Use this cheat sheet for CompTIA AI Essentials as a practical workplace AI review. The goal is not to memorize vendor slogans. The goal is to use AI assistants safely, prompt them clearly, evaluate their output, and know when not to use them.
AI assistant workflow
flowchart TD
T["Define the task"] --> C["Add context and constraints"]
C --> O["Specify output format"]
O --> R["Review for accuracy, privacy, bias, and policy"]
R --> I["Iterate with corrections or examples"]
I --> H["Human approves before business use"]
The strongest habit is to treat AI output as a draft, not an authority.
Prompt structure
| Prompt part |
What it does |
Weak version |
| Task |
Says what the assistant should do. |
“Help with this.” |
| Context |
Gives audience, goal, source material, and constraints. |
No audience or business situation. |
| Role or perspective |
Frames how the response should reason. |
Random expert persona with no useful boundary. |
| Output format |
Requests bullets, table, checklist, email, code, or summary. |
“Make it good.” |
| Quality criteria |
Tells the assistant what to optimize for. |
No criteria for correctness or usefulness. |
| Safety boundary |
Excludes private data, unsupported claims, or risky actions. |
No privacy or policy instruction. |
Hallucination checks
| If the answer contains… |
Do this |
| Facts, numbers, legal claims, medical claims, or security advice |
Verify against a trusted source before using. |
| A confident claim without evidence |
Ask for sources, then check whether the sources actually support it. |
| A summary of uploaded material |
Compare against the original document for omissions or distortions. |
| Code, formulas, or configuration |
Test in a safe environment before production use. |
| Strategic recommendations |
Ask for assumptions, alternatives, risks, and failure modes. |
Sensitive-data rule
Do not paste confidential, personal, regulated, customer, employee, credential, source-code, or proprietary data into an AI tool unless your organization explicitly allows that use in that tool.
| Data type |
Safer habit |
| Customer records |
Use anonymized examples or approved internal AI tools only. |
| Credentials and secrets |
Never paste them. Rotate if exposed. |
| Legal, HR, or medical data |
Follow policy and require human review. |
| Proprietary strategy |
Generalize the scenario and remove identifiers. |
| Production logs |
Redact tokens, names, addresses, account IDs, and internal URLs. |
AI agent boundaries
| Agent capability |
Boundary to set |
| Reads information |
Limit sources and verify retrieval quality. |
| Writes or edits content |
Review before publishing or sending. |
| Calls tools |
Scope permissions and require confirmation for high-impact actions. |
| Schedules or sends messages |
Add approval, audit trail, and rollback path. |
| Handles ambiguous requests |
Ask a clarifying question instead of guessing. |
Quick eliminations
| Question clue |
Avoid |
Prefer |
| “first prompt failed” |
abandon the tool |
revise context, constraints, and examples |
| “important decision” |
trust first output |
verify and involve accountable human reviewer |
| “sensitive company data” |
paste raw data into public AI |
anonymize or use approved internal tooling |
| “AI made up facts” |
ask it to be more confident |
verify sources and constrain the task |
| “agent can take action” |
broad tool access |
least privilege and confirmation for risky steps |