Every Friday I answer one question I actually get asked. This week’s is the one people ask right after “what does it cost?”: how much time does AI agent maintenance actually take once the thing is running?
I didn’t want to guess, so this morning I counted. Every agent in my stack files an Asana card when it finishes something or needs me. Since mid-June that adds up to 575 cards in my Inbox and 171 “please review” cards in Drafts. Then I pulled every card I’d closed by hand, meaning anything closed hours or weeks after it was filed, not the ones the agent closed itself in the same second.
The short answer: my time went to keys, not content
There were 17 cards I closed by hand. Here’s every one of them, grouped:
| What needed me | Cards |
|---|---|
| Social scheduler login expired, so posting stopped | 5 |
| Outreach email bounce rate over the safe line | 5 |
| Google analytics/search access needed re-pairing | 3 |
| Image compression hit its monthly quota | 1 |
| A scraping API key was blank | 1 |
| An outreach campaign needed creating by a human | 1 |
| A data field the next agent needed was empty | 1 |
That’s 9 about access, 6 about limits and 2 about plumbing. And there were zero about content: no “rewrite this post” and no “this caption is off.”
Meanwhile the Drafts column, the one built for me to review the writing, has 167 open cards. The oldest is 116 days old. Every one of those posts went live on schedule and the business kept running. So the review queue turned out to be a paper trail, not a gate.
One honest caveat: a closed card tells you what needed a human, not how many minutes it took. Some of those 17 got fixed quickly and closed weeks later in a cleanup sweep. So I won’t give you an hours-per-week figure I can’t back up. What I can tell you is where the hours go, and it isn’t where people expect.
Why AI agent maintenance is about keys, not writing
When you picture maintaining an AI agent, you picture babysitting its output. In practice the output is the stable part. A well-briefed agent writes the same quality on day 90 as on day 9.
⚡ GET THE AI EDGE
Weekly AI tips that actually save you time and money. No fluff, no hype — just what works.
What decays is everything around the agent. Logins expire on someone else’s schedule. Free tiers run out mid-month. Email lists go stale and start bouncing. Agents are bad at none of this. They just aren’t allowed to fix it, because fixing it means logging in as you.
That’s also why the running cost of an agent is the wrong number to obsess over. The real cost is the re-pairing work, and it arrives without warning.
The bigger time sink: finding the 17
Here’s the part I didn’t enjoy counting. Those 17 blockers sat in the same Inbox as 557 still-open cards, mostly status reports and FYIs. The urgent ones weren’t hard to fix. They were hard to see. A login that expires on a Tuesday gets fixed on a Friday because it was card 40 in a pile of reports.
If your agent can’t tell “I’m done” apart from “I’m stuck and only you can unstick me,” you’ll spend more time sorting cards than fixing anything.
What to do about it this week
- List every login your agents depend on. Note when each one expires and who re-pairs it. Put the renewal dates on your calendar before they turn into an outage.
- Set a usage alert at 80% of every quota. Image credits, email sends, API calls. Hitting the limit at 100% means it’s already a blocker.
- Split your agent’s messages into two channels. Status reports go to one place you skim weekly. Anything needing your hands goes somewhere that buzzes. One pile for both is how a 5-minute fix waits 5 days.
- Be honest about review. If you’re never going to approve 167 drafts, don’t design a system that pretends you will. Decide what genuinely needs eyes, gate only that, and let the rest ship.
If you’re still deciding whether to hand work over at all, start with what to automate first. If trust is what’s stopping you, these five questions are where I’d begin. And if you’re weighing an agent against a human assistant, the maintenance math above belongs next to what actually transfers in that swap.
Want someone to map your logins, quotas and alert channels before they bite? Book a strategy session and we’ll do it together.
See you next Friday. — Jon
📥 FREE: THE AI PLAYBOOK
The exact tools and workflows I use to run a one-person agency. 25 years of marketing experience distilled into an actionable guide. Yours free.
AI & Digital Marketing Strategist
With over 25 years in digital marketing, e-commerce, sales, and international business, Jon Jones is a seasoned entrepreneur dedicated to empowering small businesses through innovative AI solutions. His journey began in investment banking on Wall Street, guiding companies through capital acquisition and complex markets, and has expanded to include expertise in web development, brand design, product development, and courses and training.
Fluent in Mandarin and with a rich foundation in both finance and tech, Jon launched JonJones.ai to provide accessible, transformative tools for small businesses to streamline operations, generate leads, and build sustainable growth. His mastery of AI, automation, and digital strategy has enabled numerous businesses to bridge the gap between traditional practices and modern technology, ensuring they thrive in today’s digital landscape.
