RESEARCH REPORT PART XV APR 2026 ARUP DIGITAL ADVISORY WORKFLOWS

Researcher +
Analyst.

The two named reasoning agents every M365 Copilot Premium user has and most never use. One does deep multi-step research across work + web. One runs Python to analyse data files. Both take longer than standard Copilot — and deliver output standard Copilot can't. They share a monthly query allowance; the discipline of spending it well is the whole skill.

The premise
Researcher and Analyst aren't "better Copilot Chat." They're structurally different — reasoning agents that take minutes rather than seconds, produce decision-grade outputs rather than drafts, and work across evidence in ways standard Copilot can't. For advisory consultants specifically — where deliverables are often "what does the evidence say?" — these are the two agents that most directly map to billable work. The shared monthly query allowance forces discipline. This chapter is about spending those queries where the return is largest.
2 agents
Pre-pinned
in the Copilot app
Shared
Monthly query allowance
(check admin centre)
Critique
Multi-model review
(GPT + Claude)
Python
Analyst's real
code execution
01 / CORE SHIFT

Reasoning agents vs prompting.

Before the features, the reframe. Everything else in this playbook has been about prompting — you ask, Copilot answers. Researcher and Analyst are different. You commission; the agent plans, retrieves, reasons, reviews, and delivers. The difference isn't speed; it's shape of output.

Think about how you'd ask a senior colleague for help with two different tasks. Task 1: "Can you tell me what the last ORR guidance said about rail accessibility?" They tell you in a sentence. Task 2: "Can you prepare a briefing on the UK transport sector's current approach to digital asset management, covering main operators, regulatory context, industry research, and Arup-relevant angles?" Different request. They'd go away for an afternoon, come back with something substantive. Researcher and Analyst are the AI equivalents of the second kind of request.

"Researcher is designed for deeper reasoning of your tasks. By taking longer processing time, Researcher delivers a more comprehensive output, making it ideal for detailed research projects."

Three specific shifts this produces:

From seconds to minutes

Standard Copilot Chat returns answers in 2-30 seconds. Researcher and Analyst take 5-15 minutes to complete a query — sometimes longer for complex requests. That isn't a bug; it's the mechanism. During those minutes, the agent is planning the approach, retrieving sources iteratively, reasoning over what it finds, cross-checking, drafting, and revising. The output shows the work that standard Copilot can't produce because it's optimised for speed.

From answers to reports

Standard Copilot produces a paragraph or short bulleted answer. Researcher produces a multi-page structured report with sections, citations, visual elements, sometimes charts. Analyst produces a reasoned analysis with the Python code visible, intermediate results, and business-language interpretation. These are decision-grade artefacts — the kind you'd attach to a client briefing, paste into a Word document, or present to a team.

From tool-you-use to tool-you-commission

The shared monthly quota changes the relationship. You don't fire off a Researcher query for something you could get from standard Copilot in 10 seconds — that's a wasted query. You save them for tasks where a reasoning agent's depth and rigour genuinely justify the cost. Each query becomes a small commissioning decision. The discipline forces better use.

For Arup advisory specifically, the two agents map directly to the two most common billable activities: researching a topic to produce a deliverable (Researcher) and analysing data to produce an insight (Analyst). These aren't marginal productivity boosts — they're agents that do work consultants would otherwise spend half-days on.

Why these matter disproportionately for advisory
Parts V-XIV covered Copilot features that make existing work faster. Researcher and Analyst go further: they produce work that a consultant might otherwise have to do entirely from scratch. A market landscape report. A sector due-diligence summary. A data analysis against client spreadsheets. For consultants whose billable output includes structured research and quantitative analysis, these two agents change the shape of the working day more than almost anything else in this playbook. They're also the features that most directly justify the M365 Copilot (Premium) licence for knowledge workers. Researcher replaces hours of manual desk research. Analyst replaces the need to export data to Python notebooks or hire ad-hoc data analysts for one-off queries.
02 / THE SIX SURFACES

Six distinct named-agent capabilities.

Two agents, six distinct capabilities worth knowing. Core Researcher and Analyst are the starting points; the additional capabilities (Critique, Computer Use, output conversion, the shared quota) shape how you actually use them.

