Chapter 14: Copilot Cowork — Delegating Real Work

1Chapter 14: Copilot Cowork — Delegating Real Work¶

“It is easy to have a dozen tasks in flight at once, each one moving forward while you focus on what only you can do.”

It is 4:40 p.m. on the last day of move-out at a 480,000-square-foot show in Las Vegas. A GES show manager is standing in an aisle that used to be a technology pavilion, watching the last of the aisle carpet come up, holding a radio in one hand and a phone in the other. The marshaling yard is running forty minutes behind. Two exhibitors still have crates unaccounted for. The organizer wants a preliminary reconciliation summary “sometime tomorrow, ideally before my board call.”

That reconciliation package is real work. It means pulling the labor hours from the on-site sheets, comparing them to the forecast, reconciling material handling weights against the freight targets, cross-referencing the exhibitor orders that changed on the floor against what was actually delivered, and writing a narrative that a show organizer can read in six minutes and understand.

It is four hours of work, minimum. It requires SharePoint, Excel, Outlook, Teams, and a document library. And the person who has to do it is currently standing in a mostly empty hall with a dying phone battery and a 7:15 a.m. flight.

Here is the change this chapter is about. That show manager can open Copilot on their phone, describe the outcome they want, attach the show’s document library, and walk away. The task keeps running when the phone goes into airplane mode. Somewhere over Denver, the work is still progressing. By the time they land, there is a draft reconciliation workbook in the show folder, a narrative summary in Word, and a proposed email to the organizer waiting for approval — not sent, waiting.

That is Copilot Cowork. And the skill it demands is not a new kind of prompting. It is a new kind of management.

1.11. The Shift: Copilot Now Has Three Gears¶

Every chapter before this one taught you to work with Copilot. This chapter teaches you to delegate to it.

The distinction matters more than it sounds. Up to now, the pattern has been consistent: you prompt, Copilot responds, you evaluate, you refine. You are the loop. Every step passes through you. That is a genuinely powerful way to work, and for a large share of daily tasks it remains the right one.

Cowork breaks that pattern. You describe an outcome. Cowork plans the steps, opens the apps, retrieves the files, does the analysis, produces the documents, and pauses only when it needs a decision from you. You are no longer the loop. You are the reviewer.

Microsoft frames the Copilot family in three modes, and this table is the single most important thing in the chapter. Learn it well enough to explain it to a colleague in thirty seconds.

The professional behavior underneath this is worth naming plainly, because it is the actual skill.

The prompt-craft skill of Copilot Chat is describing a task. The skill of Cowork is describing an outcome and then reviewing like a manager.

Anyone at GES who has ever run a crew knows exactly what that means. You do not tell a lead installer which screw to pick up. You tell them what the booth looks like when it’s done, what the constraints are (union jurisdiction, the venue’s rigging rules, the 6 p.m. hard stop), what you want to be consulted on, and then you let them work — and you inspect the result before the exhibitor sees it. That is precisely the relationship Cowork asks for.

1.22. What Copilot Cowork Actually Is¶

1.2.12.1 The one-sentence definition¶

Copilot Cowork executes complex, long-running, multi-tool tasks end-to-end across Microsoft 365 and returns finished artifacts — a deck, a report, an email, an updated workbook — not a draft suggestion or a recommendation.

1.2.22.2 The timeline¶

Cowork did not appear overnight, and its history tells you something about how carefully it was built.

At GA, more than half of the Fortune 500 were already using it, including Accenture, Avanade, Advance Local, Capital Group, Koch, LTM, Ooredoo Qatar, and Zurich Insurance. That adoption curve is unusual, and it’s the strongest available evidence that this is not a novelty feature.

1.2.32.3 The models — and the Anthropic relationship¶

Cowork’s underlying intelligence is worth understanding, because it explains both its capability and its cost model.

Microsoft worked closely with Anthropic and integrated the technology behind Claude Cowork into Microsoft 365 Copilot. At general availability, Cowork runs on Anthropic’s frontier models — Opus 4.8 and Sonnet 4.6. Microsoft’s own fine-tuned model, Cowork 1, was post-trained for substantially lower cost and is releasing shortly after GA.

Microsoft’s positioning here is deliberate and, frankly, sensible: your work is not limited by one brand of models. Copilot hosts the best innovation from across the industry and chooses the right model for the job regardless of who built it. A lightweight formatting task does not need a frontier reasoning model. A 40-source deep research synthesis does. Multi-model design means the platform can match the model to the task — which is a capability story and a cost story at the same time.

1.2.42.4 The five things that make Cowork structurally different¶

1.33. What Cowork Can Actually Do¶

Cowork’s capability surface breaks into five areas. Read this section with your own week in mind — the goal is for at least three of these to trigger a “wait, I do that every show” reaction.

1.3.13.1 Communication¶

Cowork can draft and send email through Outlook — new messages, replies, forwards. It can post updates in Teams channels and send direct messages in 1:1 or group chats. It can build and send HTML newsletters. It can manage an inbox: sorting into folders, deleting, responding inline. And it can prepare full stakeholder communication sets — status updates, announcements, follow-ups.

At GES: the pre-move-in exhibitor notice that goes to 800 exhibitors on a show; the daily on-site status post to the show team’s Teams channel; the “your freight has arrived at the advance warehouse” confirmation wave; the organizer weekly update during the 90 days before a show.

1.3.23.2 Documents and files¶

Cowork can create Word documents, Excel workbooks, PowerPoint decks, and PDFs from scratch. It can edit and refine documents shared into the session. It can browse your entire Work IQ to pull in the content it needs. It can create SharePoint and OneDrive folders and reorganize existing files into them.

At GES: the show operations manual; the exhibitor service kit refresh; a post-show recap deck; the labor forecast workbook; a Spiro spec document; the show folder structure itself, built and populated at kickoff.

