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

~12 min read · MACCOME · Last updated: July 11, 2026

Who should read this? Sales, marketing, finance, ops, product, and engineering leads who already know what ChatGPT Work is but need a Monday-morning playbook. On July 9, 2026, OpenAI launched ChatGPT Work and merged Codex into the new desktop app. You get: three usage principles, mode and environment decision tables, a five-step universal workflow, copy-paste prompts for six roles, Scheduled Tasks recipes, usage optimization tactics, a 30-day roadmap, and six FAQs. Structure: principles → framework → role templates → automation → cost control → pitfalls → onboarding → conversion close. For launch context and Cowork comparison, see the companion launch guide.

bolt

TL;DR — 30-second verdict

  • Start with a task you already know: month-end variance, campaign brief, or sales meeting prep — you can verify quality fast.
  • Describe outcomes, not steps: Work plans its own path; pin data with @AppName and review Plan Mode before execution.
  • Six roles, ready prompts: Sales (3), Marketing (2), Finance (2), Ops (2), Product (1), Engineering (2) — adapt plugin names to your stack.
  • Automate last: run manually 2–3 times, then convert to Scheduled Tasks with a safety checklist.
  • Watch usage: same workflow can cost 5x more depending on design — draft in Chat, trim Plan steps, request concise outputs.

Six Pain Points: Why Knowing ChatGPT Work Is Not Enough

OpenAI's onboarding advice is simple — start with a task you already know well — but most teams stall before the first successful run. These six blockers show up across early adopters at Zapier, Nvidia, Virgin Atlantic, and internal OpenAI sales teams:

  1. Wrong mode, wasted quota: Running multi-hour cross-app jobs in Chat mode stalls; using Work for deep code review underperforms Codex. Picking the wrong lane burns included usage before you deliver anything.
  2. Step-by-step prompts that fight the agent: Work mode plans its own path. Micromanaging "open Salesforce, export, then…" produces brittle runs and longer step chains.
  3. Disconnected plugins: Gmail, Slack, and Drive must be authorized before the task starts. Vague references like "the CRM" fail; explicit @Salesforce pins the data layer.
  4. Skipped Plan Mode on high-stakes work: External emails, financial reports, and client deliverables need step-by-step approval. Auto-executing without review is the fastest path to rework.
  5. Premature automation: Scheduling before two or three manual validation runs leads to silent failures — especially when a laptop sleeps or a dashboard URL changes.
  6. Usage shock at month-end: Work and Codex share a metered pool. Verbose outputs, duplicate data pulls, and GPT-5.6 on lightweight tasks can make the same workflow cost five times more than a trimmed design.

This guide removes those blockers with principles, templates, and a phased rollout plan — not another launch recap.

Three Principles That Decide Success

Before copying any prompt, internalize how ChatGPT Work differs from everyday Chat:

PrincipleWhat it meansPractical tip
Describe outcomes, not steps Work mode plans its own execution path Avoid: "Open Salesforce, export data, then…" — Prefer: "Build a weekly pipeline PPT from @Salesforce deals in the last 30 days, flagging at-risk opportunities"
Connect tools first Plugins are Work's data layer Authorize Gmail, Slack, Drive before starting; use @AppName to pin sources explicitly
Plan Mode is your brake Review the plan before execution For high-stakes deliverables (external emails, financial reports, client docs), approve every step

Pick the Right Mode: Chat / Work / Codex

The new ChatGPT desktop app runs three modes behind one shell. Using the wrong one wastes quota:

Your needRecommended modeWhy
Quick Q&A, brainstorming, single-turn copy Chat Lightweight, fast response
Cross-app multi-step projects, finished deliverables, hours-long tasks Work Plugin integrations + Plan Mode + Computer Use
Code review, PR management, multi-repo development Codex Developer-native workflows preserved post-merge
Weekly recurring, unattended background tasks Work + Scheduled Tasks Triggered or scheduled execution

Desktop vs Web: Where to Run Each Workflow

ScenarioRecommended environment
Local file read/write, Computer Use, free-tier trial Desktop (Mac / Windows)
Team collaboration, monitor task progress anywhere Web / mobile (Plus and above)
Sales meeting brief auto-generation + email notification Web Workspace Agent + scheduled dispatch
Local Excel reconciliation, folder batch processing Desktop Work mode

Universal Five-Step Workflow Framework

Regardless of role, run your first task through this sequence:

workflow
1. Connect plugins → 2. Write goal + output format → 3. Review Plan Mode → 4. Steer mid-flight → 5. Accept deliverable & iterate

Work Mode Prompt Formula

prompt
[Role] + [Data sources @plugins] + [Task] + [Output format] + [Constraints] + [Acceptance criteria]

Example skeleton:

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

Plan Mode Review Checklist

Before approving execution, confirm each item:

  • Are data sources correct (right account, right month)?
  • Any high-risk actions (send external email, delete, overwrite files)?
  • Does output match your team's template?
  • Can any steps be removed to save usage?
  • Do you need a human approval checkpoint?