FlagshipOpenAI deep research modelMultilingual
1 · Researcher — the deep research agent
Pre-pinned in the M365 Copilot app under Agents. Powered by OpenAI's deep research model combined with Microsoft 365 Copilot's orchestration and deep search capabilities. Reasons across your work data — emails, meetings, files, chats, SharePoint — and the web, integrating both into a single structured report. Can pull from third-party connectors (Salesforce, ServiceNow, Confluence) to bring in data from systems outside M365. Output is a multi-section report with citations, visual elements, sometimes charts. May ask clarifying questions before starting, especially for ambiguous requests. Think of it as commissioning a junior analyst — give it a specific brief with context, constraints, and desired structure; come back in 10 minutes to a usable draft.
Where it shines at Arup
  • Sector / market landscape briefs for engagement kickoffs
  • Regulatory research (policy, guidance, standards)
  • Competitive analysis of firms in a market
  • Client due-diligence and background research
  • Thought-leadership research for LinkedIn / internal briefings
  • RFP response research on the target client
Effective commissioning
  • Be specific about topic, scope, and audience
  • Define search scope — work data, web, or both
  • Specify desired structure (sections, length)
  • Engage clarifying questions when they come up
  • Allow 5-15 minutes — don't abort early
March 2026Multi-modelQuality upgrade
2 · Researcher with Critique & Council
Researcher uses multi-model intelligence — a generation model plans research and produces a draft, then a separate review model (typically Claude) acts as expert reviewer, checking accuracy, completeness, and citation integrity before the report is delivered. This is the Critique architecture: GPT drafts, Claude reviews. Council is a related capability showing multiple model responses side-by-side with a cover letter on where they agree, diverge, and what each brings. Both are broadly available in the Frontier program. For Arup advisory this means: the quality of Researcher output is meaningfully better than single-model deep research, with multi-model review catching single-model failure modes.
Why this matters
  • Single-model research has consistent blind spots
  • Multi-model review catches factual errors and citation misattribution
  • Council shows where models genuinely agree vs disagree — useful for contested topics
  • Reduces manual fact-checking burden on consultants
  • Ahead of Perplexity, Gemini, and standalone OpenAI deep research on benchmarks
Practical notes
  • Check if your tenant is in Frontier for Council access
  • Critique runs in background — you see final output, not the review
  • Claude integration requires Anthropic-as-subprocessor enabled
  • Still verify critical claims before commercial use
Frontier programBrowser + terminalSecure virtual env
3 · Researcher with Computer Use
In Researcher's home screen, select Computer Use (Frontier program only). Researcher operates in a secure virtual environment where it can: access gated content (sites behind logins or paywalls, with your approved credentials); navigate real interfaces — click, type, interact with websites while you stay in control; run code via a secure terminal for data analysis; tailor reports to your work — create presentations, spreadsheets, or applications using code generation. Watch the desktop view as Researcher works. Confirm actions when prompted. Never bypasses organisational security controls. For advisory research that needs to reach beyond public web pages — regulatory portals requiring login, subscription databases, sector-specific industry tools — this extends Researcher's reach substantially.
Good Computer Use scenarios
  • Research requiring gated industry databases (Gartner, Forrester)
  • Regulatory research from login-gated portals
  • Client meeting prep pulling from multiple web sources
  • Analysis tasks needing script execution alongside research
  • Generating presentations from research findings
Governance notes
  • Frontier program enrolment required
  • Admin must enable agent access
  • You approve credential use — don't approve what you wouldn't manually
  • Organisational DLP and compliance still apply
FlagshipOpenAI reasoning modelPython execution
4 · Analyst — the Python data scientist
Pre-pinned in the M365 Copilot app under Agents. Powered by an OpenAI reasoning model, optimised for data analysis. Attach data files (Excel, CSV, PDF, XML, PowerPoint, databases) via the + icon; ask a question in plain language. Analyst uses chain-of-thought reasoning to plan the analysis, writes Python code to execute it, runs the code, interprets results into business-language insights. The Python code is visible in real time — check methodology, verify logic, catch errors. Output is a structured analysis with insights in prose, visual elements (charts, tables), and specific findings. Handles statistical analysis, trend detection, outlier identification, forecasting, what-if analysis, visualisation generation. Supports 8 languages (expanding). Takes 3-10 minutes typically.
What it does well
  • Multi-spreadsheet consolidation and analysis
  • Statistical analysis (means, distributions, correlations)
  • Time-series analysis and basic forecasting
  • Outlier detection and segmentation
  • Chart generation for presentations
  • What-if scenario comparison
  • Data cleaning and structural analysis
Check the Python
  • Code is visible — read it for anything commercial
  • Watch for mismatched column names / data misinterpretation
  • Verify calculation logic, especially for financial data
  • Don't trust the headline number; trust the code behind it
March 2026Format conversionNo re-work
5 · Output format conversion
Since March 2026, Researcher reports support multiple output formats. Once a report is produced, convert it into: PowerPoint presentation (slides with report structure, visuals preserved — for meetings and briefings); PDF (formatted document for sharing or archiving); infographic (one-page visual summary, uses Create module generation — see Part XIV); audio overview (podcast-style narration, for on-the-go consumption, similar pattern to Copilot Notebooks Audio Overview — see Part X). Content isn't regenerated — same reasoning and citations flow into whichever format you pick. Doesn't consume an additional query — conversion is part of the original Researcher query's output. This closes the last-mile gap between "Researcher produced a good report" and "I have a deliverable for my meeting."
Format → audience mapping
  • PowerPoint — client meetings, internal briefings, workshops
  • PDF — formal documentation, archival, external sharing
  • Infographic — LinkedIn, one-pagers, executive summaries
  • Audio overview — commute consumption, team listening, accessibility
Practical workflow
  • Run Researcher query → get report
  • Review content for accuracy first
  • Convert to primary format needed
  • Optionally convert to secondary format for reuse
  • Edit converted output as needed
Structural constraintShared poolMonthly reset
6 · The shared monthly quota
Every user with an M365 Copilot (Premium) licence gets a shared monthly query allowance across Researcher and Analyst. The pool is shared — using it for Researcher reduces what's available for Analyst, and vice versa. Resets at the start of each calendar month. Check the current limit in the M365 admin centre — Microsoft adjusts the cap as the agents evolve. This quota shapes everything: frivolous queries waste budget that mattered. A good Researcher or Analyst query takes 5-15 minutes of compute and produces a substantive report; a casual query takes the same allocation and produces content you could have got from standard Copilot in 10 seconds. The skill is reserving queries for tasks where the agent's depth is the feature. See Use Case 6 for explicit budgeting strategies.
Quota-aware habits
  • Plan queries weekly, not ad-hoc
  • Pre-draft prompts carefully before submitting
  • Save reports to OneDrive / Notebooks for reuse
  • Use standard Copilot for anything that fits
  • Reserve named-agent queries for >30-min-of-work tasks
What doesn't count
  • Standard Copilot Chat prompts (unlimited)
  • Copilot in Word/Excel/PowerPoint/Teams (unlimited)
  • Agent Builder agents you or others built
  • Output format conversion of an existing report
  • Re-reading or sharing past reports
The difference from Agent Builder and Copilot Studio
Researcher and Analyst are pre-built agents made by Microsoft, available to every M365 Copilot Premium user. Agent Builder and Copilot Studio (covered in Part IV) are for creating your own agents — custom assistants tailored to specific workflows. The distinction is "use" vs "build." Researcher and Analyst are used as-is; Agent Builder / Studio let you create new ones. An advisory team likely wants both: named agents for generic research and analysis patterns, plus custom agents for Arup-specific workflows (e.g. an "Arup Engagement Brief Agent" that knows the team's template and produces briefs in Arup's format). Don't confuse the categories — they're genuinely different tools.
03 / SIDE-BY-SIDE