1.3.33.3 Calendar and meetings¶

Cowork can schedule meetings from natural language, manage a calendar (adding, moving, resolving conflicts, declining with a reason message to the organizer), surface meeting intelligence and prep insights, and deliver a daily briefing of what’s ahead.

At GES: the week of a show, when a single account lead has an organizer walkthrough, three exhibitor escalations, a union steward meeting, and a client dinner all colliding on the same afternoon.

1.3.43.4 Research and search¶

Cowork can search across the organization for documents, messages, and information; run deep research that synthesizes many sources into a comprehensive report; and browse SharePoint and OneDrive folders to select files.

At GES: “What did we learn at this venue last year?” — a question whose answer is currently distributed across three people’s memories, a OneNote site survey, and an email thread nobody can find.

1.3.53.5 Automation¶

Two forms, and they’re the ones most people miss:

• Scheduled prompts — run a prompt on a schedule so recurring tasks happen automatically. Every Monday at 6 a.m., produce the freight-target compliance summary for all shows in move-in this week.

• Event-driven tasks — run when something happens. When an email arrives from the venue’s exhibitor services desk, extract the request, classify it, and post it to the show channel.

1.44. The Session Experience — What It’s Actually Like to Use¶

A Cowork session has a rhythm, and knowing it removes most of the first-week friction.

Step 1 — Describe your task. You type or dictate the outcome you want. The input field accepts up to 250,000 characters, so you are not constrained to a one-liner — paste the whole organizer brief if that’s what’s useful. You can drag files directly into the chat, use Add work context to attach specific files, people, emails, Teams chats, channels, or meetings, upload from your device, or attach cloud files from OneDrive, SharePoint, or Teams.

There is a microphone button for voice input. This is not a gimmick for GES. Dictating a task while walking a show floor, hands full, is a materially different experience from typing it.

The home page offers suggested prompts — Catch me up, Organize my inbox, Organize my week, Prep for a meeting, Plan an event, Prepare for my 1:1, Research a company — and lists your recent tasks so you can resume any previous session.

Step 2 — Watch Cowork work. Cowork breaks the request into steps and works through them one at a time, visibly. You see a thinking indicator, step-by-step status updates (“Searching OneDrive,” “Composing your email”), responses streaming word by word, and interactive cards. Active skills appear in the side panel as they load. A connection status indicator shows Connecting / Connected / Reconnecting / Failed, with a Retry option.

Step 3 — Steer, interrupt, or queue. You can send another message while Cowork is busy. Queued messages are processed in order, and Cowork adjusts direction if what you send changes the plan. You can pause at any point to add context or correct a wrong assumption. Cowork will also ask you clarifying questions — often with multiple-choice options — and you can Skip them if the answer doesn’t matter to you.

Step 4 — Approve actions when asked. Cowork pauses before consequential actions. This is important enough to have its own section (§7).

Step 5 — Review the results. Download the documents, check what was sent, and request changes in the same session.

1.55. Skills and Plugins — How Cowork Gets Specialized¶

Cowork loads specialized skills as it works. You don’t invoke them by name; Cowork selects them based on what the task needs, and the active ones appear in the side panel so you can see what it’s using.

Custom skills. You can create up to 50. Cowork discovers them automatically at the start of each session. For GES, the obvious candidates are the things that are specific to how this company works: the standard post-show reconciliation format, the freight-target compliance calculation, the exhibitor service kit house style, the Spiro spec document template, the ESG emissions data schema.

Plugins. Cowork supports plugins from the Microsoft 365 App Store. These add new skills and connectors — specialized expertise (financial analysis, legal research) or connections to external data sources. Admins can deploy plugins org-wide, which is the right pattern for anything that touches a system of record.

1.66. The Five-Part Cowork Prompt — The Core Skill of This Chapter¶

Here is the honest truth about Cowork: most disappointing results are scoping failures, not capability failures.

“Clean up my calendar” leaves almost everything open to interpretation. Cowork has to guess what you value, and guessing shapes the result. The same request, scoped properly, produces something you’d actually ship.

Microsoft’s official guidance gives a five-part structure. Learn it, use it every time, and your success rate with Cowork changes immediately.

The before-and-after that makes this click:

Help me prep for my offsite next week.

Outcome: A briefing pack and a draft agenda for the offsite on June 12.

Inputs: The offsite calendar invite, the three pre-reads I've attached, 
and the notes from our last two leadership syncs.

Definition of done: One Word briefing document and one draft agenda, 
both saved to the Leadership Offsite folder in SharePoint.

Constraints: Keep the briefing to four pages. Use our standard agenda 
format. Do not include budget figures — those come from Finance separately.

Approval scope: Don't send anything to attendees. Just save the files 
and tell me when they're ready.

1.77. Approvals and Control — The Governance Heart of Cowork¶

If you read only one section of this chapter twice, make it this one.

Cowork can send email. It can post in Teams. It can update records and move files and decline meetings. That capability is the whole point — and it is also the reason the approval model exists and the reason you need to understand it properly rather than click through it.

1.7.17.1 The approval dialog¶

Before a sensitive action, Cowork pauses and shows you what it intends to do. For many actions you get a rich preview — the actual draft email, the actual Teams message, the actual meeting invite. For others you get a summary. Medium and high risk actions carry a risk level indicator.

You then have four choices.

1.7.27.2 The permission model — the part that should give you confidence¶

Every Cowork task runs with your permissions and sees only what you can see.

That sentence deserves to be read slowly. Cowork does not have a privileged view of GES data. It is not an administrator. If you cannot open a SharePoint site, Cowork cannot open it on your behalf. If an organizer’s contract folder is restricted to the account team, Cowork working for someone outside that team cannot read it.

