2026 CHATGPT
WORK_
6 ROLES_
PIPELINE.

ChatGPT Work role-based workflows prompt templates 2026

Вводная: 9 июля 2026 OpenAI выкатил ChatGPT Work. Практический вопрос: что деплоить в production завтра? Официальный entry point — задача, которую вы уже умеете верифицировать. Этот гид разбирает Sales, Marketing, Finance, Operations, Product, Engineering на воспроизводимые pipeline с prompt-шаблонами, Plan Mode audit checklist, Scheduled Tasks recipes и usage pool optimization. Launch context + Cowork diff: ChatGPT Work + Codex merge — технический разбор.

1. Root cause: знаете Work, но не эксплуатируете

  1. Mode mismatch → до 5× usage delta — trivial Q&A в Work вместо Chat сжигает shared consumption pool.
  2. Prompt описывает steps, не deliverable — Work строит execution path сам; вы задаёте output contract.
  3. Scheduled Tasks без security preflight — unrestricted plugin scope + auto-send = compliance incident waiting to happen.

2. Три принципа (non-negotiable)

ПринципМеханикаImplementation
Outcome, не procedureAutonomous path planning❌ «Open Salesforce, export…» → ✅ «Из @Salesforce deals 30d: weekly PPT с risk flags»
Plugin stack firstPlugin catalog = data layerGmail/Slack/Drive authorized; explicit @AppName — снижает false pulls
Plan Mode = circuit breakerPlan-before-executeExternal email, finance reports, customer deliverables: line-by-line approval

2.1 Mode routing table: Chat / Work / Codex

RequirementModeRationale
Fast Q&A, brainstorm, single-turn copyChatMin latency, min token burn
Cross-app multi-step, deliverables, hour-long runsWorkPlugin stack + Plan Mode + Computer Use
Code review, PR ops, multi-repo devCodexDev-native toolchain
Recurring unattended background jobsWork + Scheduled TasksCron/trigger без manual kickoff

2.2 Desktop vs Web deployment matrix

ScenarioTarget env
Local files, Computer Use, Free-tier probeDesktop (Mac / Windows)
Team progress monitoring, mobile checkWeb / Mobile (Plus+)
Sales brief pipeline + email notifyWeb Workspace Agent + scheduler
Excel reconciliation, folder batchDesktop Work mode

3. Universal 5-step execution pipeline

1. Wire plugins → 2. Define target + output schema → 3. Audit Plan Mode → 4. Mid-run steer → 5. Accept deliverable & iterate

3.1 Work prompt contract formula

[Role] + [Data source @plugin] + [Task] + [Output format] + [Constraints] + [Acceptance criteria] Skeleton: Ты — [role]. Pull из @Salesforce и @Gmail [time range] [data type]. Execute [action], output в [Google Docs / Excel / PPT / Sites]. Constraints: [no source mutation / 2 decimal places / no auto-send]. On complete: [Slack notify / save to path].

3.2 Plan Mode audit checklist

  • Data source + time window корректны (client/month)?
  • Plan содержит send/delete/overwrite — high-risk ops?
  • Output schema matches team template?
  • Redundant steps removable для token savings?
  • Human-in-the-loop gate required?

4. Шесть ролей: workflow + prompt templates

Compiled from OpenAI cases, Zapier/Nvidia/Virgin Atlantic early adopters, Workspace Agent Cookbook. Swap @plugin под ваш stack.

4.1 Sales

Scenario A: Daily meeting brief (Scheduled) — Baseline: 1–2 h/day manual → target <15 min review.

Scheduled Task: каждый weekday 16:00. 1. Scan tomorrow @Google Calendar client meetings (exclude internal) 2. Per meeting: - @SharePoint / @Salesforce: 30-day notes & activity log - Public news + exec moves (30 days) - 2–3 sentences per external attendee 3. 2–3 page brief per meeting → @Google Drive 4. @Gmail summary with links Subject: «Client briefs tomorrow — [date]» | Table: Client | Time | Topic | Link

OpenAI internal: Discovery → custom PoC за 24h vs weeks.

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

From @Salesforce [account]: pipeline, contacts, recent activity. 1. Interactive Sites dashboard: pipeline, 7-day signals, next actions 2. Scheduled Task: weekdays 08:00 refresh 3. Major delta → @Slack DM Constraints: no external auto-mail; amounts = raw CRM.