Six Role-Based Workflows with Prompt Templates

Templates below draw on OpenAI official cases, early tester feedback (Zapier, Nvidia, Virgin Atlantic), and the Workspace Agent Cookbook. Replace @plugin names with your actual stack.

Sales

Scenario A: Daily customer meeting briefs (scheduled)

Pain point: Reps spend 1–2 hours daily assembling client background, recent news, and meeting agendas.

Work solution: Scan tomorrow's calendar, pull CRM notes, search recent news, generate and archive briefs.

OpenAI internal reference: Sales teams converted a Discovery conversation into a customized PoC proposal within 24 hours — a process that traditionally took weeks.

prompt — sales A
Create a scheduled task running every weekday at 4pm:

1. Check tomorrow's customer meetings in @Google Calendar (exclude internal-only)
2. For each customer meeting:
   - Pull 30-day account notes and interaction history from @SharePoint / @Salesforce
   - Search 30-day public news and executive updates for that company
   - Write a 2–3 sentence background summary for each external attendee
3. Generate a 2–3 page brief per meeting, save as @Google Drive documents
4. Email me a summary with links via @Gmail

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

Scenario B: Live account command center (Sites + daily refresh)

Pain point: Enterprise account data lives across CRM, email, and Slack; reps manually maintain account plans.

Work solution: Build a live Sites dashboard that refreshes daily.

prompt — sales B
From all opportunities, contacts, and recent activity for [Account Name] in @Salesforce:

1. Create an interactive account command center (Sites) including:
   - Pipeline overview (stage, amount, expected close date)
   - Key signals from the last 7 days (email, meetings, support tickets)
   - Recommended next actions (priority sorted)
2. Set a Scheduled Task: auto-refresh the Site every weekday at 8am
3. DM me via @Slack when major changes occur

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

Scenario C: Lead review and pipeline repair (Zapier-style workflow)

Pain point: Thousands of monthly leads with invisible follow-up gaps.

Work solution: Cross-reference CRM + email touchpoints; output an executive dashboard.

prompt — sales C
Analyze @Salesforce leads from the last 30 days and cross-reference @Gmail sales correspondence.

Find:
1. Leads with no follow-up for 48+ hours (grouped by source)
2. Broken handoff points (where response rate drops sharply after a step)
3. Estimated pipeline loss amount

Output:
- Excel detail table (Lead ID | Source | Last follow-up | Gap type | Recommended action)
- 1-page executive summary PPT highlighting seven-figure potential loss opportunities
- A repeatable weekly review workflow (for Scheduled Task use)

Marketing

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

Pain point: Research, campaign briefs, and regional assets are split across people; context gets lost between handoffs.

Work solution: One instruction spans the full pipeline with context carried forward.

prompt — marketing A
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 Campaign Brief (Google Docs) with messaging 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 "Campaign / [product name]" folder

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

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

Scenario B: Slack / Teams sync to meeting agenda (weekly scheduled)

Pain point: Weekly agendas go stale; someone must manually scan multiple channels.

Work solution: Auto-summarize channel activity and refresh the agenda doc.

prompt — marketing B
Set a scheduled task running every Monday at 7am:

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 needing alignment in the meeting
3. Update the "Weekly Agenda" document in @Google Drive (preserve version history)
4. Post a summary of 5 bullets or fewer to @Slack #leadership

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

Finance

Scenario A: Month-end variance analysis (OpenAI-validated use case)

Pain point: Month-end close and forecast adjustments take days; most time goes to finding numbers and building tables.

Work solution: Auto-locate source data, populate Sheets, reconcile, and generate a management deck.

OpenAI internal result: Month-end close and forecast workflows compressed from days to hours.

prompt — finance A
Assist with [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 narrative (Google Docs) explaining likely causes by Revenue / COGS / OpEx
4. Build a 5–8 slide management deck (with charts, following attached template style)
5. List 3 key judgment calls requiring human finance sign-off

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

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

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

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

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

Operations

Scenario A: Daily dashboard change monitoring (scheduled)

prompt — ops A
Run automatically every weekday at 6:30am:

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) structured as:
   - TOP 3 items to watch today
   - Metrics change table
   - Recommended 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

prompt — ops B
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. Rank by frequency × impact × implementation effort
3. Output a product review backlog (Notion / Google Docs format)
4. Set a Scheduled Task to auto-refresh this document every Friday

Constraints: anonymize all customer references; no customer names in output.

