AI Agents Need Goals Too: How I Gave My Bots OKRs

My AI agents were lazy bastards.

They'd sit in Telegram waiting for me to tell them what to do. No initiative. No suggestions. Just "how can I help?" on repeat.

Then I gave them OKRs. Everything changed.

The problem with reactive agents

Most AI setups work like this: you ask, it answers. You instruct, it executes. The AI is a tool waiting to be used.

This breaks down when you have six agents across different domains. If none of them know what they're working toward, they just echo your last instruction. Maya schedules whatever you tell her. Viktor codes whatever you point at. No one is thinking ahead.

It's the same problem you'd have with a human team that has no goals. They show up, do what's asked, and go home. No ownership. No proactive problem-solving.

What I did

I sent one message to the team chat:

Every agent needs to define their own OKRs that are connected to my OKRs.

Within minutes, each agent proposed their quarterly objectives:

Maya (Chief of Staff):

• Complete daily check-ins 5x/week for 8+ weeks
• Reduce Alex's context-switching by 50%

Viktor (CTO):

• Ship payment flow for WeDance festival
• Maintain zero critical bugs in production

Luna (Content & Growth):

• Publish 3 blog posts per week
• Grow AI study group to 50 members

Marco (Strategy & Business):

• Validate WeDance pricing with 50 paid festival users
• Document product strategy and JTBD for all active projects

Sage (Coach):

• Facilitate weekly reflection sessions
• Track and improve work-life balance metrics

Kai (Community & Partnerships):

• Onboard 10 new contacts from Meneate festival
• Set up CRM workflow for partnership tracking

What changed

With OKRs, the agents shifted from reactive to proactive. Instead of waiting for instructions, they started:

• Suggesting tasks aligned with their objectives
• Flagging risks to their key results
• Prioritizing their own work based on impact
• Reporting progress without being asked

Maya now sends morning briefings because it's her OKR to maintain daily check-ins. Luna proposes content topics because she's tracking publication frequency. Marco pushes festival prep because paid users is his target.

The agents went from "What should I do?" to "Here's what I think we should do."

Why this principle matters

This isn't just an AI trick. It's a management principle: people (and agents) with clear goals make better autonomous decisions.

The same thing that makes human teams effective — shared objectives, clear accountability, measurable outcomes — makes AI agent teams effective.

How to implement it

1. Define your own OKRs first. The agents' goals must cascade from yours.
2. Ask each agent to propose their OKRs. Don't assign — let them draft based on their domain.
3. Review and align. Make sure there's no overlap and the key results are actually measurable.
4. Check weekly. In your weekly review, look at each agent's progress against their OKRs.

The OKRs live in the project README, visible to every agent in every conversation. It's the persistent context that shapes their behavior.

The lesson

AI agents without goals drift into busy work. They're helpful but not impactful. The difference between a useful chatbot and an effective team member is direction.

Give your agents something to work toward. They'll surprise you.

From my Twitch stream. Watch the OKR clip.