Copilot Researcher Agent
Deep research that cites its sources and does the thinking for you.
8
Cards
20
Minutes
30
Prompts
Prerequisite: Basic Copilot familiarity (Module 01 recommended). Requires a Microsoft 365 Copilot licence.
Chat vs. Researcher - Why This Is Different
Standard Copilot chat answers your question in one pass. The Researcher agent investigates it across multiple steps - reading, reasoning, cross-referencing, and synthesising.
Key differences:
Chat
Single-pass answer. Instant. No source citations.
Researcher
Chain-of-thought reasoning. 5-45 min. ~10 citations per report.
When to use chat: quick factual lookups, drafting, summarising a single document.
When to use Researcher: multi-source synthesis, competitive analysis, strategic briefings, policy research, anything requiring cited evidence.
How to access: Requires M365 Copilot licence. Find Researcher in the Agents pane at microsoft365.com/copilot. If it's greyed out, your admin needs to enable it in the Copilot Control System.
Try these prompts
What is the Researcher agent in Microsoft 365 Copilot and how does it differ from standard Copilot chat?
Show me a list of all agents available in my Microsoft 365 Copilot.
Which task is best suited for the Researcher agent?
Correct.
Multi-source competitive analysis is exactly what Researcher is built for - it searches multiple sources in parallel, reasons through the findings, and delivers a structured report with citations.
Your First Deep Research
The Researcher workflow has five phases. Understanding them lets you get better results from every prompt.
Submit prompt
You describe what you need researched.
Clarification questions
Researcher asks 1-2 questions to narrow scope. Don't skip this - it's the difference between generic and useful.
Parallel search
Searches web + your organisational data simultaneously. Multiple queries from multiple angles.
Chain-of-thought synthesis
You can watch it reason - which sources it's consulting, patterns it's finding, how its thinking evolves.
Structured report
Executive summary, detailed sections, charts, source citations. Ready to use.
Processing times: Simple research: under 5 min. Moderate synthesis: 10-20 min. Complex multi-source analysis: 30-45 min. Start a deep research before your meeting, not during it.
Try these prompts
Analyse the current state of [industry] in [region]. Include market size, key players, emerging trends from the last 12 months, and regulatory changes. Cite all sources.
Before you research, tell me what clarifying questions you'd ask to make this analysis most useful for my role as [role].
Organisational Data + Web
Researcher's superpower: it searches your Microsoft Graph data (emails, files, meetings, Teams, SharePoint) AND the web simultaneously. No other research tool does this.
The dual engine
Work data
Emails, files, meetings, Teams chats, SharePoint, OneDrive
Web data
Market trends, competitor info, regulatory changes, industry reports
Semantic understanding: Researcher understands concepts, not just keywords. Searching for "revenue targets" surfaces documents about "sales goals", "profit margins", and "financial objectives" - even when those exact terms aren't used.
Privacy boundaries: Researcher can only access data you already have permission to see. It doesn't bypass document permissions or access other people's mailboxes.
Try these prompts
Search my recent emails and Teams messages about [project name]. Summarise the current status, key decisions made, and any open questions - then research industry best practices for this type of project and compare.
Based on documents in our [SharePoint site/team name], identify the top 3 risks for [initiative]. Then research how other organisations in [industry] have mitigated similar risks. Cite internal and external sources separately.
When you ask Researcher to "compare our onboarding process to industry best practices", what data sources does it search?
Correct.
Researcher's dual engine searches your Microsoft Graph data (emails, files, SharePoint, Teams) and the web at the same time. That's what makes it uniquely powerful for contextualised research.
Structuring the Perfect Research Prompt
A high-quality research prompt has five components. Missing any produces a generic report.
Topic
"AI in healthcare" is too broad. "How Australian regional health services are using AI for clinical governance reporting" is researchable.
Scope
Time period, geography, industry segment, number of competitors, depth (overview vs deep dive).
Audience
"Write for a clinical governance committee" produces different output than "write for a CEO board briefing."
Output format
"Include an executive summary, findings by theme, a comparison table, and 3 recommendations."
Constraints
"Under 2,000 words", "Only cite sources from the last 6 months", "Exclude [source]."
Complete research prompt template
Acting as a research analyst briefing [audience], investigate [topic]. Scope: [geography], [time period], [industry]. Include: executive summary, key findings organised by theme, comparison table of [items], and 3 evidence-based recommendations. Under [word count] words. Cite all sources.
I need a competitive landscape analysis of [3-5 specific competitors] in the [market] space. Compare: pricing models, target customer segments, key differentiators, and recent product announcements (last 6 months). Format as a comparison matrix with a narrative summary. Written for [role].
Which prompt will produce the best Researcher output?
Correct.
It includes all five components: specific topic, scope (geography + time), audience, output format, and constraints. Researcher will produce a focused, actionable report.
Export and Share
Researcher produces a structured report - but the real skill is turning it into the right deliverable for your audience.
Paste into any app. Loses some formatting.
Creates a collaborative Loop page. Team can contribute, comment, refine. Researcher updates in real-time alongside.
Fully formatted document. Best for board papers, compliance docs, formal reports.
Hover over charts/graphs, copy, paste into PowerPoint or Word.
Convert to conversational audio via Notebooks. 8 languages, adjustable length/speed.
Pro tip: For presentations, don't export the whole report. Ask Researcher: "Convert the key findings into 6 PowerPoint slide outlines with speaker notes for a 15-minute executive briefing."
Try these prompts
Reformat this research as a 1-page executive brief: 3 key findings, 2 risks, 1 recommendation. Under 400 words. Suitable for email distribution to senior leadership.