Scenario C: Lead audit & pipeline repair (Zapier-adapted)

@Salesforce leads 30d + @Gmail correspondence cross-ref. Find: leads >48h no follow-up (by source), broken chains, estimated pipeline loss. Deliverables: Excel detail + 1-page exec PPT (7-figure risk) + weekly review process.

4.2 Marketing

Scenario A: Research → Brief → multi-market assets (E2E)

Input: [attachment / @Google Drive link] Phase 1 — Brief: audience, pain, positioning → Google Docs Phase 2 — Assets: 1 acquisition email, 3 LinkedIn, 1 landing outline → Drive Phase 3 — Localization: US, EU, APAC; flag sensitive copy Halt after each phase until approval.

Scenario B: Slack/Teams → weekly agenda (Monday 07:00)

1. @Slack #product-launch + @Microsoft Teams «Go-to-Market» (7 days) 2. Extract: decisions, open questions, blockers 3. Update @Google Drive «Weekly agenda» doc (versioned) 4. @Slack #leadership: max 5 bullets Constraint: public threads only.

4.3 Finance

Scenario A: Month-end variance (OpenAI-validated) — Days → hours close cycle.

[Month] budget variance analysis: 1. @Google Drive «Finance/Actuals» + «Forecast» 2. @Google Sheets: actual vs plan by dept; flag >5% or >$50K 3. Performance narrative (Docs): revenue/cost/opex 4. 5–8 slide management PPT (template style) 5. List 3 manual verification checkpoints Constraints: sources immutable; cite cell refs.

Scenario B: Invoice/payment reconciliation (AP automation)

Payment register vs invoice list (@Google Drive). Flag: variance >2%, missing tax ID, duplicate invoice #, vendor mismatch. Review table; no auto-payment trigger.

4.4 Operations

Scenario A: Daily dashboard delta monitor (06:30 cron)

1. [Dashboard URL / @SharePoint report] 2. Delta vs yesterday: >10% swing or new red KPIs 3. 1-page morning brief (Docs): TOP 3, metric table, owner suggestions 4. @Gmail to ops-leads@company.com On access failure: report in Plan, zero hallucinated metrics.

Scenario B: Feedback clustering → product priority

14-day window: @Slack #customer-feedback, @Gmail «NPS-Detractor», @Google Drive tickets. 1. Cluster into 5–8 themes (representative quotes) 2. Priority: frequency × impact × implementation cost 3. Product backlog candidates; Friday auto-refresh via Scheduled Task Constraint: anonymized quotes, no customer PII.

4.5 Product

Scenario A: Jira + GTM launch readiness (Nvidia-adapted)

Launch readiness [feature]: 1. @Jira: Epic/Story status, open blockers 2. @Google Drive «GTM Plans»: milestones 3. @Slack #product-launch: unresolved threads (7 days) 4. Report: R/Y/G score, blocker list, Go/No-Go recommendation No auto Jira mutation; high-risk = human decision.

4.6 Engineering — Work + Codex split

Codex = code path. Work = cross-team docs. Same desktop app context switch.

Scenario A: PR review + release notes

Codex: review [repo] PR #123 (security/perf/test coverage) → Release Notes draft. Work: @Confluence format + @Slack #engineering announcement draft (no auto-send).

Scenario B: Multi-repo weekly engineering report

Codex: [frontend-repo] + [backend-repo] — merged PRs, open P0/P1 → Markdown. Work: Google Docs + burn-down from @Jira; Scheduled Task Friday 17:00.

5. Scheduled Tasks recipe library

RecipeTriggerPipelineRole
Monday agenda refreshMon 07:00Slack digest → agenda docMarketing / Ops
Daily KPI briefWeekday 06:30Dashboard delta → emailOps / Finance
Feedback weekly clusterFri 16:00Multi-channel → themes → priority listProduct
Account pulseWeekday 08:00CRM delta → Sites updateSales

5.1 Scheduled Task prompt syntax

Configure Scheduled Task: - Frequency: [daily / Mondays / 1st of month / @Slack keyword trigger] - Time: [timezone + exact time] - Action: [workflow spec] - Notify: [Slack channel / email / none] - Human gate: [steps requiring pre-approval]

5.2 Unattended preflight security checklist

  • Plugin scope minimized (least privilege)
  • Auto-send disabled unless explicitly required
  • Fixed archive path — no overwrite of shared assets
  • Enterprise: agent network policy verified in Admin Console
  • 2–3 manual dry-runs before cron activation

6. Usage optimization: consumption pool economics

Work и Codex share единый usage billing pool. Identical workflow, different design: до 5× delta.