Researcher vs Analyst.

Two named agents. Similar framing (deep reasoning, minutes-to-answer, pre-pinned in the app). Very different capabilities. Knowing which to commission for which task is the primary skill.

Agent 1 · Deep research

Researcher

Model: OpenAI's deep research model, with Critique multi-model review.

Works on: web + work content (emails, files, meetings, chats) + third-party connectors.

Output: multi-section structured report with citations. Convertible to PowerPoint / PDF / infographic / audio.

Takes: 5-15 minutes typically. Longer for complex topics.

Languages: multiple supported — check current coverage in M365 Copilot.

  • Sector / market landscape reports
  • Regulatory research and policy analysis
  • Competitive analysis and benchmarking
  • Client background / due-diligence
  • Thought-leadership research
  • RFP response context-gathering
Agent 2 · Data analysis

Analyst

Model: an OpenAI reasoning model, optimised for analysis.

Works on: data files you attach (Excel, CSV, PDF, XML, PowerPoint). Also M365 Graph data.

Output: reasoned analysis with Python code visible, insights in prose, charts, tables.

Takes: 3-10 minutes typically.

Languages: 8 supported (expanding).

  • Multi-spreadsheet consolidation analysis
  • Statistical analysis of client data
  • Trend detection and outlier analysis
  • Time-series / forecasting from historical data
  • Scenario comparison (what-if analysis)
  • Data-driven chart generation

The decision rule

A simple rule that captures 95% of cases:

  • If the task is "tell me about X" → Researcher. You want synthesis from multiple sources. The output is prose.
  • If the task is "what does this data say?" → Analyst. You have a data file. The output is analysis.

The edge cases where both might apply:

  • "Tell me about this data and its context" — Researcher with the file attached; you want meaning, not numbers.
  • "Analyse this spreadsheet against industry benchmarks" — Analyst for the data; Researcher for the benchmarks; synthesise.
  • "Forecast this trend and explain why it's happening" — Analyst for the forecast; Researcher for the explanation.

They're complementary, not interchangeable

The strongest advisory workflows often use both agents sequentially. Typical pattern: Analyst crunches the client's data to identify trends; Researcher investigates what's driving those trends in the broader market; you synthesise the two into a client briefing. Two quota queries for a deliverable that used to take two days; one day's turnaround is still faster than most competitors.

The mental model that makes commissioning easier
Think of Researcher as a sharp junior analyst — give them a brief, they come back with a well-structured, well-cited report. Think of Analyst as a data scientist with a laptop — give them a data file and a question, they come back with analysis and show you the code. The choice between them is the choice you'd make between commissioning those two different colleagues. If you wouldn't ask the data scientist for a market landscape report, don't ask Analyst. If you wouldn't ask the research analyst to forecast from a spreadsheet, don't ask Researcher. Match the agent to the colleague you'd have asked.
04 / PREREQUISITES

Licence, quota, expectations.

