How the advisory team should use Copilot in Edge — to compress an afternoon of research into twenty minutes, interrogate PDFs without reading them cover-to-cover, and compose briefings from whatever's open in the browser. Six distinct Edge AI surfaces, six advisory workflows, and the one visual check that separates safe Arup work from unprotected data exposure.
Before the features, the reframe. Web research has always been a slow, sequential process — read page, take notes, close tab, repeat. Copilot in Edge turns that into a parallel one. The browser reads with you. Whatever's in the tab becomes queryable.
Think about how an advisory consultant usually researches a new topic. Open search. Click through ten blue links. Read each carefully. Copy-paste excerpts into a Word doc. Keep twenty tabs open for days. Try to remember which page said what. Eventually write something synthesising it all, losing most of what you read along the way.
"The real value here isn't speed, it's selectivity. A research task rarely involves reading one source thoroughly — it involves scanning many sources and going deeper into the ones that matter. Copilot makes that dramatically faster."
Copilot in Edge doesn't replace reading — it restructures the sequence. Three specific shifts:
Old pattern: read every page that looked vaguely relevant, hoping something useful emerged. New pattern: open 10-12 tabs in parallel, sidebar-summarise each in 30 seconds, close the ones that aren't useful, deep-dive only the ones that are. The afternoon research session becomes a 20-minute session with better output because your reading time is spent on the sources that actually matter.
Old pattern: the page is the interface. You read whatever the author put in whatever order. New pattern: the sidebar is the interface. Ask specifically what you want to know about the page — "what does this say about procurement frameworks for transport authorities?" — and get a targeted answer. You interrogate the page; you don't just receive it.
Old pattern: to compare three websites, you'd open them in sequence and take notes, then manually build a comparison. New pattern: with Copilot Mode, Copilot reads all your open tabs at once and produces a synthesised comparison. The browser becomes a research assistant that can read everything simultaneously — the closest thing to delegation in the Microsoft stack.
For Arup advisory work — where engagements routinely require reading regulatory documents, client annual reports, competitor websites, industry research, and government guidance — the compounding return is enormous. A consultant who builds these habits produces more-informed deliverables in a fraction of the time. The research time doesn't just shrink; the quality of what you actually know at the end of the session goes up.
"Copilot in Edge" covers six related but distinct capabilities. Page summarise is the one everyone knows. The other five — PDF analysis, multi-tab reasoning, video summarise, Compose, and right-click Rewrite — are where the research workflow really lives.
Ctrl+Shift+. to open the sidebar. Type "Summarise this page" for a default structured overview, or ask a specific question — "What does this say about procurement constraints?", "What are the author's three main recommendations?", "Is this source primary or secondary?" — and get a targeted answer. Citations link back to specific sections of the page. Works on articles, reports, blog posts, government guidance, news sites. First use requires permission — Edge asks if Copilot can access page content. If your org has disabled page-context access, the sidebar works but answers aren't grounded in the page. For prompts that don't need page content (like "help me plan a trip"), the URL, page title, and history are the only things shared.aka.ms/copilot-mode). Once enabled and after you grant permission, Copilot can access the content of all your open tabs at once. Open four competitor websites and ask "compare these four in a table with columns for target market, core services, pricing, and delivery model" — get a synthesised comparison rather than four separate summaries. Works for: competitive scans, framework comparisons, cross-referencing multiple sources on the same question, assembling itinerary-like outputs from multiple inputs. Visual indicators appear whenever Copilot accesses page content so you always know when multi-tab analysis is happening. You can disable Copilot Mode at any time.The single most important visual check in this whole chapter. Copilot in Edge has two modes — consumer and enterprise — and they look almost identical. One is safe for Arup work. The other quietly isn't. The only difference visible to you is the green shield icon at the top of the sidebar.
When you open the Copilot sidebar in Edge, look for the green shield icon at the top of the pane. That icon means you're signed in with your Microsoft Entra work account and Enterprise Data Protection (EDP) is active.
EDP means: prompts and responses are not used to train foundation models. Content is handled under your organisation's compliance policies. Enterprise-grade security protections apply to anything you ask Copilot about.
No green shield means you're in consumer mode — the same privacy regime as using free Copilot Chat on a personal account. Treat it the same way you'd treat ChatGPT on a personal laptop: useful for generic queries, not safe for anything Arup-confidential.
The green shield appears when all three of the following are true:
If any of those isn't met, you're in consumer mode. Signing into Edge with a personal Microsoft account — even on an Arup laptop — does not give you EDP. You have to use your work profile specifically.
The most common mistake: consultants who use Edge for personal browsing slip into personal profile for a search, then forget to switch back before asking Copilot about work-related content. The shield is gone and they don't notice. Check the shield every time you ask Copilot anything about Arup, clients, or any sensitive material. If no shield, switch to work profile first; or don't use Copilot at all for that query.
