Agency AI Coordination Playbook

You build an elaborate lead scoring system. After launch, you discover that 90% of leads come through one channel and a simple FIFO queue would have worked. Two weeks of build time wasted.

The team has 30 agents. 8 of them have overlapping responsibilities. 5 of them run but nobody reads their output. The maintenance burden exceeds the value. Debugging takes hours because nobody knows which agent is responsible for what.

An agent remembers that Client X is on a $5K/month plan. But the config file was updated to $8K/month last week. The agent reports inaccurate revenue numbers for 2 weeks until someone notices.

The human call center manager leaves. Nobody monitors daily performance for 3 weeks while you hire a replacement. Speed to lead doubles. Appointment rates drop 15%. You lose $20K in potential revenue.

A new lead comes in at 10:02 AM. The first call attempt happens at 2:15 PM. By then, the lead has talked to 3 competitors. Your close rate on this lead drops from 30% to 5%.

The agency finds out a client is leaving when they get the cancellation email. All the warning signs were there: no calls for 6 weeks, emails taking 5 days to reply, budget questions. Nobody was watching the pattern.

A guarantee client's Google Ads are performing well but their call center conversion dropped 40%. The ad performance report looks green. The call center report is on a different page. Nobody connects the dots for 2 weeks.

A $15K/month client has gone silent for 10 days. Everyone assumes someone else is handling it. The client is actually evaluating competitors. By the time someone notices, the RFP is out.

The sales agent sends an upsell proposal to a client who is unhappy with current performance and considering leaving. The client feels tone-deaf and accelerates their departure.

Every agent-to-agent handoff requires David to be in the middle. The sales agent finds a churn risk but cannot tell the retention agent directly. It sits in a report until David reads it 3 days later. The client has already left.

Five agents all query the Slack API independently. Rate limits hit. Data is inconsistent because each agent queried at a different time. One agent sees a message the others missed.

The morning briefing includes 200 Slack messages from 40 channels. Most are irrelevant. David stops reading briefings because the signal-to-noise ratio is too low.

The calendar scanner failed silently 2 days ago. The briefing keeps reporting "no meetings today" because it reads the stale file. David misses a client call.

Every small account triggers alerts constantly because the dollar thresholds are set for large accounts. Alert fatigue sets in. When a real overspend happens on a large account, nobody notices because alerts are ignored.

The pacing agent fires an overspend alert for a client that is offboarding next week. The team wastes 30 minutes investigating a non-issue. This happens 3 times a week. Trust in alerts erodes.

The portfolio "average CPL" is $45, which looks fine. But the median is $28, and one client at $180 CPL is dragging the average up. The expensive client's problem is invisible in the aggregate.

The sales agent and the retention agent both pull client spend data independently. They use different date ranges. Their reports contradict each other. The founder does not know which number is right.

An overspend alert fires. Nobody responds. The alert fires again the next day. And the next. After 2 weeks, someone notices the client burned through their monthly budget in 10 days.

A new sales agent sends 50 cold emails with a pricing error. No human reviewed them. The mistake is discovered when prospects reply with confusion. Brand damage is done.

The sales agent sends a reactivation email to a former client who left on bad terms. The agent does not know the relationship history. The email reopens a wound and generates a complaint.

A new project manager agent is deployed with full Accelo write access on day one. It creates duplicate tasks, assigns work to the wrong people, and marks projects complete that are not. Cleanup takes longer than the agent was supposed to save.

The founder corrects the same mistake 5 times across 5 sessions. Each session starts fresh. The agent has no memory of being corrected. The founder gets frustrated and stops trusting agents.

The team adopts 3 new frameworks in a month because they looked promising. None are properly integrated. The existing tools were fine. Six weeks of work is wasted on horizontal moves disguised as improvements.

The team adds advanced features (agent-to-agent messaging, autonomous outreach) while basic reliability (data accuracy, consistent formatting) is still broken. The foundation crumbles under the weight of complexity.

The founder is 60% billable and proud of it. Revenue looks good. But zero time went into sales, strategy, or agent improvement for 3 months. The pipeline is empty. When current clients churn, there is nothing to replace them.

A designer has 7 overdue tasks. Nobody knows because the tasks are scattered across Todoist, Slack, and Accelo. A client deliverable is 2 weeks late. The founder finds out when the client complains.

Each agent sends its own report at different times. The founder reads 8 separate updates, tries to mentally compile them, misses a conflict between the calendar and the task list. A client meeting overlaps with an internal deadline nobody caught.

Every day has meetings. Deep work happens at 9pm. Strategy work happens on weekends. The founder burns out in 6 months. The business stalls because the person who makes decisions never has time to think.

All agent outputs look the same. You cannot tell which agent produced a report at a glance. Debugging takes 3x longer because you are mentally mapping names to roles every time.

Two agents both think they own client communication. One sends a follow-up email while the other schedules a call. Client gets confused, trust drops. Or: nobody owns it because "that is the other agent's job."

You update the sales agent's prompt and suddenly the project manager stops including budget context in reports. Shared state or shared prompts created invisible coupling.

The sales agent sends a prospecting email to a competitor of your biggest client. The client finds out and questions your loyalty. Or: the agent emails someone the founder knows personally who already declined.

The CRM shows $200K in pipeline. Half of it is dead deals nobody closed out. The founder makes hiring decisions based on a pipeline that does not exist. Actual pipeline is $80K.

A proposal is signed on Friday. The sales agent updates the CRM. But nobody tells accounting. The client starts work on Monday. Three weeks later, the first invoice goes out late. The client is annoyed.