Saas AI Coordination Playbook

A customer downgrades from Enterprise to Pro on the 15th. The billing agent charges the Enterprise rate on the 1st because the plan change did not propagate. The customer disputes the charge. The support agent issues a refund. But the billing agent charges Enterprise again next month because the root cause (sync failure) was never fixed.

A payment fails on a Tuesday. The billing agent sends emails. But the product agent independently revokes access on Wednesday because it detected an unpaid invoice. The customer was mid-presentation using the product. Their screen goes blank in front of their team. They had every intention of updating their card. Now they are humiliated and furious.

A customer upgrades from $49/mo to $99/mo on day 15. They expect to pay ~$50 for the remaining half-month. The billing agent charges $99 for the full next month plus a $50 proration they did not expect. The customer contacts support. The support agent cannot explain the math because they cannot see the billing agent's calculation. Three agents, zero coordination.

The metering agent counts 1.2M API calls. The billing agent calculates based on 1.4M due to a timezone mismatch in the aggregation window. The customer is overcharged $800. They catch it because they track their own usage. One billing error on a usage-based plan erodes all trust in the billing system permanently.

A customer hits their 50-seat limit. 12 invitations fail. They email support asking for "a few more seats." Support replies with the upgrade pricing page. The customer wanted 5 seats, not a tier upgrade. A sales-aware CS agent would have offered a seat expansion at the current tier price. Instead, the customer sees a 60% price increase and starts evaluating competitors.

An Enterprise customer stops using the advanced reporting feature they specifically asked for. Login frequency drops 50% over 3 weeks. Nobody notices because they are still paying. 60 days later, they send a cancellation email. The CSM calls in a panic. The customer says "we made the decision 6 weeks ago." The $48K/year account is gone.

An annual contract renewal comes up in 2 weeks. The CSM reaches out for the first time in 3 months. The customer says "actually, we have been meaning to talk to you about some issues." There is no time to fix anything. The customer downgrades or churns. A 90-day runway would have surfaced the issues early enough to address them.

A migration adds an index to a 500M row table. It locks writes for 8 minutes. During that window, 200 subscription renewals fail silently. The billing agent retries them an hour later, but 15 customers were double-charged because the retry logic did not check for partial writes. The migration itself worked perfectly. The coordination failure caused the outage.

A deploy goes out at 2 PM. Error rates double. The deploy agent auto-rolls back at 2:15 PM. But the support agent has already sent 30 "we are investigating" replies with no context. When the rollback completes, the support agent keeps sending the old template. Customers get conflicting messages.

The codebase has 347 feature flags. 200 of them are for features shipped 8+ months ago. Nobody knows which are safe to remove. New developers are afraid to touch them. The testing matrix is exponentially complex. A junior dev accidentally toggles an ancient flag in production and breaks billing for 10% of customers.

A deploy goes to 100% immediately. A subtle bug causes checkout failures for customers on a specific plan tier. It takes 45 minutes to detect because overall error rates only increased 3%. But 100% of Enterprise checkout attempts failed. A canary on 5% of traffic would have caught it in 2 minutes with zero customer impact.

Sales promises "full API access" during the demo. The customer's plan does not include API access. CS discovers this during the kickoff call when the customer asks about API documentation. CS either honors a promise they cannot afford or breaks trust on day one. Either outcome is bad.

A $2K/month Enterprise customer is routed to the self-serve onboarding flow because nobody configured the routing threshold. They get a generic welcome email and an in-app tooltip tour. The buyer who signed a $24K annual contract feels like a number. They escalate to their sales rep. "Is this really how you treat Enterprise customers?"

A customer signs up for a $149/month plan. They log in once, look around, get confused, and never come back. The onboarding email sequence fires on schedule ("Day 3: Set up your first workflow!") but the customer is not reading them. 30 days later, they cancel. The product worked fine. The onboarding failed to get them to value.

The product team deprecates API v1 with 90 days notice in the changelog. The CS agent is never told. 3 Enterprise customers rely on v1 for critical integrations. They discover the deprecation when their integration breaks. "We would have migrated if someone told us" is the universal customer response to surprise deprecations.

The sales agent lobbies hard for Feature X because 3 prospects asked for it. The product team builds it. Post-launch, nobody uses it. Meanwhile, the support agent has been logging 50 tickets/month about Feature Y which would have retained 10 churning accounts worth $120K ARR. The loudest channel won. The most valuable signal was ignored.

A database failover causes 15 minutes of read-only mode. The monitoring agent fires an alert but nobody classifies it. Support starts getting tickets but does not know the scope. The status page still shows "all systems operational." Enterprise customers discover the issue from their own monitoring and call their CSMs. The response is reactive, uncoordinated, and slow.

A customer requests data deletion. The production database is wiped. But their name and email remain in the analytics warehouse, the email marketing tool, and 3 support ticket transcripts. A GDPR audit discovers the incomplete deletion. The fine for non-compliance starts at 20M euros or 4% of global annual revenue.

A vulnerability is discovered in the file upload feature. The security team patches it in 4 hours. But nobody tells customers that their uploaded files may have been exposed. A customer discovers the vulnerability independently and reports it publicly. The company is now doing crisis communications for a vulnerability they already fixed, because they skipped the coordination step.

SOC 2 audit is in 30 days. The compliance team asks engineering for 12 months of deployment logs. Engineering has 9 months. The other 3 months were on a system that was decommissioned. Nobody exported the logs. The auditor flags a gap. The company either delays the audit or receives a qualified opinion. All because evidence collection was not built into the daily workflow.

A deploy introduces a bug in the CSV export feature. 80 customers file tickets. 80 support agents manually investigate. 80 responses are sent individually. Each takes 15 minutes. 20 hours of support time for a bug that could have been deflected with a single known-issue notice posted in the first 10 minutes.

The "integrations" feature area generates 200 tickets per month. Support hires two more agents to handle volume. Nobody tells the product team that the integration setup flow has a 40% failure rate. The product team thinks integrations are fine because they only look at completion metrics, not support ticket volume.

An Enterprise customer with $200K ARR submits a critical ticket. It enters the general queue. A junior support agent picks it up and sends a templated response. The customer's CTO emails the CEO directly. The ticket sat in a queue for 4 hours because the triage agent did not know the customer's tier or health score.