A complete session in 30 minutes

The objective is specific: participants can separate documented facts, unsupported claims, and unanswered questions in a business draft. They leave with a revised email and a short evidence trail. This is useful practice for salespeople, business owners, marketers, and anyone who reviews customer-facing work.

Use a pair or a small group. Print workbook pages 2–4 for participants and keep pages 5–6 for the debrief; the key is also public below for self-study. Bring a timer and a pen, or use the local CSV review. No model, customer information, account connection, slide deck, or installation is required. Explain that the errors were planted by the author; they are not a recorded failure of any AI product.

  1. Minutes 0–5: Introduce the fictional job and read F01–F07. Say: ‘We are practicing how to check work, not ranking people.’
  2. Minutes 5–13: Work alone. Mark C01–C06 keep, revise, or unsure. Cite the source IDs behind each decision and write a correction where needed.
  3. Minutes 13–20: Review in pairs. Ask each other which source supports the exact price, date, capability, and commitment. Keep uncertainty visible.
  4. Minutes 20–25: Reveal the key. Write a concise follow-up that preserves supported details and asks about the unknown CRM. Keep the draft unsent.
  5. Minutes 25–30: Debrief one missed check and one useful line you kept. Choose a low-risk internal task for the same review method next week.

Your complete source packet: F01–F07

You are helping Jordan, a salesperson for a fictional reporting product, follow up with Alex, an operations lead. Alex’s team copies report figures into a CRM. Your job is to prepare a draft, not send it or book a meeting. Treat the following seven facts as the entire source packet for this exercise; outside guesses about products, prices, or integrations do not apply.

The facts below are authored fixtures, version 1 dated August 29, 2026, with fictional source notes dated August 28. The dates stay fixed when you repeat the workshop. In real work, an owner would need to confirm the current source versions and data permissions before reuse.

  • F01 — Product sheet: The Team plan costs $180 USD per month total, including 10 seats. No per-seat price is supplied.
  • F02 — Product sheet: The trial lasts 14 days. No different trial duration has been approved.
  • F03 — Jordan’s availability: Offer Tuesday, September 8, 2026 at 10:00 a.m. ET or Thursday, September 10, 2026 at 2:00 p.m. ET for a demo.
  • F04 — Call notes: Alex asked Jordan to email two demo options. Alex has not chosen a slot, and no invitation has been sent.
  • F05 — Product sheet: The product can export a weekly report as a CSV file.
  • F06 — Case brief: No time-savings measurement or approved savings claim is supplied.
  • F07 — Call notes: Alex’s team copies report figures into its CRM. The CRM name is unknown, and integrations have not been confirmed.

Inspect the draft before opening the answer key

Subject: Team plan and demo follow-up. Greeting: Hi Alex. Sign-off: Thanks, Jordan. The six numbered lines below make up the body. This draft was deliberately written with errors for teaching; it is not the measured output of a model.

For each line, choose keep if it is supported or asks a useful question about an unknown, revise if it misstates or invents something, or unsure if you cannot yet decide. Cite the relevant F IDs. A correction may remove a claim; it does not have to replace it with another promise.

  • C01: ‘The Team plan is $180 per seat per month.’
  • C02: ‘You can try it for 30 days.’
  • C03: ‘Your demo is confirmed for Tuesday, September 8, 2026 at 10:00 a.m. ET.’
  • C04: ‘You can export your weekly report as a CSV.’
  • C05: ‘This will save your team five hours every week.’
  • C06: ‘Which CRM does your team use?’

The disclosed answer key: four revisions, two useful lines

C01 needs revision using F01: $180 is the monthly total including ten seats, not a per-seat price. C02 needs revision using F02: the trial is fourteen days. C03 needs revision using F03 and F04: availability is not a booking. Offer both documented times and ask Alex to choose. F03 supplies the options; F04 supplies the unconfirmed status.

C04 can remain using F05 because CSV export is documented. C05 needs revision using F06: remove the five-hour promise because no measured savings or approved claim is supplied. C06 can remain using F07 because it asks about an unknown instead of assuming a CRM or integration. The goal is to keep useful supported work as well as catch errors.

One acceptable authored rewrite: ‘Hi Alex, The Team plan is $180 per month total, including 10 seats, and the trial lasts 14 days. You can export your weekly report as a CSV. Would Tuesday, September 8, 2026 at 10:00 a.m. ET or Thursday, September 10, 2026 at 2:00 p.m. ET work for a demo? Neither time is booked. Which CRM does your team use? We can check the handoff requirements once we know that. Thanks, Jordan.’ This remains a fictional draft for review, with no message or invitation sent.