Product

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

Pain point: Launch readiness requires checking engineering progress, marketing plans, and support docs — manual and error-prone.

Work solution: Pull status across systems; output a Go/No-Go readiness report.

prompt — product A
Launch readiness review for [product/feature name]:

1. From @Jira: pull linked Epic / Story completion status and open blockers
2. From @Google Drive "GTM Plans": pull the corresponding launch plan and check key milestones
3. From @Slack #product-launch: extract unresolved discussions from the last 7 days
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.

Engineering — Work + Codex in the Same App

Engineering workflows benefit from Codex for code and Work for cross-team documents. Switch modes inside the same desktop app — no tool change required.

Scenario A: PR review → release notes → team announcement (Codex-led)

prompt — engineering A
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 side panel
3. If approved, draft Release Notes

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

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

prompt — engineering B
In Codex mode, across [frontend-repo] and [backend-repo]:
1. Summarize this week's merged PRs 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 (pull from @Jira)
4. Set a Scheduled Task to auto-generate every Friday at 5pm

Scheduled Tasks Recipe Library

OpenAI recommends four high-frequency scheduled patterns you can adapt directly:

Recipe nameTriggerTask descriptionBest for
Monday agenda refresh Mon 7:00 Summarize Slack activity → update agenda doc Marketing / Ops
Daily metrics brief Weekdays 6:30 Visit dashboard → compare yesterday → email report Ops / Finance
Feedback clustering weekly Fri 4:00pm Multi-channel feedback → theme clusters → priority list Product
Account daily refresh Weekdays 8:00 CRM changes → update Sites command center Sales

Scheduled Task Prompt Pattern

prompt
Set 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 approval: [which steps require my sign-off first]

Safety Checklist Before Going Unattended

  • Plugin access scoped to necessary tools only
  • Auto-external-send disabled unless explicitly required
  • Output archive path set to avoid overwriting others' files
  • Enterprise: agent network policy confirmed with admin
  • Run 2–3 manual single executions before switching to scheduled

Usage Optimization: Do More for Less

ChatGPT Work shares a metered usage pool with Codex — not a flat monthly feature. The same workflow can cost 5x more depending on how you design it.

Official Billing Logic (Simplified)

FactorImpact on usage
Number of task steps More steps = higher consumption
Context size More documents and emails pulled = higher consumption
Output length Output token cost is roughly 6x input token cost
Cache hits Re-reading the same document: cached input costs roughly 1/10 of fresh input
Model selection GPT-5.6 complex reasoning consumes more than lightweight tasks need

Seven Cost-Saving Tactics

  1. Draft in Chat first, then hand a tight brief to Work for execution
  2. Trim Plan Mode steps, especially duplicate pulls from the same data source
  3. Reuse template documents in Scheduled Tasks to leverage cache discounts
  4. Request concise outputs: "table + 3 bullet summary" beats a full narrative report
  5. Split large projects: Phase 1 confirms direction, Phase 2 generates deliverables — avoids expensive full re-runs
  6. Free users: run small desktop tasks first; measure consumption before scaling
  7. Enterprise teams: set workspace / group / individual limits in Admin Console

Pre-Launch Usage Test Method

runbook
1. Pick a real task you know the human time cost of (e.g., month-end variance table — usually 2 hours manually)
2. Run once in Work mode with Plan Mode; record step count
3. After execution, check consumption against your plan's included usage
4. Extrapolate: if run daily / weekly, is monthly consumption within budget?
5. If high → optimize per section 6.2 and re-run to compare

Common Pitfalls and Troubleshooting

IssueCauseFix
Work mode cannot find installed Codex projects Incomplete app migration Update Codex app → becomes ChatGPT desktop; if broken, reinstall from chatgpt.com/download
Plugin authorized but no data returned Insufficient scope or wrong @name spelling Re-check plugin directory permissions; use explicit @Salesforce not "the CRM"
Plan looks right, output is wrong Stale context or AI inference Pause and steer mid-flight; attach explicit source files or links
Scheduled task did not fire Device asleep or desktop logged out Use web Workspace Agents for long-cycle tasks; desktop tasks need device awake and logged in
Usage higher than expected Verbose output, redundant pulls, too many steps See usage optimization section; Enterprise: set limits in Admin Console
Unclear whether to use Work or Cowork Different workflow types Cloud SaaS collaboration → Work; local folder batch processing → Cowork (see companion comparison)