Fewer technical prerequisites than most chapters — the named agents work out of the box for any M365 Copilot (Premium) user. But the quota constraint and the "this is not standard Copilot" expectation need explicit setting.

The five-point setup

1. Confirm M365 Copilot (Premium) licence. Researcher and Analyst are only available with a full M365 Copilot licence. Copilot Chat (Basic) users cannot access them. Check at copilot.cloud.microsoft — label under your name should read M365 Copilot (Premium). See Part XII for full tier discussion.

2. Verify admin enablement. Copilot administrators can enable or disable Researcher and Analyst at tenant level. If you don't see them in the Copilot app's Agents section, check with IT — it may be a policy setting rather than a licence gap.

3. For Critique and Computer Use, confirm Frontier program status. Multi-model Critique and Computer Use require tenant enrolment in the Frontier program. If enrolled, these capabilities are available; if not, you get standard Researcher. Frontier enrolment is an admin decision.

4. Understand the quota. Researcher and Analyst share a monthly query allowance — confirm the current cap in the M365 admin centre, as Microsoft adjusts it. Resets at the start of each calendar month. Count is not currently surfaced in-product — you're informed when close to or at the limit. Plan accordingly.

5. Set expectations on timing. Researcher and Analyst take 5-15 minutes per query. That is not slow — it's the mechanism. Submit the query, do other work, come back. Don't abort mid-query; don't re-submit while waiting. Consultants new to these agents often don't realise they're supposed to leave the query and return. Setting expectations up front saves frustration.

The Anthropic subprocessor setting
For Researcher with Critique to use Claude as the reviewer model, Anthropic as a subprocessor must be enabled at the tenant level. Being rolled out gradually; full availability expected by end of March 2026 (so likely live in most tenants by now but worth confirming). If not enabled, Researcher still works in single-model mode but loses the Critique multi-model review. Admin setting; worth confirming with IT if Researcher output seems to have the old single-model characteristics rather than the new multi-model polish. See Microsoft Learn documentation on "Use Claude with Researcher in Microsoft 365 Copilot" for full admin configuration.
05 / USE CASE 1

Researcher for engagement kickoff.

The highest-value default use of Researcher for advisory consultants. Day 1 of a new engagement. Brief the agent with client, scope, and specific questions. Come back to a multi-page report. An afternoon's research becomes 15 minutes.

~30 min totalLow effort1 query · high valueEngagement kickoff
Commission the engagement brief
Use this when
Day 1 of any engagement where you need substantive context. Typical scenarios: new client engagement requiring sector research; proposal response requiring industry and competitor analysis; due-diligence assignment; sector-entry work for a practice area; onboarding to a new engagement team.
01
Copilot app → Agents → Researcher
02
Write structured brief (see template below)
03
Select scope — work + web for most cases
04
Answer clarifying questions if asked
05
Wait 10-15 min; review; convert to deliverable

The commissioning template

The single biggest difference between a good Researcher output and a poor one is the specificity of the brief. Vague prompt, vague report. Structured brief, structured report.

Commissioning brief for this research: ## Context I'm an Arup advisory consultant starting an engagement with [client — e.g. "a UK metropolitan transport authority"]. The engagement is about [scope — e.g. "digital advisory support for their estates modernisation programme, specifically BIM Level 2 compliance and lifecycle asset management"]. I need deep background before the kickoff meeting on [date]. ## What I need Produce a structured briefing covering: 1. **The client organisation** — recent developments, current leadership, strategic priorities, any publicly-available budget signals, known governance issues 2. **Sector / regulatory context** — current UK regulatory environment for [area], recent policy changes, relevant standards (ISO, PAS, DfT), what's changing in the next 12 months 3. **What peers are doing** — how comparable organisations are approaching similar challenges, notable exemplars, common approaches, what's working 4. **Arup's relevant angle** — past Arup work with this client or similar ones (from my work data if available), Arup capabilities directly relevant, sensible positioning for the engagement 5. **Key kickoff questions** — the non-obvious questions a sharp consultant would raise based on this research ## Sources - Use both web and work data - Cite specific sources for every substantive claim - Flag where sources disagree - Note areas where information is thin or speculative ## Output Structured report suitable for converting into slides for the internal pre-kickoff briefing. Professional tone, advisory-consultant audience. Aim for 4-6 pages of dense, well-sourced content.

The review pass

Even with Critique multi-model review, verify critical claims before relying on them commercially. Three checks:

  • Sources — are they authoritative? Click through to 2-3 to confirm they say what the report claims.
  • Specificity — are numbers, dates, quotes exact? AI occasionally misattributes or paraphrases inaccurately.
  • Gaps — is anything obviously missing? Your domain knowledge may catch things the agent didn't.

Budget 15-20 minutes for review. Total time: 30-45 minutes for a report that used to take half a day.

Convert to deliverable

Once reviewed, use the March 2026 format conversion: PowerPoint for internal pre-kickoff briefing; PDF for the engagement's Loop workspace archive; infographic for a one-page client-facing summary; audio overview if you'll listen during your commute to the kickoff.