In addition to EDP, Arup's Microsoft Purview Data Loss Prevention policies apply to Copilot in Edge on work profile. If a page is subject to DLP (sensitive client extranet, restricted SharePoint, confidential document portal), Copilot won't be able to summarise it — DLP blocks the access. This is working as designed, even when it's inconvenient. Don't treat DLP blocks as a problem to work around; they're a signal that the content isn't appropriate for Copilot processing anyway.
Edge Copilot has lower prerequisites than most chapters in this playbook — but the ones it has are structural. Five minutes of setup on day one saves the team from the preventable mistakes.
1. Confirm the work profile. Open Edge. Click your profile icon (top-right). Make sure your Arup/Entra work profile is active — not "Personal" and not "Guest". If you see multiple profiles, pin the work profile as default. This single step ensures EDP applies.
2. Verify the green shield. Open the Copilot sidebar (Ctrl+Shift+.). Confirm the green shield icon at the top of the pane. If not, something is wrong with profile or policy — don't use Copilot for Arup work until fixed.
3. Grant page-context permission deliberately. On first use, Copilot asks whether it can access page content. For any page where you want page-grounded answers, grant it. You can revoke per-page if a specific page is sensitive. For the research workflows in this chapter, page-context permission is essential — without it, Copilot answers from general knowledge not from what you're reading.
4. Decide on Copilot Mode. Settings → AI innovations → Copilot Mode. Enabling it gives you multi-tab reasoning (Use Case 3) plus the revised New Tab Page layout. The multi-tab capability is genuinely useful for advisory research. The New Tab Page changes are cosmetic and some consultants dislike them — you can enable multi-tab without committing to the interface changes via the granular settings under AI innovations.
5. Review the privacy toggles. Under Edge sidebar → "…" menu → Settings → Privacy:
Once these are set, you rarely touch them again. Five minutes up front, the whole team benefits for the rest of the year.
The core Edge Copilot workflow. The one that changes how research actually gets done. Open everything in parallel; sidebar-summarise each in 30 seconds; close the useless ones; deep-dive the useful ones. Afternoon becomes half an hour with better output.
The reason advisory consultants' research tends to sprawl is that they read every page in full because they're worried they'll miss something. The Copilot-assisted triage inverts that: the sidebar tells you in 30 seconds whether the page is worth reading, so you can afford to close the ones that aren't. The anxiety-reducing function of the sidebar matters as much as the time-saving function. You're not skimping — you're letting Copilot identify the signal so your reading time can focus there.
Once you've triaged the 10-12 tabs and deep-dived the 3-4 useful ones, switch to the Compose tab in the sidebar. Ask for a briefing note based on everything you've been reading — Compose inherits session context. See Use Case 5 for the pattern. The whole research-to-briefing loop fits in under 30 minutes.
The advisory-team-specific workflow. Long regulatory documents, industry standards, government guidance — 100-300 page PDFs that used to mean an afternoon of reading. Drag into Edge, interrogate with narrow targeted questions, get what you need in minutes.
Every stage does something the others don't:
For any claim you're going to rely on in a client deliverable or a commercial decision, navigate to the cited section manually and read it in full. Copilot's summaries are excellent starting points but compress nuance, and for regulatory or legal content, nuance is often the whole point. Treat Copilot as the research assistant that points you at the right paragraph; you're still the one who reads it carefully.
Findings that matter should move out of the ephemeral browser session and into something durable. Options:
Edge's unique trick. Open three or four competitor websites, platform pages, or regulatory frameworks in separate tabs. With Copilot Mode on, ask for a synthesised comparison. Get a structured side-by-side rather than four separate summaries — and the table drops straight into a deck.
1. Consultancy competitive scan. Five tabs — Arup, Mott MacDonald, Atkins, Jacobs, WSP. Prompt for comparison on advisory services focus, target client type, digital/BIM capability, sustainability positioning, notable recent engagements. Output drops into an internal competitive analysis deck.
2. BIM framework comparison. Three tabs — UK ISO 19650, PAS 1192 legacy, and a client's own BIM standards document. Prompt for a comparison on common information management principles, differences in how they treat collaboration, evolution over time, and what's most relevant for a transport authority. Output becomes the briefing for the project team.
3. Power Platform vs alternatives. Four tabs — Microsoft Power Platform, ServiceNow, Salesforce, Mendix. Prompt for a comparison on target use cases, enterprise integration, low-code vs pro-code balance, pricing structure, and Arup-relevant deployment patterns. Output feeds a client platform-recommendation document.
Multi-tab comparisons look authoritative — which is both their strength and their risk. Always check any cell that makes a material difference in your recommendation against the original source. Copilot occasionally conflates content from adjacent tabs or imports plausible-sounding but incorrect information. For anything going to a client, spot-check at least one row per tab before the comparison leaves your desk.