Data stays in your tenant. Existing user and admin permissions are respected. Identity and compliance policies apply by default. And Cowork actions are auditable — there is a record of what was done, when, and on whose behalf.

This is architecture, not policy language. It’s the same principle that governs Microsoft 365 Copilot generally, extended to a tool that can now act rather than only answer.

1.7.37.3 Where this meets T.R.U.E.¶

GES’s values are not decoration in this chapter. Two of the four map directly onto the approval model, and one maps onto the whole idea of delegation.

1.7.47.4 Auditability as an operational asset¶

It’s easy to read “actions are auditable” as a compliance checkbox. At GES it’s more useful than that.

An audit trail means that when an organizer asks “when did you tell us about the freight target change?”, the answer is retrievable. It means that when a show team hands off from pre-show to on-site crew, the record of what was communicated and when is intact. It means a post-show dispute about what an exhibitor was told has evidence behind it rather than three conflicting recollections.

Delegation without a record is risk. Delegation with a record is process.

1.88. Economics and Governance — What Cowork Costs and How to Think About It¶

Cowork is not free, and pretending otherwise makes it harder to adopt well.

1.8.18.1 The licensing structure¶

• The Microsoft 365 Copilot User Subscription License (USL) is a prerequisite. Cowork sits on top of it.

• Cowork itself is billed on usage, denominated in Copilot Credits.

The USL separately includes Copilot Chat; Copilot in Word, Excel, PowerPoint, Outlook, and Teams; the Work IQ context engine; multi-model frontier intelligence; the pre-built agents (Researcher, Analyst); and custom agents via Agent Builder. Cowork is the delegation layer added on top of all of that.

1.8.28.2 The four cost inputs¶

The price of a Cowork task is calculated from four things:

Notice that three of the four are directly controlled by prompt quality. A well-scoped prompt is a cost-control mechanism, not just a quality mechanism. That is not a coincidence and it’s worth saying to anyone at GES who has to sign off on the budget.

1.8.38.3 Light, medium, heavy¶

Microsoft observes three task patterns in practice:

Microsoft also identified four user personas with distinct usage patterns, and publishes a customer estimator spreadsheet at aka.ms/CustomerCoworkEstimator (its figures assume Anthropic Opus 4.8). The underlying cost model is straightforward:

(users per segment) × (expected prompt volume across light / medium / heavy) × (cost per prompt type), summed across segments.

Microsoft’s own testing reports that Copilot Cowork averaged 30–40% cheaper per prompt than Claude Cowork operating through its Microsoft 365 connector — a consequence of efficient runtime and matched model selection.

1.8.48.4 What this means at a company of ~2,600 people¶

If everyone at GES ran three heavy tasks a day because heavy tasks are fun to watch, the bill would be real. If nobody used Cowork because they were unsure about credits, the opportunity cost would be larger. Neither extreme is the answer.

Here is a practical rule of thumb for GES.

Governance guidance for GES leaders deploying this:

• Give teams a monthly credit budget and visibility into it. People manage what they can see.

• Build a shared prompt library. A well-scoped prompt costs less and produces more. Reusing a good one is free efficiency.

• Convert the repeated heavy tasks into scheduled prompts. A weekly scheduled task with a tight scope costs less than six people improvising the same analysis.

• Watch for the “just re-run it” habit. Re-running a whole heavy task because one number was wrong is expensive. Steering mid-task, or asking for a targeted correction in the same session, is cheaper.

1.99. Microsoft’s Four Flagship Scenarios, Translated to GES¶

Microsoft leads with four demonstration scenarios. Each has a direct GES analogue, and walking the translation is a fast way to see the shape of what Cowork does.

1.9.19.1 “Clean up your calendar” → Show-week calendar triage¶

What Microsoft shows: Cowork reviews your Outlook schedule, asks what you’re prioritizing, flags conflicts and low-value meetings, and proposes changes. On approval it accepts, declines, or reschedules meetings and adds focus blocks. It can also send a prep document for a meeting.

The GES analogue: It is the Tuesday of move-in week. A GES account director has an organizer walkthrough, two internal syncs, a Spiro design review, and a supplier call — three of which overlap with the hours they physically need to be on the floor.

Outcome: A show-week calendar that protects Wednesday 7 a.m.–1 p.m. and Thursday 7 a.m.–noon for on-floor time during move-in. Inputs: My Outlook calendar for June 8–12; the Show 4412 move-in schedule. Definition of done: Conflicts resolved, focus blocks added, declines sent with a reason. Constraints: Never move or decline anything with the organizer on it — flag those for me instead. Internal meetings can be moved freely. Approval scope: Ask before any decline that goes to someone outside GES.

1.9.29.2 “Build the meeting packet and align the team” → The organizer pre-con packet¶

What Microsoft shows: Cowork pulls inputs from email, meetings, and files; schedules prep time; and produces a connected set of deliverables — a briefing document, supporting analysis, and a client-ready deck, all saved in Microsoft 365 — plus a draft customer status email with the latest files attached.

The GES analogue: The pre-conference meeting with a show organizer, ninety days out. Currently this is assembled by three people across a week from a floor plan, a labor forecast, last year’s post-show report, and an email thread.

The Cowork version produces the briefing doc, the supporting analysis workbook, the organizer-ready deck, a scheduled internal prep session the day before, and a draft email to the organizer with the deck attached — waiting for approval.

1.9.39.3 “Research a company fast” → Organizer and prospect intelligence¶

What Microsoft shows: Cowork gathers earnings reports, filings, analyst commentary, and news with an emphasis on primary sources, organizes findings with citations, and returns an executive summary formatted for email, a structured research memo with clear assumptions, and an Excel workbook with labeled tabs.

