Agent Foskett Academy • Microsoft Security Copilot • Module 1 • Lesson 1

Lesson 1 — What is Microsoft Security Copilot?

Microsoft Security Copilot is an AI-powered security solution designed to help security and IT professionals investigate threats, understand security signals, assess risk and work through complex security information more quickly.

It can summarise incidents, explain alerts, support threat intelligence, help develop KQL queries and suggest investigation pivots across Microsoft security products.

But Security Copilot does not replace the analyst. The analyst still validates the evidence, understands the business context and decides what action should be taken.

AI accelerates the investigation. Evidence closes the case.
Agent Foskett What is Microsoft Security Copilot lesson
What you will learn

This lesson introduces Security Copilot and the evidence-first mindset used throughout the Academy.

What Security Copilot is
Where analysts use it
What it can help investigate
Why human validation remains essential

Microsoft Security Copilot investigation model

Security data, alerts and threat intelligence

Microsoft Defender XDR, Sentinel, Entra, Defender for Cloud and other connected products

Analyst asks a natural-language security question

Security Copilot processes available context through connected capabilities and plugins

Copilot returns a summary, explanation, suggested query or investigation pivot

Analyst reviews the cited sources and underlying evidence

Facts are separated from assumptions and uncertainty

The analyst decides the next investigation or response action

Security Copilot at a glance

AreaWhat Security Copilot providesWhat the analyst still provides
InteractionNatural-language prompts and follow-up questions.Clear objectives, scope and investigation context.
Incident reviewSummaries of alerts, entities and possible attack activity.Validation against original alerts, logs and timelines.
Threat intelligenceExplanations of threat actors, indicators and techniques.Relevance to the organisation and current incident.
KQL supportQuery generation, explanation and refinement assistance.Schema knowledge, testing and result validation.
ReportingDraft technical and executive summaries.Accuracy, confidence, impact and final approval.
ResponseSuggested next steps and investigation pivots.Risk decisions, containment approval and accountability.

Learning objectives

  • Define Microsoft Security Copilot.
  • Understand why it was created.
  • Recognise standalone and embedded experiences.
  • Identify common security use cases.
  • Understand its relationship with Microsoft security products.
  • Recognise the limits of AI-assisted output.
  • Apply evidence-first analyst validation.

What is Microsoft Security Copilot?

Microsoft Security Copilot is a generative AI security solution built for security and IT professionals.

Natural-language security analysis

Analysts can ask questions using everyday language instead of beginning every investigation with a portal search or manually written query.

Built for security work

Security Copilot is designed for security operations, incident response, threat intelligence, posture, identity and IT administration scenarios.

Why Security Copilot was created

Security teams face high alert volumes, fragmented evidence, complex attacks and pressure to investigate quickly.

The problem is not always missing data

Many investigations already contain enough evidence. The challenge is connecting it across identities, devices, email, cloud resources and threat intelligence.

Faster initial understanding

Security Copilot can help turn a collection of alerts and entities into an initial explanation of what may have happened.

Standalone experience

The standalone Security Copilot experience provides a dedicated workspace for prompts, sessions, promptbooks and connected security capabilities.

Embedded experiences

Security Copilot capabilities can also appear inside supported Microsoft security products, allowing analysts to use AI in the workflow they are already investigating.

Microsoft Defender XDR

Security Copilot can help summarise incidents, explain alerts, review entities and support cross-domain investigations.

Microsoft Sentinel

Security Copilot can assist with incident investigation, threat hunting, KQL development and security operations workflows.

Microsoft Entra

Identity investigations may include risky sign-ins, suspicious authentication, privilege changes and compromised-account activity.

Microsoft Defender for Cloud

Cloud investigations may include resources, recommendations, attack paths, workload alerts and exposure context.

Microsoft Intune

Security Copilot can support device-management and endpoint-administration scenarios where supported capabilities are available.

Microsoft Purview

Security Copilot can support data-security, compliance and investigation scenarios through connected Microsoft capabilities.

Threat intelligence

Copilot can summarise threat actors, campaigns, indicators, tactics and techniques to provide external context.

Incident summaries

Copilot can organise alerts, entities and important events into a concise starting point for an investigation.

Alert explanations

Security Copilot can explain unfamiliar alert names, behaviours, terminology and relationships.

Investigation pivots

Copilot can suggest useful entities, questions and evidence sources for the analyst to investigate next.

Timeline assistance

AI can help arrange events chronologically, but the analyst must confirm timestamps and relationships.

KQL generation support

Security Copilot can help draft queries, but every table, field, filter and returned result must be checked.

KQL explanation

Copilot can explain what an unfamiliar query is attempting to do and identify possible investigation logic.

Promptbooks

Promptbooks organise repeatable sequences of prompts for common security workflows and investigations.

Plugins

Plugins connect Security Copilot to supported Microsoft and third-party capabilities and provide data or actions for prompts.

Grounded responses

Useful responses depend on the context, permissions, plugins and data available to the current user and session.

Permissions still matter

Security Copilot does not automatically give an analyst access to security data they are not authorised to view.

