Capture terminology, decisions, and lessons learned in the Knowledge Wiki so TestFactors’ AI gets smarter and more accurate with every release.
The Knowledge Wiki is the secret weapon of mature teams. A well-maintained wiki cuts AI generation errors by ~60% and dramatically reduces the number of clarification questions on new workbooks.
Every HR project carries a mountain of context that lives only in people’s heads: which fields are required for our subsidiaries, what naming conventions we use, why we always do X before Y, what last release’s mistake was, which managers approve which moves.The Knowledge Wiki is where that context goes — in plain English. Once it’s there:
Scenario generation uses it to propose more accurate categories and scenarios
Test case generation uses it to fill in real data instead of placeholders
FactorBot uses it to answer team questions with project-specific knowledge
Clarification questions drop dramatically because the AI already knows your conventions
New team members read it to onboard fast
Three layers of knowledge:
Layer
What it is
Examples
Wiki pages
Long-form articles written by humans
”How we handle US/Canada cross-border transfers”, “Picklist convention for Job Codes”
Uploaded documents
PDFs, Word docs, reference guides
SF release notes, vendor implementation docs, internal HR policies
Extracted facts
Auto-extracted snippets from workbooks and answered clarifications
PDFs, Word docs, and other reference materials work the same way — TestFactors extracts and indexes their content.
1
Click Upload document
In the Knowledge tab → Documents sub-tab → Upload.
2
Drag-and-drop
Drop one or more files. Accepted formats:
PDF (text and OCR for image PDFs)
DOCX
TXT, MD
HTML
Excel sheets that are reference material, not workbooks (e.g. picklist masters)
Max 25 MB per file.
3
Tag and submit
Add tags (same as for wiki pages). Click Upload.Processing happens in the background — usually 30 seconds to 3 minutes per document. You’ll get a notification when it’s done.
Upload dialog with two PDFs in the queue and a tag input.
When the AI generates scenarios or test cases and runs into something it doesn’t know about your project, it adds a question to the Open Questions queue instead of guessing. This is gold for filling wiki gaps.
Open Questions tab with 4 questions, each linking to where the gap was detected.
1
Open the Open Questions tab
Knowledge tab → Open Questions.
2
Pick a question
Each question shows:
What the AI was trying to do
What it couldn’t figure out
What it currently assumed (and how confident it was)
A direct link to Answer this (creates a wiki page draft)
3
Answer by writing a wiki page
Click Answer this. A wiki page draft opens, pre-titled with the question and pre-filled with the AI’s current assumption.Edit the answer to be correct. Save.The AI immediately picks up the new knowledge and may re-generate any affected scenarios/test cases.
Some answers are conventions that should apply project-wide (or even org-wide) — not just one-off knowledge.After answering a question, tick the Apply as rule checkbox:
Workbook scope — only this workbook
Project scope — every workbook in this project
Client scope — every workbook for this client
Organization scope — every workbook in your org (use sparingly; very high-leverage)
Rules silently apply to future generations — no question is raised, no clarification needed.
Answer form with the 'Apply as rule' checkbox and scope selector.
“We always use Position Code starting with ‘POS-’” is fine. “Because Payroll’s integration rejects anything else” is much better — future readers understand the constraint.
One topic per page
A 50-page wiki where each page covers one thing well beats a 5-page wiki where each page covers ten things badly.
Use examples generously
“For US: use ‘EXEMPT’ or ‘NON-EXEMPT’ (no other values)” — concrete, testable, AI-readable.
Promote to rules slowly
A new rule at org scope affects every future workbook — verify the answer on 2–3 workbooks before promoting widely.
Answer Open Questions weekly
Make it a 15-minute weekly habit. Compounds dramatically — fewer questions next week, better generations the week after.
Link to source documents
“Per the linked SF release note, the new field X replaces Y in v2H 2025” — citations turn wiki into knowledge.
The Knowledge tab has a search bar at the top that searches:
Wiki page titles and bodies
Uploaded document content (including PDF text)
Tags
Answered Open Questions
Results show a snippet with your search term highlighted, the source type, and a “Last updated” timestamp.For complex queries, ask FactorBot — it can run knowledge searches with full conversational context.
Search results showing 3 wiki pages and 1 PDF matching 'position code'.
My wiki page doesn't appear in AI-generated scenarios
Three checks:
Is it tagged? Untagged pages are searchable but get less weight in AI generation. Add at least an SF Module tag.
Is the scope right? A wiki page at Project A scope won’t influence generation in Project B.
Has the AI seen it? AI generation indexes new wiki pages within a few minutes. If you wrote the page right before generating, wait 5 minutes and re-generate.
An uploaded PDF says 'Processing' for over an hour
Likely an OCR job on a scanned PDF. For very large or low-quality scans this can take 30+ minutes. If it’s still processing after 2 hours, try uploading again. If it consistently fails, the PDF may be corrupted — try opening in Acrobat and re-saving.
I want to revoke a rule I promoted
Knowledge → Rules sub-tab. Find the rule, click Disable. Future generations won’t apply it. Existing generations that used it stay intact.
My team is duplicating wiki pages
Two patterns help:
Search before you write — always check if a page exists first
Use the wiki’s tag-based browsing to find existing pages by SF module / topic
If duplicates exist, merge them: open the better one, copy over content from the worse one, then delete the duplicate.
A teammate's wiki page contains inaccurate info
Edit it directly (if you have edit permission) or add a comment via the page’s comment thread (@mention the author). All edits are versioned — nothing is lost.
I can't see a wiki page another teammate said exists
Either it’s scoped above your project (e.g. client-scope page when you’re not a client member) or you’re filtered to a different SF module. Check filters and ask the author for the scope.
