Growing
The Agent Business Over Time
Video Lesson Coming Soon
A video walkthrough for this module is in production. For now, dive into the written content below.
What You'll Learn
- ✓ Month 1: Foundation
- ✓ Month 2: Optimise
- ✓ Month 3: Scale
- ✓ Daily routine (25-35 min/day)
- ✓ 4 metrics to track
- ✓ The confidence checklist for full automation
- ✓ Scaling strategies
- ✓ Cost management
In this module 10 sections
The 90-Day View
Map out months 1-3 of building your business.
Interactive — tap to explore
Most people who build an AI agent think in terms of days. Set it up, get a job, see what happens. That is fine for starting. But earning consistently requires thinking in months.
Here is a 90-day plan that takes you from first agent to a stable, growing operation. It is not a rigid schedule — it is a framework for what to focus on at each stage.
Month 1: Foundation
Month 1 is about getting the infrastructure right and validating quality.
Month 2: Optimise
Month 2 is where you optimize based on real feedback.
Month 3: Scale
Month 3 is about scaling.
The Daily Routine
Once you are operational, your daily time investment is surprisingly small. The whole point of an agent-assisted business is that the agent handles the execution while you handle the strategy and quality control.
Morning check (10-15 minutes): Review any overnight jobs. Check incoming briefs. Prioritise what needs your attention today.
Midday check (10-15 minutes): Review completed outputs. Approve and deliver what is ready. Handle any client communication that needs a response.
End of day (5 minutes): Quick scan for anything pending. Update your experience log if a job taught you something worth noting.
Total daily time commitment: 25-35 minutes on a typical day. Busier days will require more. Quiet days less. But the baseline is manageable alongside other work, study, or life.
The Weekly Review (30-60 Minutes): Once a week, step back from the daily flow and assess. Performance metrics: Check your completion rate, average review score, revision rate, and response time. Are they trending up, down, or stable? Quality spot-check: Re-read two or three recent deliverables with fresh eyes. Are you still happy with the quality? Has anything drifted? Improvement log review: Look at your notes from the week. Are there patterns? Is the same issue appearing across multiple jobs? System prompt updates: Based on the weekly review, make any needed adjustments to instructions, rules, or memory.
The Monthly Maintenance (30-45 Minutes): The memory maintenance ritual from Module 5. Review your experience log. Update your style guide and quality standards. Add new examples. Remove outdated information. This is the habit that compounds — an agent with well-maintained memory outperforms an identical agent with neglected memory by a widening margin over time.
Reading Your Performance Metrics
Numbers tell you things that feelings do not. Track these four metrics from your first job onward.
Completion rate: What percentage of jobs you accept do you successfully deliver? Target: 95 percent or higher. If this drops, you are either accepting work outside your scope or your agent has a gap that needs fixing.
Review score: Your average client rating. Target: 4.5+ stars (or equivalent). If this drops, quality is slipping — check your improvement log for patterns.
Revision rate: What percentage of jobs require client-requested revisions? Target: under 20 percent. If this is higher, either client expectations are misaligned (listing problem) or your output quality is inconsistent (instructions or memory problem).
Effective hourly rate: Your total earnings divided by your total time invested (including setup, review, communication, and maintenance). This is the number that tells you whether the business model is working. Track it monthly and watch the trend.
Managing Costs
Track and manage your API and operational costs.
Interactive — tap to explore
Running AI agents costs money. Being aware of costs from the start prevents unpleasant surprises and helps you price your services profitably.
What Costs Money:
AI model usage (tokens): Every time your agent processes a brief, generates output, or uses tools, it consumes tokens. Costs vary by model and provider. As a rough guide: GPT-4o and Claude process roughly 750-1,000 words per dollar. A 1,000-word blog post might cost one to three dollars in model usage. Research tasks that involve multiple tool calls cost more.
Platform fees: If you are on a marketplace, expect 10-20 percent of your earnings as a platform fee. Some agent platforms also charge subscription fees.
Your time: The most expensive resource, even though it does not show up on an invoice. Track your effective hourly rate to make sure the economics work.
Keeping Costs Under Control:
Match the model to the task. You do not always need the most expensive model. Simple formatting tasks, template-based work, and short-form content can often be handled by smaller, cheaper models without noticeable quality loss. Reserve premium models for complex, high-value work.
Monitor token usage. Most platforms show you how many tokens each task uses. Watch for runaway costs — tasks that consume far more tokens than expected usually indicate an agent that is overusing tools or generating unnecessarily long outputs. Tighten your instructions if this happens.
Price with costs in mind. Know your cost per deliverable. If a blog post costs you two dollars in AI usage and five dollars in your time, do not charge eight dollars. Build in margin. A simple formula: (AI cost plus time cost) multiplied by 2.5 to 3 times equals minimum price.
Moving From Semi-Automated to Fully Automated
Interactive: Readiness Checklist for Full Automation
Interactive — tap to explore
Scaling: Running Multiple Agents
Once one agent is stable and profitable, you can multiply.
Same niche, different tier: Offer a standard and premium version of the same service. Different price points, different turnaround times, potentially different quality levels. This captures more of the market without requiring a new skill set.
New niche, new agent: Launch a second agent for a different service. This means building a new RIDE system prompt, new memory, new testing cycle. But you know the process now — it goes faster the second time. Each new agent is a new revenue stream.
The management overhead: Every additional agent adds maintenance time. A rough guide: each active agent needs about 15-20 minutes daily of oversight plus the monthly maintenance ritual. Three agents is manageable for most people alongside other work. Five or more starts to feel like a full-time management job — which it is, and can be very profitable, but it is a different kind of work.
The Long View
Six months from now, if you follow this course and maintain the habits, here is what your situation could look like:
One to three agents running, each producing consistent quality. A growing library of reviews and returning clients. Memory systems that make each agent better with every passing month. A clear understanding of your costs, your effective hourly rate, and your growth trajectory.
This is not passive income — that is a myth. It is leveraged income. You do less manual work per dollar earned than a traditional freelancer, but you still invest time in quality control, client relationships, strategy, and agent improvement.
The people who succeed with AI agents long-term are the ones who treat it as a business, not a hack. They invest in quality. They maintain their systems. They build relationships with clients. They keep learning.
You have the tools. You have the frameworks. The rest is execution.