Business context matters

An AI response may identify suspicious behaviour without understanding whether that behaviour is expected in the organisation.

AI output is not evidence

A fluent explanation can still be incomplete, incorrect or based on a weak assumption.

Review cited sources

Analysts should inspect the alerts, entities, logs, threat intelligence and product sources used to produce a response.

Separate fact from inference

Record which conclusions are confirmed, likely, possible or unsupported.

Check the timeline

Confirm that timestamps create a credible sequence and that events are not being connected only because they look similar.

Check the entities

Verify that users, devices, IP addresses, mailboxes, applications and cloud resources belong to the same investigation.

Check the scope

Determine whether the activity affects one account or a broader group of users, devices and resources.

Check uncertainty

Ask Copilot to identify missing data, conflicting evidence and conclusions that cannot be verified.

Human judgement remains essential

The analyst decides whether an incident is genuine, how serious it is and what response is appropriate.

Security Copilot is a force multiplier

It helps capable analysts move faster, but it does not replace security knowledge or investigation discipline.

Example introductory prompt

Summarise this incident for a security analyst.

Identify:
1. The likely initial access method
2. The affected users and devices
3. The most important evidence
4. Any gaps or uncertainty
5. The next three investigation steps

Separate confirmed facts from assumptions.

Agent Foskett investigation: “The incident contained fourteen alerts…”

The incident contained fourteen alerts

Three endpoints

Two identities

One suspicious email

A malicious URL click

A PowerShell download

An impossible-travel alert

An OAuth consent request

Every alert made sense by itself

Together they appeared to describe a larger attack

Agent Foskett asked Security Copilot for an incident summary

Copilot proposed a likely phishing-led attack path

Agent Foskett did not accept the answer as evidence

He reviewed the email, sign-ins, endpoint timeline and OAuth activity

Some events were confirmed

One alert belonged to unrelated administration activity

The final attack path was narrower than the AI summary

Security Copilot accelerated the investigation

Evidence closed the case
The answer sounded convincing. The evidence decided whether it was true.

Security Copilot validation checklist

Validation areaQuestion to askEvidence to review
SourceWhat data supports this claim?Alerts, logs, entities, plugins and cited sources.
TimelineDo the events occur in a credible order?Event timestamps and investigation timeline.
IdentityIs the correct user or service identity involved?Sign-in, audit, role and risk activity.
DeviceIs the activity linked to the correct endpoint?Device timeline, processes, files and network connections.
EmailWas the message delivered and interacted with?EmailEvents, UrlClickEvents and mailbox evidence.
CloudAre the affected resources and permissions connected?Resource, identity, alert and attack-path evidence.
ScopeIs this isolated or widespread?Related users, devices, resources and indicators.
ContextCould legitimate activity explain the behaviour?Change records, owners and business operations.
ConfidenceIs the conclusion confirmed or inferred?Supporting and conflicting evidence.
ActionIs there enough verified evidence to respond?Risk, impact, containment options and approvals.

Key takeaways

  • Microsoft Security Copilot is a generative AI security solution for security and IT professionals.
  • Analysts interact with it using natural-language prompts and follow-up questions.
  • Security Copilot can help summarise incidents, explain alerts, support threat intelligence and assist with KQL.
  • It is available through standalone and embedded experiences.
  • It works alongside Microsoft security products rather than replacing them.
  • Permissions, connected capabilities and available context influence the response.
  • AI output may be incomplete or incorrect even when it sounds convincing.
  • Analysts must review sources, timelines, entities, scope and business context.
  • Human judgement remains responsible for conclusions and response actions.
  • AI accelerates the investigation. Evidence closes the case.

What Agent Foskett checked

  • Incident summary
  • Source evidence
  • Alert relationships
  • User identities
  • Endpoint activity
  • Email evidence
  • Cloud resources
  • Timeline accuracy
  • Unsupported assumptions
  • Remaining uncertainty

Best practices

  • Ask focused questions.
  • Provide useful context.
  • Request supporting evidence.
  • Separate fact from inference.
  • Review cited sources.
  • Confirm timelines and entities.
  • Protect sensitive information.
  • Use least privilege.
  • Document confidence and uncertainty.
  • Keep the analyst accountable.

Related Agent Foskett resources

Continue through the Security Copilot Academy and the Microsoft security learning paths that provide the evidence Copilot helps analysts investigate.

Continue the Microsoft Security Copilot Academy

Lesson 1 introduces Security Copilot and the evidence-first analyst mindset. The next lesson explains how Security Copilot processes prompts, context, plugins and connected security capabilities.

What is Microsoft Security Copilot?

Microsoft Security Copilot is a generative AI security solution designed to help security and IT professionals investigate threats, process security signals, assess risk and work across Microsoft security products using natural-language prompts.

Microsoft Security Copilot Lesson

This Agent Foskett lesson introduces Security Copilot, standalone and embedded experiences, incident summaries, alert explanations, threat intelligence, KQL support, plugins and evidence-first analyst validation.