Conference keynotes, analyst briefings, industry explainers, client executives' media appearances — often 30-60 minutes of video that nobody has time for. If a transcript is available, Copilot reads it and summarises the whole thing in minutes. Information density finally matches your calendar.
Copilot's video summarise feature reads the transcript — so anything with an available transcript works:
Pure visual content (silent demos, workflow recordings without narration) won't work. Nor will videos where the transcript has been deliberately blocked. If Copilot's summary feels vague, check whether the transcript is available — the video platform usually shows a "CC" or transcript option somewhere.
After summarising a video, switch to the Compose tab and ask for a briefing note based on the video. The output inherits the context — so you get a prose summary of the video's content with the right framing. Useful for: internal knowledge-share posts, brief-the-team emails, pre-reads for meetings where the video's content is context.
The payoff step. You've triaged the sources, summarised the PDFs, compared across tabs, absorbed the video. Don't lose it all to a blank Word doc — use Compose in the sidebar to produce the first draft while everything's still loaded in Copilot's session context.
Both generate written content grounded in research. The difference:
/filename syntax.The handoff pattern: use Compose to get the first draft out while research is loaded; paste into Word; switch to Word Copilot for proper formatting, structure, and final polish. You get the speed of Compose's session-grounded output and the rigour of Word's template-aware editing.
Compose's output quality depends on what's actually in Copilot's session memory. If you've been asking questions in Chat about the research, that conversation is context for Compose. If you've closed the tabs, that context fades. Compose works best immediately after a research session — not the next day when the session is gone. If you need to compose from research done earlier, re-open the tabs and run a quick summary of each to re-seed the context, then Compose.
The small feature used thirty times a day. Any editable text field in any web page — right-click, rewrite with Copilot. LinkedIn posts, web-based portals, third-party forms, in-browser chat, procurement responses. Eliminates the copy-to-Word round trip that most consultants still do.
LinkedIn drafting. Advisory consultants often want to share professional content on LinkedIn but don't want to fully switch to a writing tool for a 3-paragraph post. Draft rough in LinkedIn's compose box; select all; right-click → Copilot → Rewrite; adjust for "Professional tone, concise, confident voice." Review and post. Three-minute workflow replaces the 15-minute "draft in Word, polish, paste back" loop.
Procurement portal responses. Many public-sector procurement portals have character-limited free-text fields where the quality of writing matters enormously. Draft in the field; Rewrite for concision within the character limit; the output fits where your original didn't.
Web-based ticket comments. JIRA, Azure DevOps, Monday — any ticketing system where your comment will be read by multiple stakeholders. Quick Rewrite to turn "please look at this it's broken" into "this is blocked by [clear statement of issue]; could you advise on [specific question]?" Small improvement, multiplied across dozens of tickets a week.
In-browser email replies. For consultants using Outlook Web App or other browser-based email clients, the right-click Rewrite works identically to the Draft with Copilot discussed in Part VIII — but without needing to open the full Copilot pane.
One failure mode to avoid: over-formalising casual channels. If you're in a team Slack channel where the norm is "hey all, quick q" and you Rewrite every message into "Dear colleagues, I have a quick question," you sound tone-deaf. Match the rewrite to the context. Professional portals get professional tone; team chat stays informal.
Six use cases, six tools, one table.
| Your scenario | Use this | Surface | Time |
|---|---|---|---|
| Quick summary of a web page or article | Sidebar page summarise | Edge sidebar | ~30 sec |
| 10-12 sources on a new research topic | Parallel triage + deep-dive pattern | Multiple tabs | ~20 min |
| 100+ page regulatory PDF | 4-stage PDF deep-dive | PDF in Edge | ~15 min |
| Compare 3-4 websites or frameworks | Multi-tab Copilot Mode | Copilot Mode | ~5 min |
| 60-minute conference keynote on YouTube | Video summarise | Edge sidebar | ~3 min |
| Turn research into a 400-word briefing note | Compose tab with session context | Compose | ~5-10 min |
| LinkedIn post needs professional polish | Right-click Rewrite | LinkedIn compose | ~30 sec |
| Procurement portal free-text response | Right-click Rewrite (for concision) | Portal field | ~1 min |
| Find quotable claims in a long report | PDF deep-dive · verbatim quote prompt | PDF in Edge | ~2 min |
| Quick sense-check: is this page relevant? | Sidebar · relevance check prompt | Single tab | ~30 sec |
| Client annual report before a meeting | PDF deep-dive + Compose brief | PDF in Edge | ~20 min |
| Cross-reference multiple research sources | Multi-tab comparison + verify | Copilot Mode | ~10 min |
Patterns from training advisory teams on Copilot in Edge. Almost every bad outcome maps to one of these.