What is PII, and how do you keep it out of AI?

1 of 6

What counts as PII?
Personal data can be obvious or hidden inside everyday business records
PII includes direct identifiers and also details that can point to a person when combined. If it can identify, contact, or single someone out, handle it carefully.
Direct identifiers: Full names, personal email addresses, phone numbers, home addresses, and dates of birth are classic examples.
Account & ID data: Customer numbers, employee IDs, account numbers, and document IDs can also count as sensitive identifying data.
Financial & HR details: Payroll details, tax forms, bank information, benefits records, and compensation data deserve extra caution.
Indirect identifiers: A job title, ZIP code, company, and birthday may identify someone when combined, even if no full name appears.
PII is not just one field: it is any information that can reasonably point back to a person.

Where does PII leak into AI?
Most exposure happens during normal work: copy, paste, upload, connect
Teams often share personal data with AI by accident while trying to move faster. The risk usually comes from convenience, not bad intent.
Prompts & chat boxes: Employees paste emails, notes, or customer messages directly into AI to summarize or rewrite them.
Uploaded files: Spreadsheets, contracts, resumes, support exports, and reports may contain hidden personal details in the data.
Connected tools: CRM, help desk, meeting transcript, or storage integrations can expose more data than the user realizes.
Screenshots & images: Screenshots can include names, phone numbers, invoices, dashboards, and other visible identifiers.
If the workflow feels easy, that is exactly when hidden personal data can slip in unnoticed.

How do you keep PII out of AI?
Minimize, redact, substitute, and use only approved tools
The safest habit is to reduce personal information before it ever reaches an AI system. Simple process changes go a long way.
Redact before you share: Remove names, emails, phone numbers, addresses, account details, and anything unnecessary for the task.
Use placeholders: Swap real details for labels like [CUSTOMER], [EMPLOYEE], or [CASE ID] so AI can still help without seeing the person.
Summarize, do not dump raw data: Give the model the situation, goal, and relevant facts instead of full records, transcripts, or exports.
Stick to approved tools: Use tools with business controls, known settings, and clear rules for retention, training, and access.
If the task works with less personal data, less is the better default.

What if AI really needs context?
Give AI only the minimum safe context needed to do the job
Sometimes AI is useful on real business work. When that happens, the goal is controlled context, not unrestricted access to raw personal data.
Start with the business goal: Define the task first so you know what information is truly necessary and what can be left out.
Pseudonymize where possible: Replace real names with IDs or role labels and keep the lookup key outside the AI workflow.
Keep source data in the system of record: Let CRM, HR, finance, or ticketing tools hold the original records instead of copying everything into AI.
Review outputs before they leave: Check summaries, drafts, and recommendations for accidental exposure before sharing them with customers or coworkers.
Use AI on the task, not on more identity data than the task truly requires.

Small business takeaways
You do not need a huge program to reduce PII risk in AI workflows
Keep it simple: teach your team what PII is, reduce what goes into AI, and use tools with clear controls.
Make a short ‘do not paste’ list: Names, personal emails, phone numbers, HR records, financial data, and customer identifiers should be on it.
Approve a small toolset: Choose which AI tools your team may use and which ones are off-limits for business data.
Give staff safe templates: Provide prompt templates with placeholders like [CUSTOMER] or [EMPLOYEE] so people do not improvise with raw data.
Check vendor settings: Know the retention, logging, admin, and training-related settings of the AI products you pay for.
Train and review regularly: A short recurring reminder beats a policy nobody reads. Reinforce examples from real daily work.
If your team cannot explain what personal data is entering AI, you do not have control yet.

Lesson done.

Up next

What is data retention, and who decides it?

Where does this fit your business? Get your AI plan

Swipe to see each slide.

The short answer

PII, or personally identifiable information, is any data that can point to a specific person, such as names, emails, phone numbers, account numbers or dates of birth, and even combinations of details that identify someone. To keep it out of AI, share only what the task needs: redact it, swap in placeholders like [CUSTOMER], and use only approved tools.

Three things to remember

01PII is more than names: IDs, account numbers, payroll details and combined clues can all identify someone.
02Leaks happen in normal work: pasted text, uploaded files, connected tools and screenshots.
03Redact first, use placeholders like [CUSTOMER], summarise instead of dumping records, and stick to approved tools.