The value proposition per query
This is the single highest-value Researcher query pattern for advisory work. For roughly 45 minutes of consultant time (brief + review + convert) you get the equivalent of 4-6 hours of manual desk research — with citations, structure, and Arup-relevant framing. Across a year of engagements, spending 4-6 of your 25 monthly queries on engagement kickoffs is probably the single most defensible use of the quota. Make this the first habit to build; the rest of Researcher's use cases follow.
06 / USE CASE 2

Competitive and market analysis.

The second natural Researcher use case. Who are the other firms in this space? What are they offering? Where are Arup's distinctive angles? What's moving in the market? Historically a 1-2 day desk research effort; now a 20-minute commission.

~25 min totalLow effort1 query · high valueMarket intelligence
Map the market, find the whitespace
Use this when
You need structured market intelligence. Typical scenarios: proposal response needing competitive field understanding; new practice-area entry where Arup needs positioning; sector-specific thought leadership; internal strategy work identifying whitespace opportunities; understanding a client's competitive context before a meeting.
Produce a competitive landscape analysis. ## Market scope [area — e.g. "digital advisory services for UK transport authorities, specifically BIM adoption and lifecycle asset management"] ## Competitors to cover 1. [firm 1 — e.g. "Mott MacDonald"] 2. [firm 2 — e.g. "Atkins"] 3. [firm 3 — e.g. "Jacobs"] 4. [firm 4 — e.g. "WSP"] 5. [firm 5 — e.g. "AECOM"] If there are notable smaller specialists worth including, suggest them. ## Dimensions for comparison For each firm: - Core services in this space - Target client profile - Notable recent engagements (last 18 months) - Digital / BIM capability positioning - Sustainability positioning - Pricing model if publicly knowable - Key people in the practice if publicly identifiable ## Analysis Beyond the comparison: - **Where the market is moving** — emerging themes, increasingly-demanded capabilities - **Areas of convergence** — where all firms are investing - **Areas of differentiation** — where specific firms are uniquely strong - **Whitespace for Arup** — where none are visibly strong but demand exists ## Sources Public information only — firm websites, press releases, industry reports, tender announcements. Cite specifically. Flag speculation. ## Output Structured report: executive summary (half page), comparison table, market synthesis (1-2 pages), whitespace analysis with specific recommendations. 6-8 pages total.

The specific failure mode to watch for

Competitive analysis has a particular AI failure mode: sometimes conflating capabilities between adjacent firms. "Mott MacDonald has a strong X practice" when it's actually Jacobs — because the two firms have similar phrasing on their websites. Always verify any competitor-specific claim before using it in a pitch or proposal. Cost of miss-attribution in a proposal is real.

The Computer Use escalation

For substantial competitive research needing gated industry databases (Gartner, Forrester, industry analyst reports), Researcher with Computer Use (Frontier only) extends reach. The agent can log into subscription databases with your approval, pull the research, integrate with public-web analysis. Same one query, richer output.

The advisory habit worth building
Consider a monthly competitive-intelligence Researcher query for the practice area you work in. One query per month. Scope it: "what's changed in [my sector] in the last 30 days — new entrants, announced engagements, regulatory changes, notable thought leadership from competitors?" Review with the practice team. Over a year, you have a continuous market-intelligence stream that used to require commissioned research or industry analyst subscriptions. Compounding benefit: team-level awareness of the market stays current without anyone's calendar dedicated to maintaining it.
07 / USE CASE 3

Researcher report → deliverable.

The March 2026 capability that closes the loop. A Researcher report becomes a slide deck, PDF, one-page infographic, or audio briefing — without re-writing. The workflow that makes Researcher's output immediately useful rather than "a source document for my actual deliverable."

~5-15 min per formatVery low effort0 additional queriesFormat conversion
One report, four formats
Use this when
A Researcher report needs to travel to different audiences in different forms. Typical scenarios: research → internal briefing (PowerPoint) + client one-pager (infographic); engagement research → team deck + formal client PDF brief; thought leadership → LinkedIn infographic + blog post draft; sector research → commute audio + slides for the partners' meeting.

The format-to-audience matrix

  • PowerPoint — meetings (client, internal, partner, team briefings). Slides give you structure for verbal presentation. Edit afterwards to apply Arup template and refine.
  • PDF — formal sharing (email attachments, Loop workspace archive, SharePoint). Travels well, preserves formatting, read-only by design.
  • Infographic — single-page summaries (LinkedIn, client one-pager, team briefing hand-out). Visual, scannable, social-shareable. See Part XIV.
  • Audio overview — on-the-go consumption (commute, gym, walking review). Most useful for self-consumption — team members prefer to read but listen during dead time. See Part X for equivalent pattern in Copilot Notebooks.

