We are adopting AI aggressively, but responsibly. This handbook separates the tools that help us from the habits that hurt us.
Free tools like standard ChatGPT or Claude are "Public." Anything you type into them becomes their property for training. We must sanitize our inputs.
Engineers pasted proprietary semiconductor source code into ChatGPT to "optimize it." That code was absorbed into the model's training data.
Consequence: Internal BanCorporate lawyers warned staff after seeing ChatGPT generate text that closely mimicked internal Amazon data, suggesting previous leaks.
Consequence: Legal AlertPublicly available info or generic templates.
Internal drafts that can be stripped of specifics.
PII, Financials, Passwords, Secrets.
AI has weaponized social engineering. Trust nothing without verification.
Attackers can clone a Partner or CEO's voice with just 3 seconds of audio from a YouTube video or podcast.
Scenario:
"You receive a call from the CFO. It sounds exactly like him. He says he's in a meeting and needs you to wire $50k to a vendor immediately."
Hackers use AI to scrape your LinkedIn, find your recent conference attendance, and write a personalized email referencing it perfectly.
The AI Email:
"Hi [Name], great seeing you at the Fintech Summit in Austin last week. Here is the deck we discussed regarding the Q3 merger..."
AI generates probabilistic text, not facts. In 2023, lawyers were sanctioned for submitting a brief with fake case citations invented by ChatGPT. If you use AI for research, you must verify every single claim against a primary source.
The difference between a generic answer and a strategic asset is Context.
"Write a job description for a Project Manager."
Result: Generic, boring, identical to 1,000 other listings.
"Act as a Senior Recruiter in the Fintech sector. Write a job description for a Project Manager.
Context: The team is remote and fast-paced.
Tone: Exciting, challenging, avoiding corporate jargon.
Include: A section on 'Why this role is hard'."
Why the firm invests $360/year/user for Enterprise Copilot.
Public AI knows the internet. Copilot knows us.
Your inputs are discarded after the session. They are never used to train the model.
No copy-pasting. It lives inside your apps.
| Feature | Public AI (Free) | Enterprise Copilot |
|---|---|---|
| Data Privacy | They view/train on your data | Data never leaves tenant |
| Context | Limited to chat window | Reads OneNote, Loop, Emails, Chats |
| Security | None | Inherits M365 Permissions (RBAC) |
AI-generated assessment. Questions change every time.
Click below to ask Gemini to generate 5 scenario-based questions tailored to this guide's content.
Consulting AI Model...
Generating scenario-based questions...
Your results have been securely transmitted.
Connection failed.