The GES analogue: GES is pitching to become GSC for an association’s annual convention. Chuck Grouzard’s exhibition sales team needs to know the show’s history, its growth trajectory, its stated sustainability commitments, who currently holds the contract, and what the association has publicly said about attendee experience.

Same three artifacts: an email-ready executive summary, a cited research memo with assumptions stated plainly, and a workbook with tabs for show history, exhibitor counts, square footage trend, and competitive incumbency.

1.9.49.4 “Create the launch plan” → The new service or product rollout¶

What Microsoft shows: Cowork builds a competitive comparison in Excel, distills differentiation into a value proposition document, generates a customer pitch deck, and outlines milestones, owners, and next steps.

The GES analogue: Rolling out a sustainable modular exhibit product line into a new region — the competitive comparison against other modular offerings, the value proposition document that sales can actually use, the exhibitor-facing pitch deck, and the milestone plan with owners across creative, operations, and the warehouse network.

1.1010. Nine GES Scenarios, Worked in Full¶

This is the practical heart of the chapter. Each scenario names the persona, the trigger, the badly-scoped prompt most people would write first, the properly structured five-part version, what Cowork does step by step, where it pauses, and what comes back.

Read the ones closest to your role carefully. Skim the rest — the patterns transfer.

1.10.1Scenario 1 — Post-show reconciliation package¶

Persona: Show Manager, GES Exhibitions Trigger: Move-out finished last night. The organizer wants a preliminary reconciliation before their board call in 36 hours. You have a flight in four hours.

Do the post-show reconciliation for the show that just ended.

Outcome: A preliminary post-show reconciliation package for Show 4412 
that the organizer can read before their board call on Thursday morning.

Inputs: The Show 4412 SharePoint library — specifically the on-site 
labor sheets, the material handling and freight manifest workbook, the 
exhibitor order change log, and the original labor forecast. Also my 
email thread with Karen at the organizer since May 1.

Definition of done: (1) An Excel workbook with tabs for labor 
forecast vs. actual, material handling weights vs. freight targets, 
and exhibitor order variances. (2) A two-page Word summary written 
for the organizer, not for internal ops. (3) A draft email to Karen 
with both files attached. Save all files to the Show 4412 library 
under /Post-Show.

Constraints: Use the FY27 reconciliation template. Flag any variance 
over 8% in a callout list rather than explaining it — I'll add the 
narrative. Do not contact any exhibitors. Do not include internal 
margin figures anywhere in the organizer-facing summary.

Approval scope: Create and save the files without asking. Ask me 
before sending anything to the organizer.

What Cowork does: Locates the Show 4412 library → opens the labor sheets and forecast → builds the variance analysis → opens the freight manifest and computes material handling weight against the agreed freight targets → reads the exhibitor order change log and reconciles against delivered orders → builds the workbook on the FY27 template → writes the two-page summary in organizer-appropriate language → drafts the email.

Where it pauses: At the email to Karen. You get a rich preview showing recipient, subject, body, and the two attachments.

What comes back: A reconciliation workbook, a two-page organizer-ready summary, and an email sitting in draft — reviewed on your phone at the gate, approved after you check the variance callouts and confirm no margin figures leaked into the summary.

1.10.2Scenario 2 — Move-in / move-out planning¶

Persona: Operations Director Trigger: A 620,000 sq ft show at a venue with tight dock access and a hard union jurisdiction boundary between freight and I&D labor. Move-in starts in eleven days.

Plan the move-in for the big show next month.

Outcome: A move-in and move-out plan for Show 5108 that the on-site 
crew leads and the marshaling yard supervisor can work from directly.

Inputs: The organizer's approved floor plan, the venue dock schedule, 
our labor forecast workbook, the target freight arrival windows for 
the advance warehouse, and last year's post-show report for this 
same show at this same venue.

Definition of done: (1) A Word operations plan with a day-by-day 
move-in sequence by hall, dock assignments, and crew allocation. 
(2) An Excel schedule showing carrier arrival windows against 
marshaling yard capacity by hour. (3) A one-page crew briefing 
document for the I&D leads. All saved to the Show 5108 library.

Constraints: Respect the union jurisdiction boundary — freight 
handling and booth install crews must be scheduled as separate 
resources, never interchangeable. Assume the advance warehouse 
closes to receiving 7 days before move-in. Flag any hour where 
marshaling yard demand exceeds capacity rather than smoothing it 
silently.

Approval scope: Files only. Do not post anything to the show 
Teams channel or email the venue — I'll circulate it myself.

What Cowork does: Reads the floor plan and hall breakdown → maps the dock schedule against hall access → pulls the labor forecast and allocates crews by trade respecting the jurisdiction constraint → models carrier arrivals against marshaling yard capacity hour by hour → surfaces last year’s documented problems from the post-show report and checks whether the new plan repeats them → produces the three artifacts.

Where it pauses: Nowhere externally — you scoped it to files only. It may ask one clarifying question about crew shift length.

What comes back: A plan, a capacity schedule with over-capacity hours flagged in red, and a crew briefing sheet. Your job is now the interesting part: deciding what to do about the three flagged hours on Tuesday morning.

1.10.3Scenario 3 — Site survey writeup¶

Persona: Show Operations Lead Trigger: You just walked a new venue for four hours. You have 60 photos, a OneNote page of scrawled notes, and a voice memo.

Write up my site survey.

Outcome: A venue site survey document for [Venue] that a show 
manager who has never been there could plan a move-in from.

Inputs: My OneNote site survey page from yesterday, the photos in 
my OneDrive /Site Surveys/[Venue] folder, and the venue's published 
exhibitor rules PDF I've attached.

Definition of done: A Word document saved to the Venue Knowledge 
Base library in SharePoint, following our standard site survey 
structure: access and docks, ceiling heights and rigging, floor 
load, power and utilities, labor and jurisdiction notes, marshaling 
and staging, known constraints, and open questions.

