Preparing your data for AI

Everything the platform does for you — Chat answers, workflow builds, your FDE’s understanding of your business — draws on the documents you upload. Clean, current, well-chosen documents produce specific answers and systems that work on day one; a dump of stale files produces generic advice and rework. This guide covers what to upload, how to prepare it, and how to handle sensitive information.

The mechanics of uploading are covered in Upload your documents (during onboarding) and the Documents section (any time after). This guide is about what to put in.

Choose documents in priority order

You don’t need to upload everything — start with the files that best describe how your business operates today, and add more as the engagement surfaces gaps. Work down this list:

1. Process documents first. SOPs, process maps, and flowcharts that describe repeatable work. These are where automation has the biggest impact, so they’re what your FDE reads first.
2. Structure second. An org chart or a one-page who-does-what, so recommendations land with the right people.
3. Numbers third. Financial summaries, cost breakdowns, or budget documents relevant to your objective — they anchor the ROI conversation.
4. Customers fourth. Anonymized customer lists, survey results, or feedback reports, if your objective touches sales, marketing, or service.
5. Marketing materials last. Brochures, pitch decks, and brand guidelines help the AI match your voice, but they rarely change what gets built.

Check formats and size

The platform accepts:

Maximum file size is 25 MB per file. If a file is larger, split it into sections or ask your Advizr team to handle the import.

Where you have the choice, export PDFs from the source application rather than scanning paper — the platform extracts text content from your files, and a clean text layer beats a photographed page.

Name files so people can find them

The platform doesn’t require a naming convention, but your future self and your FDE both benefit from one. A pattern that works:

<what-it-is>-<scope>-<date>.<ext>

invoicing-sop-2026-05.pdf
org-chart-2026-06.png
q1-financial-summary-2026.xlsx
customer-feedback-survey-2026-04.csv

Two rules matter more than the exact pattern:

• Descriptive over clever. “invoicing-sop” beats “process_doc_final_v2_FINAL”.
• One current version per document. Upload the version that reflects reality today. If you upload the wrong file, upload the correct one and tell your FDE in Chat which to disregard.

Clean spreadsheets before uploading

Spreadsheets carry the most data and cause the most confusion. Five minutes of cleanup goes a long way:

• Put column headers in the first row, one header per column
• Keep one table per sheet — delete stale tabs and scratch areas
• Use consistent date and currency formats throughout
• Avoid merged cells and embedded charts where you can
• Remove rows that are notes-to-self rather than data

Handle sensitive information

Your documents are stored in a database provisioned for your organization alone, encrypted in transit and at rest, never shared with other clients, and never used to train AI models. Data and privacy covers the full picture. Within that boundary, two habits are still worth keeping:

• Anonymize where identity isn’t needed. If a customer list is going in as context for, say, a feedback analysis, names and emails usually add nothing — strip or pseudonymize them before export.
• Don’t over-strip. Systems are built against your real data, and systems built against fake data break against real data. If you’re unsure whether a field is needed, ask your FDE in Chat before redacting it.

If you need a document gone later, remove it from the Documents section; for complete deletion of processed data, your FDE can arrange it.

Keep documents current

Uploaded documents don’t update themselves. When an SOP changes, a report cycle closes, or a team restructures, upload the new version — the AI always works from the most recent files you’ve given it, so a stale SOP means confidently wrong answers. A quick monthly pass over your Documents library is enough for most businesses.

What Advizr does and what you do

Data preparation checklist

Data preparation — client checklist

Choose
[ ] 2-3 documents describing repeatable processes (SOPs, process maps)
[ ] Org chart or a one-page who-does-what
[ ] A recent financial summary tied to your objective
[ ] Anonymized customer data, if relevant to your objective

Clean
[ ] One current version per document; drafts and duplicates removed
[ ] Descriptive file names with dates (invoicing-sop-2026-05.pdf)
[ ] Spreadsheets: headers in row 1, one table per tab, consistent dates
[ ] Sensitive fields anonymized where identity isn't needed
[ ] Every file in a supported format and under 25 MB

Upload and maintain
[ ] Upload via the wizard or the Documents section; confirm "ready" status
[ ] Tell your FDE in Chat which documents matter most
[ ] Re-upload when a process or report changes
[ ] Remove mistaken uploads and flag them in Chat

Keep going

• Upload your documents — the upload flow, step by step
• Documents — managing your library after onboarding
• Data and privacy — where your data lives, who can see it, and how to export or delete it