The multi-format strategy

One query, three deliverables. Run Researcher query. Once report is ready, convert to PowerPoint for Tuesday's internal briefing. Convert to PDF for the engagement's Loop workspace archive. Convert to infographic for the one-page summary to share with the client partner. Three deliverables, one query, maybe 10-30 minutes of conversion + editing time across all three. This is the workflow pattern that turns Researcher from "a research tool" into "an output-production engine."

The edits still needed

Converted outputs are good starting points, not final deliverables:

  • PowerPoint — apply Arup template styles; adjust slide order for narrative flow; tighten dense slides
  • PDF — verify formatting transfer; add Arup cover if formal use; check pagination
  • Infographic — review visual elements for clarity; add Arup brand-kit styling; adjust hierarchy
  • Audio — preview for mispronunciations (technical terms, specific names); usually just for you so minimal editing
Why this quietly changes Researcher's ROI
Before the March 2026 format-conversion feature, Researcher reports had a last-mile problem. The report was useful but required manual re-formatting into whatever deliverable the consultant actually needed — typically 30-60 minutes per format. Now it takes 5-15 minutes including editing. For a consultant producing multiple deliverables from one research query, this multiplies the practical value of each of the 25 monthly queries by 2-3x. The feature doesn't feel like it changes much; in practice it changes the consumable output per query meaningfully. Build it into every Researcher workflow from day one.
08 / USE CASE 4

Analyst for client data deep-dive.

The primary Analyst use case for advisory work. Client sends spreadsheets. You need to understand what they say, what patterns are in them, what insights emerge. Historical pattern: export to Excel, build pivot tables, eyeball for patterns, hope you didn't miss anything. New pattern: attach the files, ask the question, read the analysis.

~10-15 minLow effort1 query · high valueClient data
Attach, ask, interpret
Use this when
A client has provided data that needs interrogation. Typical scenarios: client sends operational data (transactions, stakeholder register, project portfolio); financial spreadsheets for benchmarking; performance data across regions; survey responses; historical time-series for forecasting; multi-file datasets needing consolidation.
01
Copilot app → Agents → Analyst
02
Attach files (+ icon → Attach content)
03
Ask a specific question about the data
04
Wait; watch Python code for verification
05
Interpret output; follow up with refinements
Analyse these files and help me understand the data. ## Attached - [file 1 — e.g. "client_portfolio_2024.xlsx"] - [file 2 — e.g. "project_outcomes_2024.xlsx"] - [file 3 — e.g. "client_portfolio_2025.xlsx"] ## Context I'm an advisory consultant. The client is [description]. The data represents [what it is — e.g. "their active project portfolio over two calendar years with outcome metrics"]. I'm preparing for a portfolio review meeting. ## Questions 1. **Overall shape** — how many projects per year, total value, distribution across categories? 2. **Change year-over-year** — what's materially changed between 2024 and 2025? New categories, growing areas, shrinking areas? 3. **Outliers** — any projects that look unusual (very large, very long, very high variance on outcomes)? 4. **Outcome patterns** — do certain project types or sizes correlate with better or worse outcomes? 5. **Risk signals** — any patterns that might indicate portfolio-level risk (concentration, ageing projects, declining outcomes)? ## Output - Clear narrative interpreting the data - Charts / tables where they illuminate - Callouts for anything unexpected - Suggested follow-up questions to ask the client based on what the data shows Show your Python code so I can verify methodology. Cite specific rows / columns when making claims about the data.

What a good Analyst output looks like

Expect 2-4 pages of structured output:

  • Executive summary — 3-5 key findings in plain language
  • Each question addressed with the data-driven answer
  • Charts or tables — generated by the Python code
  • Python code blocks — visible, annotated, reviewable
  • Data caveats — limitations, missing data, ambiguous fields
  • Follow-up suggestions — what you'd want to analyse next

Verify the methodology

The Python code being visible is the feature. Read it:

  • Column name mapping — did Analyst correctly identify what each column represents?
  • Calculation logic — especially for financial data, is the maths right?
  • Data filtering — are any rows being excluded? Why?
  • Aggregation approach — sums, averages, medians chosen appropriately?

Don't trust headline numbers; trust the code that produced them.

The capability that changes client-data work
Before Analyst, a consultant receiving client data had three options: (a) accept the client's own interpretation at face value; (b) spend half a day in Excel doing analysis; (c) commission internal data analyst support. Option (a) is risky, (b) is slow, (c) is rare for small data requests. Analyst opens a fourth option: serious analysis in 10-15 minutes, with code visible for verification. For advisory consultants in roles that regularly receive client data — portfolio reviews, performance analysis, benchmarking work — this is the single highest-frequency Analyst use case and the most defensible way to spend monthly queries.
09 / USE CASE 5

Analyst for Arup portfolio insights.

The second Analyst use case. Not client data — Arup's own data. Team portfolio performance, engagement metrics, practice-area trends, internal benchmarks. The analyst-on-call you never had, available for the 15-minute "I'm curious about this" questions that normally go unanswered.

