ChatGPT Work Tutorial: 6 Role-Based Workflows, Prompt Templates & Automation Recipes (2026)

Before copying prompts, understand how ChatGPT Work differs from ordinary Chat. These pain points show up most often on frontline teams:

• Writing steps instead of goals: Work plans its own path. Micromanagement-style prompts limit the agent.
• Assigning tasks before authorizing plugins: Without Gmail, Slack, Drive, and other sources connected, jobs stall mid-run.
• Skipping Plan Mode on high-stakes work: External emails, financial reports, or client deliverables executed without review create uncontrolled risk.
• Using the wrong mode and burning usage: Simple Q&A in Work or complex cross-app jobs in Chat can cost up to 5× more.
• Desktop vs web mismatch: Local Excel reconciliation on web, or scheduled tasks on a sleeping desktop machine—jobs never trigger.

Regardless of role, follow this five-step flow:

1. Connect plugins → 2. Define goal and output format → 3. Review Plan Mode → 4. Intervene mid-run to correct course → 5. Accept deliverables and iterate

Work mode prompt formula:

[Role] + [Data source @plugin] + [Specific task] + [Output format] + [Constraints] + [Acceptance criteria]

Example skeleton:
You are a [role]. Pull [data type] for [time range] from @Salesforce and @Gmail.
Complete [specific action], output as [Google Docs / Excel / PPT / Sites].
Constraints: [do not modify source data / amounts to two decimals / do not send external email].
When done, [notify me on Slack / save to specified folder].

Plan Mode review checklist (confirm each item before execution):

• Are data sources correct (wrong customer or wrong month)?
• Are there high-risk actions like "send externally," "delete," or "overwrite file"?
• Does the output format match team templates?
• Can intermediate steps be removed to save usage?
• Do you need a human confirmation checkpoint?

How long should a prompt be? Focus on data source + output format + constraints—150–400 words is usually enough. Do not write every manual step; that is what Work mode automates.

These templates draw on OpenAI official examples, early tester feedback (Zapier, Nvidia, Virgin Atlantic, and others), and the Workspace Agent Cookbook. Swap @PluginName for your actual stack.

3.1 Sales

Scenario A: Automated customer meeting brief (daily schedule) — Pain point: reps spend 1–2 hours daily assembling customer context. Work approach: scan calendar on schedule → pull CRM notes → search news → generate brief and archive. OpenAI internal case: sales teams turned Discovery conversations into customized PoC proposals within 24 hours (traditionally weeks).

Create a scheduled task that runs every weekday at 4:00 PM.

1. Check my customer meetings in @Google Calendar for tomorrow (exclude internal meetings)
2. For each customer meeting:
   - Pull account notes and activity from the last 30 days from @SharePoint / @Salesforce
   - Search public news and executive updates for that company in the last 30 days
   - Write a 2–3 sentence background summary for each external attendee
3. Generate a 2–3 page brief per meeting, saved as @Google Drive documents
4. Send me a summary @Gmail email with links to each brief

Output format: email subject "Tomorrow's Customer Meeting Briefs — [date]", body as a table (Customer | Meeting time | Key topics | Brief link)

Scenario B: Account command center (Sites + daily refresh)

Based on all opportunities, contacts, and recent activity for [account name] in @Salesforce:

1. Create an interactive account command center (Sites) with:
   - Pipeline overview (stage, amount, expected close date)
   - Key signals from the last 7 days (email, meetings, support tickets)
   - Recommended next actions (priority ordered)
2. Set a Scheduled Task to auto-update the Site every weekday at 8:00 AM
3. DM me on @Slack when there are major changes

Constraints: do not auto-send any external email; amounts must match CRM source data.

Scenario C: Lead review and pipeline repair (adapted from Zapier case)

Analyze new leads from the last 30 days in @Salesforce and follow-up records, cross-referenced with sales correspondence in @Gmail.

Find:
1. Leads with no follow-up for more than 48 hours (grouped by source)
2. Break points in the follow-up chain (where response rate drops sharply)
3. Estimated pipeline loss amount

Output:
- Excel detail sheet (Lead ID | Source | Last follow-up | Break type | Recommended action)
- 1-page executive summary PPT highlighting seven-figure-level opportunity loss
- A weekly repeatable review process (for use as a Scheduled Task)

3.2 Marketing

Scenario A: Research → brief → multi-market assets (end-to-end pipeline)

I uploaded the following customer research: [attachment / @Google Drive link]

Complete the end-to-end marketing workflow:

Phase 1 — Brief:
- Extract target audience, core pain points, competitive positioning
- Output a Campaign Brief (Google Docs) with message pillars and channel recommendations

Phase 2 — Asset generation:
- From the brief, generate: 1 acquisition email, 3 LinkedIn posts, 1 landing page copy outline
- Save to @Google Drive folder "Campaign / [product name]"

Phase 3 — Regional adaptation:
- Adapt core assets for US, Europe, and APAC (language, cultural references, compliance wording)
- Flag sensitive phrasing that needs human review in each version

Pause after each phase and wait for my confirmation before proceeding.

