Every automation consultant will tell you the build cost. Most won't tell you what happens six months later when Twilio changes their pricing tier, your EHR pushes an API update, or your office manager changes the scheduling rules and nobody tells the automation.
Automations are not set-it-and-forget-it. They're more like cars — they need maintenance, and the maintenance cost depends on how complex the system is and how frequently your business changes.
What actually breaks
API changes are the most common cause. Third-party services update their APIs, deprecate endpoints, or change authentication flows. This isn't rare — it happens several times a year across a typical automation stack. Most changes are minor (a field name changes, a response format shifts), but if you're not watching for them, the automation silently fails or starts producing wrong results.
Edge cases at scale are the second biggest issue. An automation that works perfectly for 20 patients a day might start hitting rate limits at 60. A lead routing system that handled 50 leads a month breaks when a viral listing brings in 300. These aren't bugs — the automation was never built for that volume.
Business process changes are the third. Your practice adds a new insurance payer. Your real estate team hires agents who specialize in a new area. Your financial firm starts offering a new service. The automation doesn't know about any of this unless someone updates it.
What maintenance actually costs
For most workflow automations, ongoing maintenance runs $200-500/month on a retainer, or $150-250/hour for ad-hoc fixes. That covers monitoring, minor adjustments, and quick responses when something breaks.
For AI agent systems, maintenance is typically higher — $500-1,000/month — because the systems are more complex and the failure modes are less predictable. An agent that starts producing lower-quality output needs prompt tuning, not just a field mapping fix.
For MCP servers, maintenance is usually lower — $150-400/month — because the architecture is more stable once deployed. The main maintenance tasks are adding new tools/queries and updating authentication when provider tokens rotate.
The honest breakdown
I include 30 days of post-launch support with every project. During that window, issues get fixed at no additional cost. After that, you have three options:
- 1Monthly retainer — I monitor the system and handle fixes proactively. Best for business-critical automations where downtime costs real money.
- 2Ad-hoc support — You call when something breaks, I fix it. Cheaper month-to-month but you're reactive instead of proactive.
- 3In-house handoff — I document everything and train your team (or your IT person) to maintain it. Best for technically capable teams who want full ownership.
Most clients start with a retainer for the first 3-6 months and then move to ad-hoc or in-house once the automation has stabilized and the common edge cases have been handled.
How to minimize maintenance costs
Build with well-documented APIs. Avoid tools with a history of breaking changes. Use self-hosted infrastructure (n8n over cloud-only platforms) so you control the update timeline. Build monitoring into the automation itself — don't wait for a human to notice it's broken.
The cheapest automation to maintain is one that was built with maintenance in mind from day one. That means error handling, logging, alerting, and documentation baked into the original build — not bolted on after the first outage.