Every slide, in words

01
What is PII and how do you keep it out of AI?
PII is personal information that can identify a real person, and it should not be dropped into AI casually
Personally identifiable information (PII) includes data that points to a specific person directly or indirectly. The first rule is simple: do not share more personal data with AI than the task truly needs.
PII is more than names: Email addresses, phone numbers, street addresses, account numbers, dates of birth, and employee or customer IDs can all count as PII.
AI does not need everything: Many tasks can be done with summaries, placeholders, or partial details instead of real personal information.
Minimize before you prompt: If the task works without personal data, remove it before pasting, uploading, or connecting anything to an AI tool.
Good AI hygiene starts with one habit: strip out personal data unless it is truly necessary.
02
What counts as PII?
Personal data can be obvious or hidden inside everyday business records
PII includes direct identifiers and also details that can point to a person when combined. If it can identify, contact, or single someone out, handle it carefully.
Direct identifiers: Full names, personal email addresses, phone numbers, home addresses, and dates of birth are classic examples.
Account & ID data: Customer numbers, employee IDs, account numbers, and document IDs can also count as sensitive identifying data.
Financial & HR details: Payroll details, tax forms, bank information, benefits records, and compensation data deserve extra caution.
Indirect identifiers: A job title, ZIP code, company, and birthday may identify someone when combined, even if no full name appears.
PII is not just one field: it is any information that can reasonably point back to a person.
03
Where does PII leak into AI?
Most exposure happens during normal work: copy, paste, upload, connect
Teams often share personal data with AI by accident while trying to move faster. The risk usually comes from convenience, not bad intent.
Prompts & chat boxes: Employees paste emails, notes, or customer messages directly into AI to summarize or rewrite them.
Uploaded files: Spreadsheets, contracts, resumes, support exports, and reports may contain hidden personal details in the data.
Connected tools: CRM, help desk, meeting transcript, or storage integrations can expose more data than the user realizes.
Screenshots & images: Screenshots can include names, phone numbers, invoices, dashboards, and other visible identifiers.
If the workflow feels easy, that is exactly when hidden personal data can slip in unnoticed.
04
How do you keep PII out of AI?
Minimize, redact, substitute, and use only approved tools
The safest habit is to reduce personal information before it ever reaches an AI system. Simple process changes go a long way.
Redact before you share: Remove names, emails, phone numbers, addresses, account details, and anything unnecessary for the task.
Use placeholders: Swap real details for labels like [CUSTOMER], [EMPLOYEE], or [CASE ID] so AI can still help without seeing the person.
Summarize, do not dump raw data: Give the model the situation, goal, and relevant facts instead of full records, transcripts, or exports.
Stick to approved tools: Use tools with business controls, known settings, and clear rules for retention, training, and access.
If the task works with less personal data, less is the better default.
05
What if AI really needs context?
Give AI only the minimum safe context needed to do the job
Sometimes AI is useful on real business work. When that happens, the goal is controlled context, not unrestricted access to raw personal data.
Start with the business goal: Define the task first so you know what information is truly necessary and what can be left out.
Pseudonymize where possible: Replace real names with IDs or role labels and keep the lookup key outside the AI workflow.
Keep source data in the system of record: Let CRM, HR, finance, or ticketing tools hold the original records instead of copying everything into AI.
Review outputs before they leave: Check summaries, drafts, and recommendations for accidental exposure before sharing them with customers or coworkers.
Use AI on the task, not on more identity data than the task truly requires.
06
Small business takeaways
You do not need a huge program to reduce PII risk in AI workflows
Keep it simple: teach your team what PII is, reduce what goes into AI, and use tools with clear controls.
Make a short ‘do not paste’ list: Names, personal emails, phone numbers, HR records, financial data, and customer identifiers should be on it.
Approve a small toolset: Choose which AI tools your team may use and which ones are off-limits for business data.
Give staff safe templates: Provide prompt templates with placeholders like [CUSTOMER] or [EMPLOYEE] so people do not improvise with raw data.
Check vendor settings: Know the retention, logging, admin, and training-related settings of the AI products you pay for.
Train and review regularly: A short recurring reminder beats a policy nobody reads. Reinforce examples from real daily work.
If your team cannot explain what personal data is entering AI, you do not have control yet.

More in part 3

Next step

Where does this fit your business?

Six quick questions about your business, then your AI plan: what to start with, what a person still checks, and the lessons that go with it. Or bring it straight to a free thirty-minute call.

Get your AI plan Book a free call