Things break. That’s not a failure of your system. That’s a fact of technology. The difference between frustration and resilience is knowing how to diagnose and fix the most common problems.
This article is your troubleshooting guide. Bookmark it. You’ll come back to it. Not because your system is fragile, but because even solid systems occasionally need a tune-up.
Problem: “My AI Isn’t Responding”
This is the most common issue and usually the simplest to fix.
Check your internet connection. AI tools need internet. Open a website. If it loads, your internet works. If not, fix your connection first.
Check the service status. AI providers have occasional outages. Google “[your AI provider] status” to see if the service is down. If it is, wait. Nothing you can do on your end.
Check your API key. If you’re using Claude Code or any API-based tool, your API key might have expired, been revoked, or hit its usage limit. Log into your provider’s console and check your key status.
Restart the tool. Close your AI app or terminal session and reopen it. This fixes a surprising number of issues, especially after your computer has been asleep for a long time.
Check your account. If you’re on a paid plan, make sure your payment method is current. Expired credit cards cause service interruptions.
Problem: “The AI Output Is Wrong or Low Quality”
When your AI produces garbage, the problem is almost always the input, not the AI.
Your prompt is too vague. “Help me with my email” gives the AI nothing to work with. “Draft a response to this email from my client Tom. He’s asking about the project timeline. The project is on track for June 15 delivery. Keep it professional but warm.” That gives the AI everything it needs.
Your context is stale. If your AI context document hasn’t been updated in months, it’s working with outdated information about your preferences, your role, and your goals. Update it quarterly (Article 6).
The conversation is too long. AI tools have context limits. After a very long conversation, the AI might “forget” things discussed earlier. Start a new conversation for new topics. Don’t try to handle everything in one endless chat session.
You’re asking for something outside its strengths. AI is great at text, analysis, code, and organization. It’s not great at real-time data (stock prices, weather, breaking news) unless it has specific tools for that. Know the limits (Article 8) and work within them.
Problem: “My Automation Stopped Working”
Automations fail silently. Here’s how to diagnose.
Check the cron job (scheduled task). On Mac/Linux, run `crontab -l` to see your scheduled tasks. Make sure the entry is still there and the times are correct. On Windows, open Task Scheduler and verify your task is active.
Check the script manually. Run the automated script by hand. If it fails when you run it manually, you’ll see the error message. Common errors: file paths that changed, missing dependencies, or environment variables that aren’t set.
Check file permissions. This is a classic. Your script runs fine when you execute it, but the cron job fails because it runs in a different environment. File permissions, path variables, and environment settings can all differ between your terminal and the cron environment.
Check the logs. If your automation writes logs (and it should), read them. The error is usually right there. “File not found.” “Permission denied.” “API key invalid.” Specific, fixable problems.
Check disk space. If your computer’s drive is full, scripts that create files will fail. Run a disk space check: `df -h` on Mac/Linux or check Storage in Windows Settings.
Problem: “My Dashboard Shows Stale Data”
You open your dashboard and it’s showing yesterday’s data. Or last week’s.
Check the data source. Is the data file being updated? Look at the file’s modification date. If it hasn’t been modified since yesterday, the automation that updates it probably failed. See the automation troubleshooting section above.
Clear your browser cache. Browsers cache web pages to load them faster. Sometimes they cache too aggressively and show you an old version. Hard refresh: Cmd+Shift+R on Mac, Ctrl+Shift+R on Windows. Or open the page in an incognito/private window.
Check the service worker. If your dashboard uses a service worker for offline capability (Article 58), the service worker might be serving cached content. Update the cache version in your service worker file and redeploy.
Check the deploy. If your dashboard lives on a web server, make sure the latest version was actually uploaded. A failed deploy means the server still has the old files. Check your deploy script’s output for errors.
Problem: “Voice-to-Text Isn’t Accurate”
Your spoken words come out garbled on screen.
Reduce background noise. The most common cause of poor transcription. Move to a quieter location or use earbuds with a built-in microphone.
Speak at a natural pace. Don’t over-enunciate or speak unnaturally slowly. Modern speech recognition is trained on natural speech patterns.
Check your microphone. On your phone, make sure nothing is covering the microphone. On your computer, check System Preferences/Settings to confirm the right microphone is selected.
Use the right language setting. If your voice-to-text is set to a different language or dialect, accuracy drops. Make sure it matches how you actually speak.
Try a different transcription service. If the built-in dictation isn’t cutting it, try a dedicated service. Whisper (by OpenAI) is very accurate and available through various apps.
Problem: “File Permissions on the Server”
If your dashboard or data lives on a web server and you’re seeing 403 Forbidden errors, the problem is almost always file permissions.
Web servers need files to be readable (permission 644 for files, 755 for directories). When you upload files from your local computer, they sometimes arrive with restrictive permissions (600 or 700) that block web access.
The fix: After uploading files to your server, set permissions:
- Files: 644 (readable by everyone, writable only by you)
- Directories: 755 (accessible by everyone, writable only by you)
Build this into your deploy process so it happens automatically every time. Never upload files to a web server without checking permissions afterward.
Problem: “I’m Getting Rate Limited”
AI providers limit how many requests you can make per minute or per day.
Symptoms: Error messages mentioning “rate limit,” “too many requests,” or “429” status codes.
The fix: Slow down. Add pauses between automated requests. If you’re generating multiple audio files or processing many emails at once, add a 2 to 3 second delay between each request.
Long-term fix: Upgrade your API plan if you’re consistently hitting limits. Or optimize your workflows to make fewer, more efficient API calls. One well-crafted prompt that gets the right answer is better than five vague prompts that need revision.
Problem: “Something Broke and I Don’t Know What”
The worst kind of problem. Something isn’t working, but there’s no obvious error.
Step 1: What changed? Think about what you changed recently. A new script. A file move. An update. A configuration change. The thing that broke is usually related to the last thing you changed.
Step 2: Check the basics. Internet working? Services up? API keys valid? Disk space available? These cover 60% of mystery issues.
Step 3: Test components individually. If your morning briefing isn’t generating, test each piece. Can your AI access your calendar? Can it read your task data? Can it write to the output file? Can the file be deployed? Test each step to find where the chain breaks.
Step 4: Check your backups. If something got corrupted or accidentally deleted, restore from your backup (Article 57). This is why backups exist.
Step 5: Ask your AI. Describe the problem to your AI. “My morning briefing automation stopped working. Here’s the error I’m seeing: [paste error]. Here’s my script: [paste script]. What’s wrong?” AI is excellent at diagnosing technical issues when given specific information.
The Maintenance Mindset
Don’t wait for things to break. Schedule regular check-ups.
Weekly: Quick scan of your automations. Are they all running? Any error logs? Dashboard loading correctly?
Monthly: Test a backup restore. Update your AI context. Review your API usage. Check for any software updates that might affect your tools.
Quarterly: Full system review. Which workflows are you still using? Which can be improved? Any new tools worth evaluating? Update your documentation.
Problems are inevitable. Downtime is optional. With good troubleshooting habits and regular maintenance, your system stays reliable. And when something does break, you fix it in minutes instead of hours because you know exactly where to look.
[Listen to “Why AI?” on our homepage] [Book Your Free Intro Session]
—
Achievementoring helps regular people build AI-powered productivity systems through 1:1 coaching, self-paced membership content, and done-for-you setup services. Because the future of personal productivity isn’t about working harder. It’s about working with intelligence.
Want help building your own AI system? Book a free intro session and see it in action. Or browse all 10 coaching sessions to see the full program.