30-Day Onboarding Roadmap

PhaseGoalAction
Week 1 Single-task fluency Pick one familiar task; run desktop Work mode manually 3 times; practice Plan Mode review
Week 2 Plugin depth Connect 3 core tools (email + collaboration + files); complete one cross-app end-to-end deliverable
Week 3 Automation Convert Week 1 task to Scheduled Task; verify 3 successful triggers
Week 4 Team rollout Document role-specific prompt library; Enterprise teams sync admin usage limits

Six-Step Getting Started Runbook

Run this sequence on day one — before scaling to your team:

  1. Install or update the ChatGPT desktop app from chatgpt.com/download (existing Codex installs migrate in place).
  2. Connect three plugins you use daily — typically email, file storage, and one business system (CRM, Jira, or Slack).
  3. Pick one verifiable task from your role section above (variance analysis, meeting brief, or invoice reconciliation).
  4. Write the prompt using the formula — role, @sources, output format, constraints, acceptance criteria — then switch to Work mode.
  5. Review Plan Mode line by line using the checklist; approve only after high-risk steps are confirmed or removed.
  6. Iterate twice, then schedule: run manually two more times, tune the prompt, pass the safety checklist, and convert to a Scheduled Task if appropriate.

Three Hard Data Points Worth Citing

Figures below come from OpenAI's launch materials and early enterprise case studies — safe for internal briefings:

  • 1400+ — integrations in ChatGPT Work's unified plugin directory at launch
  • Days → hours — OpenAI internal finance teams compressed month-end close and forecast workflows using Work mode
  • 5x — potential usage cost spread for the same workflow depending on step count, context size, and output verbosity (metered pool shared with Codex)

Additional context: OpenAI reports 5 million weekly Codex users and 1 million+ doing non-coding work — evidence that agent workflows are crossing from engineering into general knowledge work. See the launch companion for full feature breakdown and Cowork comparison.

Closing: When Your Laptop Is the Bottleneck

ChatGPT Work delivers ROI when it removes a workflow you already resent doing manually — not when you read more launch coverage. The fastest path: pick one task you know intimately, run it three times, tune the prompt, then automate it with Scheduled Tasks.

Three gaps appear when you run agents on a personal MacBook:

  • Sleep and network jitter: Lid-close or Wi-Fi handoff kills multi-hour jobs; approved Plan steps may need full reruns.
  • Permission and data mixing: Computer Use needs Accessibility and file access alongside daily browsers and production secrets — risky on one machine.
  • No true 24/7 duty cycle: Desktop Scheduled Tasks only fire when the host stays online and logged in.

For stable AI agent automation — Work scheduled pipelines, Codex multi-repo jobs, or OpenClaw gateways — MACCOME Mac cloud hosts provide real macOS, SSH handoff, and isolated environments so agents run 24/7 on dedicated nodes instead of your daily laptop. Review Mac Mini cloud rental rates for public tiers.

Sources: OpenAI blog, OpenAI Cookbook — Sales Meeting Prep Agent, ChatGPT Learn changelog, SiliconANGLE launch coverage, Developers Digest — Codex merge analysis.

FAQ

Which ChatGPT Work workflow should I try first?

The task you know best and can verify — month-end variance analysis, a campaign brief, or sales meeting prep. OpenAI recommends these because you can quality-check output quickly.

How long should my ChatGPT Work prompt be?

Aim for 150–400 words focused on data sources, output format, and constraints. Do not micromanage every step — that is what Work mode automates.

Do Scheduled Tasks run when my laptop is off?

Desktop Scheduled Tasks require the device online and logged in. For true background automation, use web Workspace Agents on Plus or higher. For always-on desktop agents, a dedicated MACCOME Mac cloud host avoids sleep and lid-close interruptions.

What is the difference between Work mode and Workspace Agent?

Work is personal agent mode inside ChatGPT. Workspace Agents are team-built, admin-governed automations in Business or Enterprise with Admin Console controls. Same technical foundation, different entry points.

Can I use generated slides or reports externally as-is?

Treat them as 80% drafts. Always human-review financial numbers, customer names, and external statements before publishing or presenting.

What can free users run from this guide?

Desktop Work mode is available with usage limits. Start with lightweight tasks like invoice reconciliation (Finance Scenario B) before scheduling long-running automation.