Scenario B: Slack / Teams activity synced to meeting agenda (Scheduled Task)

Set a scheduled task to run every Monday at 7:00 AM:

1. Summarize important discussions from the last 7 days in @Slack #product-launch and @Microsoft Teams "Go-to-Market" channel
2. Extract: decisions made, open questions, blockers that need alignment in the meeting
3. Update the "Weekly Agenda" document in @Google Drive (preserve version history)
4. Post a summary of 5 items or fewer in @Slack #leadership

Constraints: cite only publicly discussed content; do not leak messages marked confidential.

3.3 Finance

Scenario A: Month-end variance analysis (OpenAI internal validated scenario) — OpenAI internal result: month-end close and forecasting compressed from days to hours.

Help complete [month] month-end budget variance analysis:

1. Pull corresponding tables from @Google Drive "Finance / Actuals" and "Finance / Forecast"
2. Create a reconciliation workbook in @Google Sheets:
   - Summarize actual vs forecast variance by department
   - Flag line items with variance >5% or >$50K
   - Preserve all original formulas; do not overwrite source files
3. Draft performance commentary (Google Docs), grouped by Revenue / COGS / OpEx with likely causes
4. Build a 5–8 slide management deck (charts, following attached template style)
5. List 3 key judgment calls that require finance human confirmation

Constraints: do not modify any source data; cite source cell for every number.

Scenario B: Invoice and payment reconciliation (first AP automation gate)

You are an accounts payable specialist. Compare these two datasets:
- Payment register: [@Google Drive link]
- Invoice list: [@Google Drive link]

Flag the following exceptions (return as a table):
| Issue type | Vendor | Invoice # | Amount | Recommended action |
- Amount variance >2%
- Missing tax ID
- Duplicate invoice number
- Vendor name mismatch

Do not initiate payments; output review sheet for human verification only.

3.4 Operations

Scenario A: Daily dashboard change monitoring (Scheduled Task)

Run automatically every weekday at 6:30 AM:

1. Visit [internal dashboard URL / @SharePoint report page]
2. Compare to yesterday's snapshot; extract significant changes (>10% swing or new red indicators)
3. Generate a 1-page morning brief (Google Docs) with:
   - Top 3 items to watch today
   - Metric change table
   - Suggested follow-up owners
4. Send via @Gmail to ops-leads@company.com

If the dashboard is unreachable, tell me in Plan Mode—do not fabricate data.

Scenario B: Customer feedback clustering → product priorities

Monitor new customer feedback from the last 14 days across:
- @Slack #customer-feedback
- @Gmail label "NPS-Detractor"
- @Google Drive "Support Tickets Export"

1. Cluster feedback into 5–8 themes (with representative quotes)
2. Prioritize by frequency × impact × implementation difficulty
3. Output a product review backlog (Notion / Google Docs format)
4. Set a Scheduled Task to refresh the document every Friday

Constraints: anonymize feedback quotes; no customer names.

3.5 Product

Scenario A: Cross Jira + GTM plan launch readiness review (adapted from Nvidia case)

Run a launch readiness review for [product/feature name]:

1. Pull linked Epic / Story completion status and open blockers from @Jira
2. Pull the corresponding GTM plan from @Google Drive "GTM Plans" and check key milestones
3. Extract unresolved discussions from the last 7 days in @Slack #product-launch
4. Output a launch Readiness report (Google Docs):
   - Readiness score (red / yellow / green)
   - Blocker list (owner | due date | risk level)
   - Recommended Go / No-Go judgment with rationale

Do not auto-update Jira status; flag high-risk items for human decision.

3.6 Engineering — Work and Codex together

For engineering, use Codex mode for code and Work mode for cross-team documents, switching inside the same desktop app.

