# Session 1 Handout: AI in Your Industry, Honestly

<p class="meta">FAST-07 Industry-Specific AI Applications for Massachusetts Sectors · 7formations · Instructor: Wadood Chaudhary · All examples fictional · Training, not legal advice</p>

## Why this session matters

Every sector is hearing big promises about AI. Some uses are already routine and proven, some are promising but unproven, and some are mostly marketing. If you can tell them apart, you spend time and money on what works and you can push back on claims that would put customers, patients, or crews at risk. This session gives you a sorting method, the data rules for your sector, and your first hands-on sector task.

## Key concepts

### 1. Mature, emerging, hype

| Pile | What it looks like | Example |
| --- | --- | --- |
| **Mature** | Widely used; evidence from comparable businesses that it works | Spam filtering; route optimization at regional carriers; drafting emails with human review |
| **Emerging** | Real products and pilots; results still being studied | Ambient documentation in clinics; AI clash detection on construction drawings |
| **Hype** | Big or absolute claims, little or no evidence | "Eliminates all downtime"; "replaces your adjusters" |

**The evidence question:** *Who uses it, in a business like ours, with what measured result?* A demo shows that something can work once. A logo slide shows who bought it, not whether it worked.

Surveys back this caution. McKinsey's 2026 State of AI survey found most respondents report individual productivity gains, but only about a third attribute any enterprise-level financial impact to AI. Gains are real but uneven, so sort carefully.

### 2. Your sector's sensitive data

| Sector | Never in a class prompt | Rule behind it |
| --- | --- | --- |
| All | Name with SSN, license, or account or card number | G.L. c. 93H; 201 CMR 17.00 |
| Healthcare and life sciences | Any PHI (names, dates, chart numbers with health info) | HIPAA Privacy Rule; BAA required |
| Insurance and finance | Insured or customer identifiers, claim files | 93H; DOI Bulletin 2024-10 governance |
| Manufacturing | Customer drawings, specs, pricing | NDAs; confidentiality |
| Construction | Bids, drawings, client details | Contracts; confidentiality |
| Hospitality and retail | Guest payment or contact details | 93H; PCI rules |
| Logistics | Driver and consignee personal details, rates | 93H; contracts |

### 3. The sector task pattern

Every sector lab follows the same four steps: **messy input → prompt for your industry's format → check against the source and the standard → record what you fixed.** The check is what makes the output usable.

## Worked example: Bayside Precision Parts machine log

**Input (fictional):** five spindle-vibration alarms in three days, each reset by the operator; a note of "bearing noise?"; feed reduced to 80%; tool 12 broke on lot 4471 with 6 parts scrapped; next PM 10/15.

**AI draft (typical):** "Vibration issue resolved by operator. Tool 12 replaced. No further action needed before PM."

**After the check:**

```text
Equipment: Haas VF-4, cell 2
Events: spindle vibration alarm x5 (10/05-10/07), reset each time; not diagnosed.
  Operator reported possible bearing noise 10/06. Running at 80% feed (workaround).
  10/07 tool 12 broke on BP-2207 lot 4471; 6 parts scrapped.
Open issues: root cause of vibration; link to tool break not in log.
Next step: inspect spindle bearings before 10/15 PM; quality to review lot 4471.
```

*What the check caught:* "resolved" (it wasn't), "tool replaced" (not in the log), and the missing scrap count.

## Checklist: before you trust a sector output

- [ ] Input is fictional or sanitized; passes the shared-printer test.
- [ ] Prompt names the reader and your industry's format or template.
- [ ] Every status, number, and date checked against the source.
- [ ] Nothing upgraded ("reset" is not "fixed").
- [ ] Errors and fixes recorded on the lab sheet, with time.

## Do and don't

| Do | Don't |
| --- | --- |
| Ask for evidence from businesses like yours | Accept a demo as proof |
| Sort claims before spending money | Buy because a competitor did |
| Use the sector's own template | Let the model invent the format |
| Record what you fixed | Write "looked fine" |

## Try this before session 2

Write down one AI claim you have heard at work (no company names). Sort it and name the evidence that would move it to "mature." Bring it to the review.