Constraints: Keep every observed fact separate from every inference 
— put anything you inferred rather than observed under a clearly 
labeled "To verify" heading. Do not fill gaps with assumptions from 
other venues.

Approval scope: Files only.

Where it pauses: File creation in a shared library may prompt for confirmation depending on your admin’s configuration.

What comes back: A structured survey document — and, critically, an explicit “To verify” list that tells you exactly what you failed to capture while you were there. That list is worth more than the document.

1.10.4Scenario 4 — Design brief to spec doc to client-ready concept deck¶

Persona: Creative Director / Designer, Spiro custom exhibits Trigger: An exhibitor client has approved the direction from a discovery call. You need a spec document for engineering and a concept deck for the client, and the client review is Friday.

Turn my design brief into a deck for the client.

Outcome: An engineering-ready spec document and a client-ready 
concept deck for the [Client] 40x50 island stand, both consistent 
with the approved design brief.

Inputs: The approved design brief document, my discovery call notes 
in OneNote, the client's brand guidelines PDF, the venue rigging and 
ceiling height constraints from the Show 5108 site survey, and our 
standard Spiro spec template.

Definition of done: (1) A Word spec document following the Spiro 
template — structure, materials, graphics, lighting, AV, storage, 
I&D assumptions, and open engineering questions. (2) A PowerPoint 
concept deck of 10–12 slides using our client presentation template: 
concept narrative, plan view, three key experience moments, materials 
and finish direction, and next steps. Both saved to the [Client] 
project folder.

Constraints: The stand must respect the venue's 22 ft rigging height 
limit and the show's 50% sightline rule on perimeter elements. Do not 
include any product names or launch messaging — the client's launch is 
under embargo until the show. Do not include cost estimates.

Approval scope: Files only. Nothing goes to the client from Cowork — 
Tony reviews before anything leaves.

What Cowork does: Reads the brief and discovery notes → extracts design intent → checks it against the venue constraints from the site survey → produces the spec document flagging where the brief is silent (which is exactly where engineering will push back) → builds the concept deck on the client template using brand guideline colors and typography → saves both.

Where it pauses: File creation only. Notice the constraint doing real work: the embargo instruction is the difference between a deck that’s safe to circulate internally and a deck that could genuinely damage a client relationship.

What comes back: A spec doc with an explicit list of unresolved engineering questions, and a concept deck at roughly the 80% mark — which is the right handoff point. The design judgment, the story arc, and the moment that makes the client lean forward are still yours.

1.10.5Scenario 5 — Organizer RFP response¶

Persona: Exhibition Sales Director Trigger: A major association has issued an RFP for GSC services for its annual convention. Response due in nine days. It’s 40 pages of requirements.

Write our response to this RFP.

Outcome: A complete first-draft response to the [Association] GSC 
RFP that our proposal team can edit rather than write.

Inputs: The RFP PDF I've attached; our three most recent winning GSC 
proposals in the Proposals library; the GES capability boilerplate 
library; the ESG and Net Zero roadmap materials; and the case study 
folder for association conventions of similar size.

Definition of done: (1) A Word response document structured section 
by section in the RFP's own order, with every requirement addressed 
and cross-referenced to the RFP section number. (2) An Excel 
compliance matrix listing every stated requirement, our response 
status, and the owner who needs to confirm it. (3) A short internal 
memo listing the five weakest sections and what evidence we'd need 
to strengthen them. All saved to the [Association] RFP folder.

Constraints: Do not invent any metric, client name, or case study — 
if the source material doesn't support a claim, write "[EVIDENCE 
NEEDED]" and move on. Do not commit to any pricing. Use the 
organizer's own terminology from the RFP, not our internal terms.

Approval scope: Files only. Nothing external.

What Cowork does: Parses the RFP into a structured requirement list → searches the proposals library for the strongest matching language → drafts each section in RFP order → builds the compliance matrix → runs a self-critique pass to identify weak sections → produces the internal memo.

Where it pauses: File creation. The [EVIDENCE NEEDED] constraint is the most important line in the prompt — it converts hallucination risk into a visible to-do list.

What comes back: Nine days of proposal work reduced to about two, plus a compliance matrix that stops the classic failure mode of an RFP response that reads beautifully and misses requirement 4.7.3.

1.10.6Scenario 6 — Stalled pipeline review and QBR prep¶

Persona: Account Director, Exhibition Sales Trigger: Quarter end is three weeks out. Your QBR is next Tuesday. You know some opportunities have gone quiet but you don’t know which or why.

Which of my deals are stalled?

Outcome: A ranked view of my at-risk opportunities with the specific 
follow-up that went cold on each, plus a QBR deck built on that 
analysis.

Inputs: My Outlook mail and calendar for the last 120 days; the 
account files in my Accounts SharePoint folder; the opportunity 
tracker workbook; and the last two QBR decks.

Definition of done: (1) An Excel workbook ranking open opportunities 
by risk, with columns for last meaningful contact, days since, the 
specific commitment or follow-up that lapsed, and the recommended 
next action. (2) A QBR PowerPoint on our standard template covering 
quarter performance, at-risk opportunities, and the recovery plan. 
(3) Draft follow-up emails for the top five at-risk accounts — 
drafts only, in my Drafts folder.

Constraints: "Stalled" means no substantive two-way exchange in 21 
days — a newsletter or an out-of-office does not count as contact. 
Do not send anything. Do not include any account where the client 
has explicitly told us they've gone to a competitor; list those 
separately.

Approval scope: Ask before creating anything in my Drafts folder. 
Never send.

