☄️ How to Work With An AI Teammate

When using AI for work, it’s tempting to drop a wall of text into an LLM and expect a perfect answer.

However, current models and AI agent systems still need guidance from you to deliver the output you actually want.

This is even more important when starting to work with an “AI teammate.”

An AI teammate is (typically) a multi-agent system that can tackle tasks, learn from feedback, and free you up to handle higher-leverage work.

Fortunately, there’s an easy parallel you can keep in mind for getting AI teammates up and running quickly.

Just think of it like onboarding a new (human) teammate.

If you onboarded a new (human) teammate, you wouldn’t expect them to be operating at 100% on day one. Many people get frustrated with AI output, rather than having productive collaboration with it.

Here’s a simple guide to get started working with an AI teammate:

1. Map the tasks you want to delegate to AI

2. Start small and give clear and specific instructions

3. Strengthen trust and responsibilities through verified wins

1. Map Your Workflow

Brain dump your entire workflow on one canvas. Do this in Miro, FigJam, or on a plain piece of paper. Be specific, from start to finish (i.e. final hand-off or deliverable).

The goal is to visualize your current workflow(s) in high resolution and understand what pieces can be delegated based on your AI teammates’ capabilities.

The process can look something like:

• Write out your daily/weekly/monthly recurring tasks. Examples:

  • Scheduling meetings

  • Updating CRM

  • Inbox management

• Break down each task into step-by-step sequences.

  • Example (start here and then get more detailed): Lead form submission → CRM entry → SDR qualify → AE hand-off

• Highlight pain points, like those that are repetitive, cause delays, or create errors.

  • Repeated keystrokes

  • Required approvals and sign-offs

  • Error-prone copy/paste

This helps you identify routine work for AI and spot unclear problems in the workflow.

You can even use AI to help with this process.

2. "Trust but Verify"

Treat delegation like onboarding a new hire: trust the AI but always verify the work for accuracy. This is a healthy start to your working relationship for both you and your AI teammate.

• Start with low-stakes tasks.

• Initially, review every AI-produced output carefully. Don’t expect it to be perfect. Consider the output from 3 angles.

  • Accuracy: Is the output correct and reliable?

  • Value: Did it tangibly save time or effort compared to manual work?

  • Weak Spots: What consistent errors or misunderstandings occur?

• Keep in mind that things that your AI teammate isn’t good at today can and will get better over time (i.e. more intelligence and faster learning).

3. Improve The Feedback Loop

The output you’ll get is only as good as the specificity of your feedback.

AI teammates can thrive when you explicitly define success and failure criteria.

Specifically:

• Provide direct examples of good output ("This concise summary is good because…").

• Clearly show mistakes with corrected versions next to the original ("Here's a miscalculation and how to fix it...").

• Share concrete do’s and don’ts ("Always include X, never add Y").

Early on, create an intentional process for documenting, sharing, and refining your feedback. Iterate until you have a good working cadence.

Bottom Line

McKinsey estimates that AI work automation can improve productivity by 3.4% annually. I think they’re sandbagging.

The earlier you weave AI into your daily workflow, the faster productivity and output start to compound.

Clever prompts can help, but being a good manager is a much more effective way to get the most out of your AI teammate.

Until next week,

David Lobo

Head of Growth, Workmate

P.S. At Workmate, we're obsessed with building an AI teammate you can truly trust and delegate to—starting with scheduling, and evolving toward even bigger workloads.

P.P.S. Curious, what tasks do you wish you could completely delegate to AI?