~10 minLow effort1 query · medium valueInternal insights
The "I'm curious about this" analyses
Use this when
A quick internal analysis question that would take effort to answer manually but has real decision value. Typical scenarios: team performance trends over quarters; engagement mix analysis; capacity planning from timesheet data; practice-area revenue patterns; client portfolio concentration; pipeline velocity analysis; workshop throughput metrics; internal benchmarking.

The "curiosity query" pattern

The single pattern worth establishing: when you notice yourself wondering about internal data, run an Analyst query rather than noting it down "for when I have time." The friction cost is 10 minutes; the decision value is often meaningful. Some specific Arup patterns:

1. Quarterly engagement mix review. Export the team's engagement register. Ask Analyst: "analyse this register by client sector, engagement size, and outcome — what's the overall mix, what's changed year-on-year, any concentration risks?"

2. Pipeline velocity analysis. Export pipeline data. Ask: "for engagements won vs lost over the last 12 months, what's the average time from first contact to contract? Are there patterns in what converts faster?"

3. Practice-area revenue patterns. Revenue data by practice area and quarter. Ask: "identify any seasonal patterns, growth or decline trends, and practice areas where revenue variability is unusually high."

4. Internal capacity analysis. Timesheet data across the team. Ask: "who's been most utilised over the last three months? Any signals of over-utilisation? How does utilisation map to seniority?"

5. Client portfolio concentration. Client data. Ask: "what proportion of our total revenue comes from our top 5 clients? Has concentration risk increased or decreased over 18 months?"

Governance discipline

Two rules for internal-data Analyst use:

  • Anonymise where appropriate. If analysing utilisation across named individuals, consider whether the analysis benefits from names or if role labels would do. Personal data in AI analysis follows the same principles as personal data anywhere — minimise.
  • Keep outputs appropriately scoped. A utilisation analysis stays within the partner group; a commercial-sensitivity analysis stays with engagement leadership. Don't share Analyst outputs more broadly than the source data would be.
The friction Analyst removes
Most internal-data analyses that would have been useful never get run — because the cost-benefit of commissioning them doesn't work for casual curiosity questions. Nobody's going to wait for a data analyst to turn around a 4-hour piece of work just because someone wondered. Analyst changes the cost to 10 minutes, which changes the threshold: curiosity becomes analysable. Over months, this changes how the team relates to its own data — more questions get asked because asking is cheap. That's a genuine capability shift for a knowledge-worker team, not just a productivity hack.
10 / USE CASE 6

Budgeting your monthly query allowance.

The discipline most consultants struggle with. A monthly allowance feels like a lot at first. By day 20, most consultants find they've used most of it on marginal tasks and don't have capacity for the ones that actually mattered. The skill is knowing which queries deserve the slot.

Ongoing disciplineMental overheadMonthly capThe constraint
Spend queries deliberately
Use this when
Always. The budgeting mindset needs to be on by default, not activated when the quota runs low.

A sensible monthly allocation

For a typical advisory consultant, a defensible allocation across the monthly allowance might look like:

  • 4-6 engagement-kickoff Researcher queries — highest-value pattern, covered in UC1
  • 2-3 competitive / market Researcher queries — monthly intelligence, UC2 pattern
  • 4-6 client-data Analyst queries — responding to actual client data when it lands, UC4
  • 2-3 internal Analyst queries — practice-area insights, curiosity questions, UC5
  • 3-5 opportunistic Researcher or Analyst — ad-hoc substantive research or analysis as tasks arise
  • 4-6 reserve — kept for end-of-month urgency, unexpected needs, partner-requested work

Not a prescription; a starting template. Adjust based on actual work. Senior consultants in research-heavy roles may lean more toward Researcher; consultants in data-heavy workstreams may lean toward Analyst.

The weekly rhythm

The discipline works better with a weekly cadence:

  • Monday — look at the week ahead; identify 2-3 specific queries you'll likely run
  • Wednesday — check how many queries you've used; recalibrate if needed
  • Friday — reserve the remainder for late-month urgency; don't splurge

Rules of thumb for "is this worth a query?"

  • Would I assign this to a junior analyst or data scientist? → Yes, good query. → No, probably standard Copilot.
  • Will the output shape a decision or a deliverable? → Yes, good query. → No, maybe skip.
  • Would I be disappointed if I ran out of queries and missed this? → Yes, save it. → No, use standard Copilot.
  • Is this a recurring pattern I can template? → Yes, use it to test the template, then batch runs.
  • Am I running this because I'm curious or because I need the answer? → Need, run it. → Curious only, save the slot.

What doesn't deserve a query

Things consultants sometimes run named-agent queries for that don't deserve the slot:

  • "What's the current ORR guidance on X?" — standard Copilot in Web mode handles this in 10 seconds
  • "Draft me a briefing note on Y" — standard Copilot with decent prompt
  • Simple data summaries that a pivot table answers in 2 minutes
  • Web research that's genuinely surface-level
  • "Interesting" but not actionable explorations
The end-of-month panic pattern
Most advisory consultants adopting Researcher and Analyst hit the same failure mode in month 1: they use the quota enthusiastically in the first two weeks, then find themselves rationing in the last week when a meaningful task comes up and they're at 24/25 queries. The fix is counter-intuitive — spend conservatively early, not late. By Day 15 you should ideally still have 15+ queries remaining. Reserve capacity for end-of-month needs because urgent high-value work always turns up in the last week. The consultants who get the most value from these agents aren't the ones who use queries most frequently — they're the ones who save the best queries for the tasks that actually mattered.
11 / DECISION MATRIX

Which agent, when?

