# Session 3 Handout: How Employees Use AI Today and Where People Decide

<p class="meta">FAST-15 AI Foundations for Every Employee · 7formations · Instructor: Wadood Chaudhary · All examples fictional · Suggested reading: Svekis, AI Literacy for All Employees (2026), chapter 5</p>

## Why this session matters

Knowing what AI is matters most when you decide how far to let it go. This session looks at how employees actually use AI, what research says about where it helps and where it hurts, and gives you a simple ladder for deciding who holds the decision.

## Key concepts

### 1. Five ways employees use AI

Content creation and communication; summarizing and research; data analysis and reporting; workflow automation; and decision support. Each needs a different level of human oversight: a draft email needs a read-through, a decision about a person needs a named decider.

### 2. What the research says

- In a large customer-support study (NBER Working Paper 31161, later published in the *Quarterly Journal of Economics*), an AI assistant raised issues resolved per hour by 14% on average and by 34% for newer agents.
- In a Harvard Business School study of consultants, AI helped on tasks inside its "jagged frontier," but on a task outside it, people using AI were about 19 percentage points less likely to get the right answer. AI can look capable where it is not.
- Pew Research Center (October 2025): about one in five U.S. workers said some of their work is done with AI.

### 3. The assistance ladder

| Rung | The AI... | Who decides |
| --- | --- | --- |
| 1. Inform | shows information | The person reading it |
| 2. Suggest | recommends options | A person chooses |
| 3. Draft | makes a first version | A person edits and owns it |
| 4. Act with approval | prepares an action | A person approves before it runs |
| 5. Act alone | acts within set limits | People set limits and review afterwards |

AI **agents**, tools that plan and carry out several steps across apps, sit at rungs 4 and 5. The higher the rung, the more limits and review you need.

### 4. Appropriate and inappropriate uses

Appropriate: drafting routine messages, summarizing approved documents, brainstorming, organizing information you can check. Inappropriate without strong controls: final decisions about people's jobs, pay, or benefits; anything involving restricted data in an unapproved tool; outputs nobody can check.

## Worked example

Pioneer Valley Mutual (fictional) uses AI to suggest payouts on small claims. That is rung 2. An adjuster reviews each suggestion, considers facts the model may not have (a photo, a phone call), and owns the decision. If the agency wanted to move to rung 5 for tiny claims, it would need a dollar limit, a sample review each week, and a way for customers to reach a person.

## After the course

- Finish your W4 chart and W5 card; keep W5 at your workstation.
- Next steps: FAST-01 (Applied AI Literacy at Work) to practice on your own tasks, FAST-16 (Working Well with AI) for prompting and checking, or FAST-17 (Responsible AI for All Staff).