Create a summary version of this research for 3 different audiences: (a) executive team - strategic implications, (b) project team - actionable next steps, (c) frontline staff - what changes for them. Under 300 words each.
You need your team to collaboratively refine research findings before presenting to the board. Which export option?
Correct.
Edit in Pages creates a collaborative Loop page where your team can view, comment, and co-edit the research in real-time - and you can keep asking Researcher follow-up questions alongside.
Researcher + Notebooks
One-shot research is useful. But real research is iterative - new information arrives, priorities shift, follow-up questions emerge. Copilot Notebooks solve this.
What Notebooks add
A persistent workspace where you curate reference documents (Word, PowerPoint, Excel, PDFs, web links) and Researcher grounds all responses in those materials. It's the most powerful research configuration: curated context + deep reasoning + web access.
Setting up a research notebook:
- Navigate to microsoft365.com -> Notebooks -> Create new
- Give it a descriptive name (e.g., "Q4 Market Expansion Research")
- Add references: drag in Word docs, PowerPoints, Excel files, PDFs, web links
- The Overview page auto-generates a synthesis of everything
- Select Researcher from the chat interface inside the notebook
- Now Researcher grounds its research in your curated references AND broader sources
Use case: Ongoing competitive intelligence. Create a notebook, add competitor reports and internal strategy docs. Each week, ask Researcher to update your competitive landscape. The notebook accumulates institutional knowledge.
Quick create options from notebook content:
Try these prompts (inside a notebook)
Based on all references in this notebook, identify the top 5 themes, any contradictions between sources, and 3 areas where additional research is needed.
Compare what these internal documents say about [topic] with the latest published research. Where do our internal assumptions align with or diverge from current evidence?
Advanced Techniques
Three advanced capabilities expand what Researcher can do - if your admin has enabled them.
Researcher can navigate websites in a secure virtual environment - click through pages, fill forms, sign into sites with your permission. Useful for gated content, paywalled research, interactive data sources.
Security: Runs in a temporary sandboxed environment. Can't access your device or network. Asks permission before sensitive actions. You can watch it work in a virtual desktop preview.
Switch between OpenAI GPT-5 and Anthropic Claude via a dropdown. GPT-5 excels at structured, multi-step reasoning and quantitative analysis. Claude excels at nuanced text interpretation and policy analysis.
Thinking levels (GPT-5): Minimal (fast, less accurate) through High (slow, most accurate). Use Minimal for bulk tasks, High for complex strategic analysis.
Researcher can call the Analyst agent for data visualisation, or domain-specific agents your org has built. Example: Researcher gathers competitive intelligence, then hands findings to Analyst for statistical comparison charts.
Try these prompts
Using high reasoning depth, analyse [complex topic]. Consider at least 4 different perspectives, identify where experts disagree, and present a balanced synthesis with evidence for each position.
Research [topic] and then create a visual comparison matrix of [items] with scoring across [criteria]. Use the Analyst agent to generate charts.
Your admin has enabled Computer Use. When would you use it?
Correct.
Computer Use is specifically designed for navigating web interfaces that require interaction - paywalled content, login-protected data sources, or interactive tools. It can't bypass permissions and won't send communications without explicit approval.
5 Research Workflows
Five ready-to-use workflows. Pick one, customise the placeholders, and submit to Researcher. This is the "take it to work Monday" card.
Prepare a strategic briefing on [topic] for [audience]. Include market overview, competitive landscape, regulatory environment, and 3 strategic options with pros/cons. Ground in our internal strategy documents and latest web sources. Under 2,000 words.
Conduct a competitive analysis of [3-5 competitors]. Compare: market positioning, pricing, product features, recent announcements (last 6 months), strengths, and vulnerabilities. Format as comparison matrix + narrative. Include sources.
Research the current requirements of [regulation/standard]. Compare to our current practices documented in [internal source]. Identify compliance gaps. For each gap, recommend a remediation action with evidence. Written for [governance committee].
Conduct a literature review on [topic]. Include: key themes across published sources, areas of consensus, areas of debate, methodological approaches used, and gaps in current knowledge. Minimum 15 source citations. Academic register.
We're considering [Option A] vs [Option B] for [decision]. Research the evidence for and against each option. Include case studies of organisations that chose each path. Present as a decision matrix with recommendation.
Bonus: Prompt Library - Your enrolment includes 30 copy-paste research prompts. Here's a preview by category:
Executive Briefing / SWOT Deep Dive / Trend Forecast / Market Entry Analysis / Stakeholder Landscape / Risk Assessment
Competitor Deep Dive / Feature Comparison Matrix / Pricing Intelligence / Win-Loss Analysis / Market Share Landscape / Content Audit
Regulation Summary / Compliance Gap Analysis / Regulatory Change Monitor / Cross-Jurisdiction Comparison / Precedent Research / Policy Impact Assessment
Systematic Overview / Best Practice Synthesis / Case Study Collection / Evidence Brief / Innovation Scan / Historical Analysis
Options Analysis / Business Case Research / Vendor Evaluation / Change Impact Research / ROI Evidence Gathering / Meeting Preparation Brief
Course Complete
You now know how to use the Researcher agent for deep, cited, multi-source research.
What you learned:
- Chat vs Researcher - when to use each
- The 5-phase deep research workflow
- Combining organisational data with web research
- The 5-component research prompt structure
- 5 export paths for different stakeholders
- Notebooks for persistent, evolving research
- Computer Use, model selection, multi-agent
- 5 workflows ready to use Monday
Bonus included: Your Researcher Agent Prompt Library (30 prompts) is available for download. Copy-paste ready, organised by workflow.