Six use cases, two agents, one table.

Your scenarioUse thisAgentTime
New engagement kickoff researchEngagement brief templateResearcher~30 min
Competitive landscape for a proposalCompetitive landscape promptResearcher~25 min
Monthly market intelligence checkSector change promptResearcher~20 min
Client sends multi-file data for analysisMulti-file Analyst promptAnalyst~15 min
Quick internal data curiosity questionAttach + ask patternAnalyst~10 min
Research report needs client-ready slide deckConvert to PowerPointResearcher~15 min
Listen to brief during commuteConvert to audio overviewResearcher~3 min
Gated database research (Gartner, etc)Researcher with Computer UseResearcher (Frontier)~20 min
Trend detection across time-series dataTime-series Analyst promptAnalyst~15 min
Simple factual question with current web infoStandard Copilot, NOT ResearcherCopilot Chat~30 sec
Quick email draft or summaryStandard Copilot, NOT named agentsCopilot Chat~30 sec
Small data query (single spreadsheet, basic aggregation)Excel Copilot, NOT AnalystExcel~2 min
12 / DO'S & DON'TS

Common failures — and their fixes.

Patterns from training advisory teams on the named agents. Almost every bad outcome maps to one of these.

Do
  • Treat each query as a commissioning decision, not a casual prompt. The monthly cap reinforces the discipline.
  • Write structured briefs for Researcher — context, what's needed, sources, output format. Vague input produces vague output.
  • Check Python code Analyst produces — it's visible for a reason.
  • Use format conversion on every substantive Researcher report — PowerPoint for meetings, PDF for archive, infographic for summaries.
  • Budget the monthly allowance deliberately across engagement kickoffs, market intel, client data and internal work — reserve some for end-of-month surprises.
  • Use Researcher for "tell me about X" and Analyst for "what does this data say" — the clean decision rule.
  • Verify critical claims before commercial use — Critique helps but doesn't absolve.
  • Engage clarifying questions when the agent asks — it's asking because specificity matters.
Don't
  • Run named-agent queries for things standard Copilot handles — you're wasting budget on tasks that take 10 seconds in Chat.
  • Abort queries mid-run — the 5-15 minute wait is the mechanism, not a bug.
  • Re-submit the same query repeatedly — you'll burn quota and the second run isn't meaningfully different.
  • Use Analyst for simple spreadsheet work Excel Copilot does in 30 seconds.
  • Use Researcher for web searches you could do with standard Copilot Chat in Web mode.
  • Splurge queries in the first two weeks — end-of-month panic pattern is real.
  • Trust outputs without review — even with Critique, verify anything you'll rely on commercially.
  • Confuse these with Agent Builder / Copilot Studio — those are for building custom agents; these are pre-built.
The unifying mental model
Think of Researcher and Analyst as two named colleagues, one shared budget. Researcher = sharp junior analyst who returns structured reports. Analyst = data scientist who shows their Python code. 25/month = the total hours you can commission from them, across both. The decisions each week are: which tasks need one of these colleagues vs standard Copilot, which tasks need the research colleague vs the data colleague, and whether this week's query budget supports what's coming. The discipline of treating them as commissionable agents rather than unlimited tools is the entire skill. Consultants who get it: spend fewer queries, get more value. Consultants who don't: use the budget quickly, get less value, hit end-of-month empty and miss the tasks that mattered.
Playbook implication
Researcher and Analyst are the two features that most directly justify the M365 Copilot (Premium) licence for advisory knowledge workers. Teach them last in the sequence — because the commissioning discipline only makes sense once consultants have internalised standard Copilot patterns and know what these agents are for. Without that foundation, the 25/month feels arbitrary rather than structural. With it, the quota becomes a forcing function for better use.
Series Complete · Fifteen Reports
The research corpus, closed.
Part XV closes the Arup Advisory Playbook research series. Every major Microsoft 365 Copilot surface the advisory team uses — from the Copilot app through every workspace, creative tool, and named agent — now has dedicated full-depth workflow research. The 20-item inventory is covered in its entirety. Ready for playbook integration.
PART I
M365 Copilot Reference
PART II
Competitive Landscape
PART III
Tool Arsenal
PART IV
Advanced Agent Stack
PART V
PowerPoint
PART VI
Excel
PART VII
Word
PART VIII
Outlook
PART IX
Teams
PART X
OneNote + Notebooks
PART XI
Edge
PART XII
The Copilot App
PART XIII
Loop + Pages
PART XIV
The Create Module
PART XV
Named Agents