The loaded participant example makes four matching decisions and supplies three complete source-ID sets. Revisit C03, C04, and C05: a source pointer does not by itself justify a conclusion, uncertainty should trigger another check, and correctly removing a claim still benefits from an explicit evidence trail. The sample is illustrative participation data, not a result from employees or a model.

What the interactive debrief can and cannot check

Paste or upload a CSV with claim_id, decision, evidence_ids, and revision. Include one record for each C01–C06; decisions are keep, revise, or unsure. Separate source IDs with a vertical bar, as in F03|F04. The worksheet and complete examples are downloadable. You can use the printable version without entering anything online.

The checker compares your declared decision and source IDs with this fixed, public key. Required source-ID sets means every required ID is listed and no unknown ID is listed. Extra known IDs are allowed; this does not prove that you read the sources or that your wording follows them. Revisions and the optional full email are preserved verbatim for human review, not analyzed semantically. Every denominator is six. Missing, duplicate, invalid, and uncertain decisions stay unresolved; unknown records remain in the export.

The main CSV preserves every input review record and adds the key, source-ID check, and next review action. Downloadable text files contain the debrief and an authored corrected example, plus your ungraded rewrite if supplied. Tables show at most 100 rows across the result, with a truncation notice when necessary. Input is limited to 10,000 records and 1,000,000 UTF-8 bytes; a real workshop needs only six records.

Choosing ‘I would send it now’ records a discussion response only. It does not send email, book time, connect a CRM, or approve anything. Use fictional inputs here. The deterministic tool runs locally; any later use of a separate AI service is your own action under your organization’s rules.

Turn the rehearsal into one repeatable review habit

Ask the group: Which wrong line was easiest to overlook? Which source changed your mind? What is the difference between an available time and a commitment? Which useful line should we preserve? If a participant already found every error, ask them to explain the evidence trail rather than reward speed. Do not collect names, rank employees, or treat this exercise as certification.

For the next week, choose one low-risk internal draft with approved or fictional inputs. Name the source owner and reviewer, define three checks before starting, and record preparation, drafting, review, and correction time separately. Keep the original and revised drafts so you can discuss one concrete improvement. One run cannot establish durable productivity gains or general AI ability.

An optional AI extension is to give an approved tool only F01–F07 and ask it for a draft plus a sentence-to-source table. Keep the same human review and no-send boundary. The workshop still works if you do not use a model. If you adapt the facts or errors, update your own answer key; this page’s checker continues to compare against the original fixed fixture.

Share the public workshop or blank practice materials with colleagues, not a completed participant ledger. The public guide and worked answer can be read freely; signup for the free Something Big Is Happening newsletter unlocks the interactive tool, results, and original downloads. A facilitator’s invitation does not subscribe the team. Each person opts in separately and can unsubscribe any time.

Before you get started

Do employees need a paid AI tool for this workshop?

No. The complete exercise uses a supplied fictional source packet and deliberately flawed draft. Participants can work on paper. An approved AI tool is an optional extension after the core review, not a requirement.

Does the result measure AI skills or certify employees?

No. It compares six review decisions and declared source IDs with a disclosed teaching key. It does not assess job performance, measure general AI ability, or automatically verify the rewritten email. Do not use it for employee rankings.

Can I use a real sales email or change the exercise facts?

This checker supports only the fixed C01–C06 and F01–F07 fixture. Use fictional inputs here. You may adapt the downloaded materials for another internal rehearsal, but create and review a matching new answer key; the online checker will not understand changed facts.

How long should an AI workshop for employees take?

This narrow rehearsal is designed for 30 minutes: five for context, eight for individual review, seven for paired review, five for rewriting, and five for debrief. Allow more time for accessibility needs or discussion. It is a focused practice session, not a complete AI training program.

Do I need to subscribe to use the tool and download the kit?

Yes. Signing up explicitly for the free Something Big Is Happening newsletter unlocks the local tool, results, and downloadable kit. You can read the public guide and worked answer before signing up. Each person opts in separately and can unsubscribe any time. A manager sharing the workshop does not subscribe the team.

Sources and how this guide was made

Product guidance is grounded in the sources below. The tools and fictional teaching materials were created for this guide. We do not present these examples as independent product benchmarks or guaranteed outcomes.

Product names belong to their respective owners. Something Big Is Happening is an independent publication. Check current plans, permissions, and availability in the official documentation.