What comes back: This is the direct GES analogue of one of Microsoft’s own reported customer outcomes — a sales lead who pointed Cowork at a stalled pipeline and got back a ranked list of at-risk opportunities with the exact follow-up that had gone cold on each, collapsing a week of manual review into a single morning.

The value is not the ranking. It is the “specific commitment that lapsed” column. You told the organizer you’d send the sustainability case study on May 3. You never did. Nobody’s CRM captures that. Your email does.

1.10.7Scenario 7 — Drayage and material handling cost analysis across shows¶

Persona: Logistics Manager Trigger: Three organizers in the same vertical have separately asked why material handling costs rose year over year. You suspect the answer differs by venue but you’ve never had time to prove it.

Analyze our drayage costs.

Outcome: An analysis of material handling and drayage cost variance 
across our shows at the four venues in the [Vertical] portfolio over 
the last 18 months, with the drivers identified.

Inputs: The freight and material handling workbooks in the Logistics 
SharePoint library for shows 4180 through 5210; the freight target 
agreements for each show; and the advance warehouse receiving logs.

Definition of done: (1) An Excel workbook with a tab per venue plus 
a consolidated comparison — cost per hundredweight, percentage of 
freight arriving in the advance warehouse window vs. direct-to-show, 
and freight target compliance rate. (2) A Word analysis of the top 
three cost drivers with evidence for each. (3) A six-slide deck I 
can use in an organizer conversation.

Constraints: Normalize for show size using net square footage — 
absolute cost comparisons across shows of different sizes are 
meaningless. Exclude the two shows with incomplete manifest data 
and note the exclusion. Do not include any organizer's rates in a 
deliverable that another organizer will see — build the deck with 
indexed figures, not absolute rates.

Approval scope: Files only.

What Cowork does: Opens each show’s freight workbook → normalizes by net square footage → computes cost per hundredweight and freight target compliance by venue → cross-references advance warehouse receiving logs to establish what proportion of freight arrived in-window → identifies correlation between direct-to-show percentage and cost → writes the analysis → builds the indexed deck.

Where it pauses: File creation. That confidentiality constraint — indexed figures rather than absolute rates — is a GES-specific instruction that Cowork would never infer on its own. This is exactly what the Constraints line exists for.

What comes back: The analysis nobody had time to run, with the answer to a question three clients asked. This is a heavy task. It is unambiguously worth the credits.

1.10.8Scenario 8 — Exhibitor service kit refresh and question triage¶

Persona: Exhibitor Services Manager Trigger: Two things at once. The service kit for next season needs updating across 40 pages of deadlines, rates, and venue rules. And your shared inbox is taking 200+ exhibitor questions a day during move-in.

Task A — the service kit:

Outcome: An updated exhibitor service kit for Show 5108 reflecting 
the new venue rules, the FY27 rate card, and the revised advance 
warehouse deadlines.

Inputs: Last year's Show 5108 service kit PDF; the FY27 rate card; 
the venue's updated exhibitor rules document; and the current show 
schedule with the advance warehouse and direct-to-show dates.

Definition of done: An updated Word document saved to the Show 5108 
library, plus a one-page change summary listing every change made 
and its source.

Constraints: Do not change any section where the source documents 
don't show a change — preserve existing wording exactly. Every 
deadline date must trace to the show schedule; flag any date you 
cannot verify rather than carrying it forward.

Approval scope: Files only.

The change summary is the whole point. It turns an unreviewable 40-page diff into a one-page review.

Task B — the question triage (an event-driven, scheduled pattern):

Outcome: A daily triage of the Show 5108 exhibitor services inbox 
so my team starts each morning with a prioritized queue instead 
of an undifferentiated pile.

Inputs: The Show 5108 shared exhibitor services mailbox, previous 
24 hours.

Definition of done: A Teams post to the Exhibitor Services channel 
each morning at 6:30 a.m. show-local time, grouping questions into: 
(1) deadline-critical — anything about move-in times, freight 
targets, or advance warehouse cutoffs; (2) orders and billing; 
(3) venue and rules; (4) answerable from the service kit — with 
the relevant kit section cited. Include a count per category and 
name the five most urgent.

Constraints: Do not reply to any exhibitor. Categorize only. 
Do not include exhibitor contact details in the Teams post.

Approval scope: Ask before the first post. Then only to the 
Exhibitor Services channel for the rest of the show.

Note the approval scoping in the last line — this is precisely what the More options → Only to feature is for. You approve once, scope it to one internal channel, and stop being interrupted, while everything external still stops for review.

1.10.9Scenario 9 — onPeak hotel block pickup analysis, and the ESG compilation¶

Persona A: onPeak Housing Manager Trigger: Room block pickup for a convention is at 61% with six weeks to cutoff. The organizer wants to know whether to release rooms.

Outcome: A hotel block pickup analysis for [Convention] with a 
recommendation on whether and how much to release, in a format 
I can send to the organizer.

Inputs: The pickup reports in the onPeak reporting library for this 
convention for the last three years; the current block contracts 
with attrition terms; and the current-year pickup data by property 
and rate tier.

Definition of done: (1) An Excel workbook comparing this year's 
pickup curve against the prior three years at the equivalent 
weeks-to-cutoff, by property and rate tier. (2) A two-page Word 
recommendation with the attrition exposure quantified per property. 
(3) A draft email to the organizer.

Constraints: Compare at equivalent weeks-to-cutoff, not calendar 
dates — the show moved two weeks later this year. Quantify attrition 
exposure using the actual contract terms per property, not a blended 
assumption. Do not recommend releasing rooms at any property where 
the attrition clause makes release more expensive than holding.

Approval scope: Ask before the organizer email. Files freely.

Persona B: Corporate Sustainability / ESG Reporting Lead Trigger: The annual ESG report and the EMEA Roadmap to Net Zero need current-period emissions data compiled from facilities, freight, and event-level sources across regions.

