022.

I use my playbooks and AI agents to substitute additional staff

My sister and I started AMP, initially called The AI Exchange, before ChatGPT was released in November 2022, since I was using AI in Vinebase and saw how the innovation was progressing. ChatGPT came out concerning a duration afterward and started really accelerating the AI timeline, so we went all in.

We have four full-time employees and four part-time contractors correct now, but we'd need twice if we weren't operating as much as we are on AI. Given the company's current size, we'd need at smallest four to five additional people, including a revenue and promotion manager, client assistance specialist, executive assistant, and operations coordinator.

The way we construction these systems is according to item we call playbooking. Every division in the business has its own system. Each scheme is a gathering of playbooks alongside AI agents that can run independently or together.

How we build our AI playbooks

To build an AI playbook, choice one part of activity you do complete and over, and compose downward what starts it, what goes into the task, and what should arrive out at the end. Next, compose the steps the way you'd train a new hire: all step, the decisions in it, and what "good" looks like.

Start operating it alongside AI, and during you watch, difference the outcome to how you would've done it. Then, fix the playbook, not the output, until you rely it to run without you.

I continually think through how I desire a playbook to activity and outline it by hand. Whenever you're making an AI playbook, you desire to own the thinking and the logic. Otherwise, you can autumn into the trap of including item that you don't really understand, and afterward you're not going to be capable to rely that scheme anyway.

We have an AI agent operating our revenue playbook to assessment client calls and justify they encounter our standards, fairly than having a revenue director assessment them manually. Also, alternatively of revenue squad members needing to ask their director questions, they fair ask the playbook, which afterward pulls the answers from our guide for that team.

There are two types of AI expenses we have in the business. The archetypal is the subscriptions to Claude and ChatGPT for all individual on the team. The another is since the bulk of our operations are these AI playbooks that are operating in the backdrop without us being the ones starting or triggering it. In September, we spent $8,510 on Claude Code to run our playbooks in the background, which is billed by use and is distinct from the team's monthly Claude and ChatGPT subscription costs.

Own the playbook, rent the tech

We say "own the playbook, rent the tech," which method that if you are creating really fine playbooks for your AI to follow, it really doesn't matter which tool you're operating it in.

Right now, we have a setup that's operating alongside Claude Code in the background. But I could run all of these playbooks in ChatGPT, Grok, Gemini, or Microsoft Copilot.

We try to evade a lot of the risks that group are concerned concerning alongside AI going rogue by creating these detailed outlines. We don't fair desire the agents to run on vibes. There motionless needs to be a individual to decide and scheme how the activity gets done; AI fair speeds up everything else. That's the job safety we have in this AI era.

Start small, and get everyone involved

A few of us on the squad who are additional advanced alongside AI own the bigger, team-wide playbooks that the endeavor depends on, but everyone playbooks their own individual workflow, and afterward we use that as the evaluation dirt to alumnus things into becoming additional part of authoritative systems.

My direction would be to fair commencement alongside item small and really easy, since formerly you get your archetypal win and see how AI can do that repetitive activity for you, it's benevolent of akin a snowball.

Do you have a narrative to portion concerning using AI to run your small team? Contact this editor, Agnes Applegate, at [email protected].