AI Feature Errors & Troubleshooting
Resolving OpenAI API key issues, generation failures, usage limits, and AI quote problems
What You'll Learn
- Fix API key validation errors
- Resolve "AI usage limit reached" issues
- Troubleshoot failed quote generation
- Handle over-budget AI quotes
- Understand AI cost and usage tracking
"API Key Invalid" or Connection Test Fails
WorkCentral validates your OpenAI API key when you enter it. If the test fails, check these common causes.
How to Fix
- Key must start with
sk-and be at least 20 characters long - Copy the full key from platform.openai.com — partial keys won't work
- Check that your key hasn't been revoked or expired in your OpenAI dashboard
- OpenAI outage? WorkCentral tests against OpenAI's /v1/models endpoint — if that endpoint is down, the test will fail even with a valid key
- Only the organization owner can add or change the API key (Settings → AI & Quote Settings)
Pro Tip: After entering your key, WorkCentral auto-tests it after 1.5 seconds. Look for the green checkmark before saving.
"AI Usage Limit Reached"
Each organization has a monthly AI request limit that controls how many AI-powered actions you can take.
What to Know
- Default limit: 10 requests per month — both quote generation AND description improvements count
- The limit resets monthly — your counter goes back to zero each billing cycle
- The owner can increase the limit up to 1,000 requests/month in Settings → AI & Quote Settings
- Usage is shown as a progress bar on the settings page (e.g., "3 / 10 requests used")
Important: AI usage counts API calls to OpenAI, not cost. Even a single "Improve with AI" click on a description counts as one request.
Quote Generation Returns Error or Blank Results
If AI quote generation fails or returns empty results, the issue is usually with the input data.
How to Fix
- Project description must be at least 10 characters — provide enough detail for the AI to work with
- A client must be selected before generating
- Hourly rate must be at least $25
- A budget range must be selected
- If the AI returns invalid data, WorkCentral falls back to structured text parsing — you may see simplified results
- If all parsing fails, try regenerating with a more detailed project description
- The AI model defaults to gpt-4o-mini — if a model returns an error, WorkCentral automatically falls back to it
"Generate with AI" Button Is Grayed Out
The button only enables when all required conditions are met. Check each of these:
All Must Be True
AI Quote Is Over Budget
After generation, WorkCentral validates the total against your selected budget range and shows a color-coded indicator.
How to Fix
- Manually edit line items — adjust hours or rates on any AI-generated line item
- Regenerate with a smaller budget range or a higher hourly rate to get fewer hours
Good to know: AI aims for 80% of your max budget to leave room for scope adjustments. If you need the quote closer to the maximum, edit the line items after generation.
"Improve with AI" Not Working on Descriptions
The "Improve with AI" button on the project description field uses AI to polish your text into professional prose.
How to Fix
- A valid OpenAI API key must be configured in Settings → AI & Quote Settings
- The description must have content — there's nothing to improve if the field is empty
- Counts against your monthly AI usage limit — check that you have remaining requests
- If the AI returns identical text, it's treated as a failure and the original is preserved
- You can always "Revert to Original" after an AI improvement if you prefer your original version
AI Task Breakdown Not Generating Subtasks
When a quote is converted to a project, WorkCentral can use AI to break complex line items into multiple subtasks. If this isn't happening, check these settings.
How to Fix
- Toggle "AI Task Breakdown" ON in Settings → AI & Quote Settings
- An OpenAI API key must be configured — AI features require a valid key
- Task breakdown is automatic — it happens when a client accepts a quote, not as a manual action
- If AI fails, the system falls back to creating one task per line item (no subtasks)
- The AI needs at least 2 valid subtasks per line item — fewer results also trigger the one-task-per-item fallback
Note: AI task breakdown works best with detailed line item descriptions that are at least 30 characters long and have 3+ estimated hours. Simple, short line items are processed without AI regardless of the setting.
Still Having AI Issues?
If none of these solutions resolve your issue:
- Check your OpenAI account for billing issues or API key restrictions
- Try deleting and re-entering your API key in Settings → AI & Quote Settings
- Visit our contact page to submit a support request