Outcome: A compiled emissions dataset and draft narrative section 
for the FY27 ESG report covering our reporting scope.

Inputs: The facility utility data workbooks for all 24 production 
and warehouse facilities; the freight and transport data in the 
Logistics library; the event-level emissions reporting from the 
EMEA program; and last year's published ESG report for structure 
and methodology continuity.

Definition of done: (1) An Excel workbook with a tab per data source 
and a consolidated summary using last year's methodology and units. 
(2) A Word draft of the emissions narrative section. (3) A data 
quality memo listing every gap, inconsistency, and unit mismatch 
found.

Constraints: Use exactly the same methodology and boundary 
definitions as last year's published report — flag any place where 
this year's source data doesn't support that methodology rather 
than substituting an alternative. Never estimate a missing figure; 
mark it as a gap.

Approval scope: Files only. Nothing leaves this session.

1.1111. Microsoft’s Customer Proof Points, in GES Terms¶

Three reported outcomes from Microsoft’s own customers translate directly.

1.1212. Why Cowork Fits GES Specifically¶

There is a reason this chapter exists in a GES book rather than being a generic Cowork overview.

Most software productivity features assume a knowledge worker at a desk with a laptop open. That describes a minority of GES.

GES runs 4,000+ live experiences a year, serves 150,000+ exhibitors, operates in 75+ countries, and runs 24 global production and warehouse facilities with roughly 2,600 people. The work happens on show floors, in marshaling yards, on warehouse docks, in venues at 5 a.m., and in airports between cities. The people who most need analytical and documentation work done are frequently the people least able to sit still and do it.

The property that makes Cowork disproportionately valuable here is the least glamorous one on the feature list: tasks keep running when your laptop is off.

Think about what that actually unlocks:

• A show manager delegates the reconciliation package at 5 p.m. during move-out, reviews it on their phone during a layover, and approves the organizer email from the gate.

• An EMEA account lead starts a deep research task on a prospective organizer at the end of the London day; a Las Vegas colleague picks up the finished memo at the start of theirs.

• A logistics manager kicks off a cross-venue drayage analysis before walking the floor and reads the result over lunch.

• A Spiro designer sets a spec document running before a client call and comes back to a draft with the engineering questions already flagged.

The quote from the top of this chapter is the mindset in one line: “It is easy to have a dozen tasks in flight at once, each one moving forward while you focus on what only you can do.”

For a company where the thing only you can do is often physically stand in a hall and make a judgment call, that sentence is not a productivity slogan. It is a description of how the job could work.

1.1313. Cautions — What Cowork Does Not Change¶

Everything above is genuine capability. This section is the counterweight, and it is not optional reading.

1.13.113.1 Cowork does not transfer accountability¶

This is the most important sentence in the chapter.

If Cowork drafts a number and you approve the email, it is your number. If Cowork produces a reconciliation and the organizer builds a board slide on it, that reconciliation is GES’s work product. If Cowork writes a spec and engineering builds to it, the spec is the designer’s.

There is no arrangement in which “Copilot generated it” is an acceptable answer to a client. Responsibility in T.R.U.E. means being responsible for our actions and delivering on our commitments. A delegated action is still your action.

1.13.213.2 Review before approving — properly¶

Skimming a preview is not reviewing it. A real review checks, in this order:

1. Recipients. Who is on this? Is anyone on it who shouldn’t be? Is this the client’s whole distribution list or just my contact?

2. Numbers. Spot-check at least two figures against the source. If they’re right, the method is probably right. If either is wrong, stop and re-scope.

3. Claims. Does anything assert something GES has not verified? Any commitment on timing, cost, or capability that hasn’t been approved?

4. Confidentiality. Is there anything in here that belongs to a different client, a different show, or a different internal audience?

5. Tone. Does this sound like GES — warm, precise, partnership-focused — or does it sound like software?

1.13.313.3 Oversharing risk¶

Cowork can search everything you have permission to see. That’s the feature. It is also the risk.

If you have broad access — many account directors and operations leaders do — a loosely scoped research task can pull material from one client’s show into a deliverable intended for another. Cowork is not doing anything wrong; it is doing what you asked, with the access you have.

The mitigation is entirely in your hands, and it’s the Inputs and Constraints lines of the five-part prompt:

• Name the specific libraries, folders, and date ranges. Don’t say “search our files.”

• State confidentiality boundaries explicitly. “Only use material from the Show 5108 library.” “Do not reference any other organizer’s rates.”

• When in doubt, run the task with narrow inputs and widen it if the result is thin. Widening is cheap. Un-sending is impossible.

1.13.413.4 Cowork is confident about things it should be uncertain about¶

Like every model-driven system, Cowork produces fluent output regardless of whether the underlying source material supported it. The countermeasure is structural, not vigilance-based — build it into the prompt:

• “Flag anything you cannot verify rather than carrying it forward.”

• “Write [EVIDENCE NEEDED] rather than inventing a figure.”

• “Separate observed facts from inferences under a clearly labeled heading.”

• “Never estimate a missing figure; mark it as a gap.”

Every one of those lines appeared in a scenario above. They are not decoration. They are the difference between a deliverable you can defend and one you can’t.

1.13.513.5 Some things should not be delegated at all¶

An apology to a client whose booth went up wrong. A conversation with a crew member about a safety incident. A negotiation with a union steward. The moment when an organizer needs to hear a human voice take responsibility.

Understanding — people come first — is the value that draws this line. Cowork can prepare you for those conversations. It should never have them for you.

1.1414. Getting Started — Your First Week with Cowork¶

1.14.114.1 Where to find it¶

1.14.214.2 Admin enablement¶

Cowork is enabled by administrators in the Microsoft 365 admin center. If you don’t see the toggle next to Chat, that is the first thing to check — it is an enablement question, not a licensing mystery. Admins also control which plugins are deployed org-wide, which is the right control point for anything that connects to an external data source or a system of record.