FactorConsumption impact
Step countLinear positive correlation
Context sizeMore docs/emails = more input tokens
Output lengthOutput tokens ~ input cost
Cache hitCached input ~1/10 of fresh read
Model tierGPT-5.6 heavy reasoning vs lightweight tasks

6.1 Семь hardcore cost-cut tactics

  1. Chat draft → Work execute — prompt size −30–50%
  2. Plan Mode: strip redundant data pulls
  3. Scheduled Tasks: reuse template doc (cache discount)
  4. Compact output spec: «table + 3 bullets» > narrative report
  5. Phase split: Phase 1 direction lock → Phase 2 deliverable gen
  6. Free tier: probe small desktop tasks, log consumption before scale
  7. Enterprise: workspace/group/user caps in Admin Console

6.2 Pre-deploy usage estimation (5 steps)

1. Pick known task (e.g. variance table, ~2h manual baseline) 2. Plan Mode run: count execution steps 3. Execute: log consumption vs included usage 4. Project: daily/weekly → monthly budget fit 5. Over budget → optimize per §6 and A/B re-run

7. Troubleshooting matrix

SymptomRoot causeFix
Work can't find Codex projectIncomplete app migrationUpdate Codex → ChatGPT Desktop; fallback chatgpt.com/download
Plugin authorized, zero dataScope / @ typoCheck plugin dir; @Salesforce not generic «CRM»
Plan OK, output driftStale context / inferencePause + steer; attach critical data explicitly
Scheduled Task no-fireSleep / desktop logged outLong runs: Web Agent; keep desktop awake
Usage > forecastLong output, duplicate pulls§6 optimization; Admin Console limits
Work vs Cowork?Different workflow classesSaaS cross-app = Work; local folder batch = Cowork

8. 30-day onboarding roadmap

PhaseTargetAction
Week 1Single-task proficiency1 task, desktop Work, 3 runs, Plan audit drill
Week 2Plugin deep integration3 core tools; 1 E2E deliverable
Week 3AutomationWeek-1 task → Scheduled; verify 3 triggers
Week 4Team rolloutRole prompt library; Enterprise quota config

9. Case study: Virgin Atlantic + OpenAI Finance validation

Virgin Atlantic сжал sales meeting prep с 1–2h manual до automated brief pipeline. OpenAI Finance ужал month-end variance с days до hours. Driver: repeatable process → standardized prompt + Scheduled Task — не «smarter model», а process engineering.

На Mac эти pipeline оптимальны в desktop Work mode (Computer Use для local Excel, FCP folders, Xcode simulator). При parallel Scheduled Tasks (sales brief + finance close + ops morning brief) unified memory и UI responsiveness становятся bottleneck — аргумент за tiered «Mac frontend + remote backend» architecture.

10. FAQ

Q: С какого role workflow стартовать?

Задача, которую вы быстрее всего валидируете на ошибки: variance analysis, marketing brief, sales meeting prep.

Q: Оптимальная длина prompt?

150–400 слов — data source, output format, constraints. Zero micro-step instructions.

Q: Scheduled Tasks при powered-off device?

Desktop cron требует online device. True 7×24: Web Workspace Agent (Plus+).

Q: Work vs Workspace Agent?

Work = personal agent mode. Workspace Agent = team automation с Admin governance.

Q: PPT/Excel direct external use?

80%-draft — numbers, names, claims require manual verification.

Q: Free tier — какие templates?

Lightweight tasks (invoice reconciliation) на desktop; no long-running cron until consumption profiled.

11. References

12. Conclusion: Mac local Work, long-running agents на MACGPU

ChatGPT Work окупается, когда убирает manual routine. Mac desktop — optimal host для Work + Codex + Computer Use: local Excel, folder batch, GUI automation на Apple Silicon. Но parallel Scheduled Tasks + multi-hour agent runs + multi-repo Codex съедают unified memory и battery; Windows cloud hosts не дают native Apple toolchain.

Production-grade split: Mac local — Work interaction + Plan Mode approval; sales brief pipeline, finance close, ops morning brief как Scheduled agents на MACGPU Remote Mac mini M4 — Apple Silicon unified memory для parallel runs, SSH on-demand, «frontend control + backend resident» tiered architecture.