Scenario A: PR review + release notes (Codex-led)

In Codex mode:
1. Review PR #123 in [repo/name], focusing on [security / performance / test coverage]
2. Leave line-by-line review comments in the PR sidebar
3. If approved, draft Release Notes

Then switch to Work mode:
4. Format Release Notes as an @Confluence page
5. Draft an @Slack #engineering announcement (do not auto-send)

Scenario B: Multi-repo issue weekly summary (Codex multi-repo capability)

In Codex mode, across [frontend-repo] and [backend-repo]:
1. Summarize merged PRs this week and open P0/P1 issues
2. Generate an engineering weekly report in Markdown

Switch to Work mode:
3. Convert to Google Docs and insert this week's burndown chart (pulled from @Jira)
4. Set a Scheduled Task to auto-generate every Friday at 5:00 PM

Prompt pattern for setting scheduled tasks:

Set a Scheduled Task:
- Frequency: [daily / every Monday / 1st of month / when keyword appears in @Slack channel]
- Time: [timezone + specific time]
- Action: [specific workflow description]
- Notification: [Slack channel / email / none]
- Human confirmation: [which steps need my approval first]

Safety checklist before unattended runs: restrict plugin access scope; disable auto external send unless explicitly needed; set output archive path; Enterprise users confirm admin Agent network policy; run manually 2–3 times before switching to schedule.

Six steps to run your first ChatGPT Work task:

1. Download desktop app: Install the Mac/Windows client from chatgpt.com/download (free users can trial Work).
2. Connect core plugins: Authorize the 2–3 tools you use most—Gmail, Google Drive, Slack—in the plugin directory.
3. Pick the right mode: Switch to Work mode and confirm Plan Mode is available in the header.
4. Write a clear prompt: Use the role + @data source + task + output format + constraints formula; start with a lightweight job like invoice reconciliation.
5. Review plan and execute: Confirm data sources and high-risk actions; remove extra steps, then approve.
6. Accept and iterate: Check deliverable quality, note usage consumed, then convert to a Scheduled Task when satisfied.

ChatGPT Work and Codex share a single usage billing pool. The same workflow, designed differently, can cost up to 5× more.

Seven cost-saving tactics: ① draft in Chat first, then hand a trimmed version to Work; ② delete extra steps in Plan Mode; ③ reuse the same template doc in Scheduled Tasks; ④ ask for concise output (table + 3 bullet summary); ⑤ split large jobs into Phase 1 direction check → Phase 2 deliverable; ⑥ free users run small jobs on desktop first; ⑦ Enterprise teams set three-tier quotas in Admin Console.

Hard numbers worth citing:

• Output vs input cost ratio: output tokens cost roughly 6× input—long reports cost far more than table summaries.
• Cache discount: re-reading the same document, cached input costs roughly 1/10 of fresh input.
• Workflow cost spread: poor design can cost 5× more on the same task; OpenAI internal month-end variance went from days to hours.
• Lead follow-up break point: leads with no follow-up for 48 hours are a common pipeline loss source; the Zapier-adapted case can estimate seven-figure potential loss.

Frequently asked questions

• Which role workflow should I practice first? Pick a task you know well enough to judge output quality. OpenAI recommends: month-end variance, marketing brief, sales meeting prep.
• Can scheduled tasks run when my computer is off? Desktop depends on the device being online; for true unattended background runs, use web Workspace Agent on Plus or above.
• How is Work different from Workspace Agent? Work is personal agent mode; Workspace Agent is team-built, centrally managed automation inside Business/Enterprise with Admin Console governance.
• Can generated PPT/Excel go straight to external presentation? Treat as an 80% draft—financial numbers and customer names need human verification.
• Which templates can free users run? Desktop Work is available on trial; start with lightweight tasks like invoice reconciliation.

More background: OpenAI official announcement, Sales Meeting Prep Cookbook, and our sister post ChatGPT Work launch guide.

ChatGPT Work's value is not that it exists—it removes manual processes you are already tired of. Fastest ROI: pick one task you know, run it three times, tune the prompt, then automate.

Scheduled Tasks and multi-hour cross-app jobs demand runtime stability. A sleeping Mac, flaky shared Wi-Fi, or oversubscribed VMs dropping long connections can kill agent jobs mid-run. Pure SaaS integrations also cannot replace pipelines that need local file batch processing, Metal acceleration, or 7×24 uptime for mixed Work/Codex workflows.

For production environments running ChatGPT Work, Codex mode, or self-hosted agent orchestration, JEXCLOUD multi-region bare-metal Mac nodes are often the better fit: dedicated Apple Silicon, no virtualization overselling, 120-second delivery, and flexible monthly scaling as a 7×24 host for AI coworkers and coding agents. See node specs and pricing on the JEXCLOUD pricing page.