Two prerequisites, in order: the Microsoft 365 Copilot USL, then Cowork enablement plus a Copilot Credits allocation.

1.14.314.3 Three starter tasks for your first week¶

Do these in order. They’re deliberately sequenced from low stakes to real work.

Outcome: My calendar for next week organized so I have two 
protected two-hour blocks for focused work.

Inputs: My Outlook calendar for next week.

Definition of done: Focus blocks added; conflicts flagged in a 
summary message to me.

Constraints: Do not decline or move anything — just propose 
changes and tell me. Never touch anything with an external 
attendee.

Approval scope: Ask before any calendar change at all.

Outcome: A one-page catch-up on everything that's happened on 
Show [X] in the last two weeks.

Inputs: My email and Teams messages related to Show [X], and 
the Show [X] SharePoint library, last 14 days only.

Definition of done: A Word document saved to my OneDrive, 
organized by: decisions made, open issues, deadlines in the 
next 21 days, and anything waiting on me.

Constraints: Only the last 14 days. Cite the source message 
or file for every item so I can go check it.

Approval scope: Files only. Do not send or post anything.

1.1515. Try This: Convert Your Worst Recurring Task¶

Pick the task you dread most in your show cycle. The one that always slips. The one you do at 10 p.m. the night before it’s due.

Now write it as a five-part Cowork prompt. Be specific enough that a competent colleague who has never done your job could produce the right thing from your description alone.

Then answer three questions:

1. Which part was hardest to write? For most people it’s Definition of done — because we do these tasks by habit and have never articulated what “finished” means. That’s a useful thing to learn about your own work regardless of AI.

2. What constraint did you almost forget? There’s usually a confidentiality or template constraint that’s so obvious to you it never got said out loud. Those are the ones that cause problems.

3. What’s your approval scope? If the honest answer is “I want to check everything,” that’s fine for task one. Notice how it changes by task ten.

1.1616. Productive Struggle Problem¶

You are the GES account director for a major association client. Their annual convention closed nine days ago. Three things land in the same hour on a Tuesday morning:

• The organizer emails asking why material handling costs were 14% over the estimate, and requests an explanation “with data” by Thursday.

• Your VP asks for a QBR-ready view of the account, including renewal risk, by Friday.

• An exhibitor at that show has escalated a claim that their freight was misrouted through the marshaling yard, costing them a full day of install. Legal wants a factual timeline.

You have two days of actual working time, and you’re travelling on Wednesday.

The challenge: design your Cowork approach.

• Decide which of the three is a Cowork task, which is a Chat task, and which should not be delegated at all — and justify each choice using the decision rule from §1.

• Write the full five-part prompt for each task you’d delegate.

• Specify your approval scope for each and explain what you’d refuse to let Cowork do.

• Identify one confidentiality constraint that must appear in each prompt, given that the exhibitor claim may become a legal matter and the organizer is asking about costs that involve third-party carriers.

• Estimate whether each task is light, medium, or heavy, and say whether the credits are justified.

There is no single right answer. The quality of your reasoning about what you refused to delegate matters more than the prompts themselves.

1.17Glossary¶

1.18Discussion¶

Cowork changes what “doing the work” means. When the production of an artifact is delegated, the professional value shifts from the production to two things: the quality of the brief at the front, and the rigor of the review at the back.

Consider your own role at GES. Which parts of your work are genuinely craft — where your judgment, relationship knowledge, or on-the-ground read is the value — and which parts are assembly, where you are moving information between systems and formats because nobody else will?

Then consider the harder question: if the assembly work were delegated, would you actually use the recovered time on the craft work? Or would it fill with more assembly?

1.19Leader’s Takeaway¶

Every previous chapter in this book made your people faster at work they were already doing. This one changes what they do.

That is a bigger deal than a feature release, and it should be managed like one. Three things determine whether Cowork becomes an operational advantage at GES or an expensive experiment.

First: scoping is the skill, and it has to be taught. The five-part structure — Outcome, Inputs, Definition of done, Constraints, Approval scope — is not a nice-to-have. It is the difference between a deliverable your team ships and a generic document they quietly throw away. It is also, not incidentally, the primary cost-control lever. Teach it explicitly, build a shared prompt library, and treat a well-scoped prompt as the reusable asset it is.

Second: the approval habit is a culture question, not a training question. Somebody on your team will hit Approve All on a batch containing a client email. The only reliable defense is a culture where reviewing a preview properly is understood as professional behavior rather than friction — the same way checking a booth before the exhibitor walks in is understood as professional behavior. Trust means verify before you approve. Responsibility means you own every output that carries your name. Say both out loud, repeatedly, until they’re reflex.

Third: the value is in the work nobody currently has time for. The temptation is to use Cowork on the tasks you already do, and there’s real value there. But the larger return sits in the analyses that get skipped every busy season — the cross-show cost patterns, the venue history that lives in three people’s memories, the service kit comparison nobody has weeks for. Those are the tasks where Cowork doesn’t make an existing process faster; it makes a previously impossible process routine.

GES has just spent its first period as an independent company under Truelink, setting its own technology roadmap for the first time in 55 years. onPeak already shipped an AI Smart Suite. Visit by GES already ships intelligent hardware. The organization has proven it can build AI into products.

Cowork is the same capability turned inward — on the operational work of running 4,000 events a year with lean teams and immovable show dates.

The tool will keep improving. Cowork 1 is coming. More models will follow. What will not change is the underlying professional shift: from doing the work, to briefing the work and standing behind the result.

That has always been what excellent leadership looks like on a show floor. Now it’s what excellent individual work looks like too.