| Format | Live webinar — instructor-led with live demos |
| Duration | 90 minutes |
| Audience | DOJ professionals across all components and divisions |
| Environment | Microsoft 365 Government Community Cloud (GCC) |
| Skill Level | No prior AI or Copilot experience required |
This 60-90 minute session gives DOJ professionals a fast, practical introduction to Microsoft 365 Copilot Chat. Rather than working through exercises independently, you will follow along with the instructor as they demonstrate Copilot in a live GCC environment using realistic DOJ scenarios — from drafting a Deputy AG briefing to analyzing a program budget in Excel.
By the end of the session you will understand what Copilot Chat can do for your day-to-day work, how to write prompts that get results on the first try, and how to verify AI output before it goes into any work product.
To follow along during the demos, have the following ready before the webinar begins:
- Your DOJ M365 work account signed in at office.com
- Microsoft Edge or Chrome open and ready
- Copilot Chat accessible — navigate to the app launcher and confirm you see Copilot
- Microsoft Word and Excel available (desktop app or web)
- The DOJ-After-Action-Report.md file saved to your Desktop (provided by your facilitator)
Reminder: Do not enter classified, Privacy Act-protected, grand jury, attorney-client privileged, or other legally sensitive information into Copilot Chat. All demos use fictional DOJ scenarios.
- Instructor introductions and session overview
- Confirm participants can access Copilot Chat
- Quick poll: Who has used Copilot before?
What the instructor will cover:
- What Microsoft 365 Copilot Chat is — and what it is not
- How to access it in GCC: the M365 portal, Teams, and within Office apps
- A guided tour of the Copilot Chat interface
- Enterprise Data Protection: what it protects and what it does not
Live demo — follow along:
The instructor will open Copilot Chat and enter this prompt. Try it yourself at the same time:
"What are three things I should consider when drafting talking points for the Deputy Attorney General about DOJ's current priorities in combating violent crime and drug trafficking? Keep the response under 200 words."
The instructor will then:
- Check and click a source citation
- Ask a follow-up to reformat the response as a two-column table
- Show where to verify Enterprise Data Protection is active
Key takeaway: Copilot Chat responds to plain-language requests and improves with follow-up — just like working with a colleague.
What the instructor will cover:
- The five elements of a strong prompt: Role · Goal · Context · Source · Expectations
- Side-by-side: a weak prompt vs. a strong prompt for the same DOJ task
- AI hallucinations — what they are, why they happen, and why they matter especially in legal and law enforcement work
- Five techniques to prevent hallucinations in your prompts
Live demo — follow along:
Part A — Strong vs. Weak
First, try the weak version: "Tell me about federal drug enforcement."
Then try the strong version: "I am a senior policy advisor to the Deputy Attorney General drafting a briefing paper. Summarize the key pillars of the federal approach to combating drug trafficking, including the roles of DEA, FBI, and U.S. Attorneys' Offices. Focus on interagency coordination. Present as 4–5 bullet points, one sentence each, in formal policy-brief tone."
Part B — Hallucination Prevention
The instructor will demonstrate three techniques back-to-back:
Constrain the source: "Summarize the current status of DOJ's Project Safe Neighborhoods program. Use only information from official DOJ press releases and justice.gov. If specific details cannot be confirmed, say so rather than estimating."
Require attribution: "List the major DOJ grant programs that fund state and local law enforcement. For each, include the administering office and purpose. After each entry, cite your source. Flag anything unverified."
Ground in an uploaded document: The instructor will upload DOJ-After-Action-Report.md and prompt: "Using only the attached document, summarize the three most important operational gaps identified. Do not add information from outside this document."
Part C — Fixing a Flawed Prompt
The instructor will enter this broken prompt and ask participants to spot the problems before fixing it live: "Give me a detailed one-sentence summary of all DOJ civil rights cases since 2020 in bullet format as a single paragraph."
Key takeaway: The quality of Copilot's output is a direct reflection of the quality of your prompt. Hallucination prevention is not optional in DOJ work — fabricated legal citations are a real and documented AI risk.
Exercise: Hands-On Prompt Playground
Participants practice writing and refining their own prompts across three categories of DOJ work: legal tasks, investigative tasks, and administrative tasks. Each section provides a scenario and a starter prompt with guided iteration — no single right answer, just hands-on repetition.
What participants will do:
- Compare weak and strong prompts side by side for legal research, investigative summaries, and executive correspondence
- Apply hallucination-prevention techniques (source constraining, attribution, document grounding) in realistic DOJ contexts
- Iteratively refine Copilot responses using targeted follow-up prompts
- Build a prompt from scratch for their own work context using the five-element framework
This exercise can be completed independently after the webinar or as a guided lab in the full 4-hour workshop.
What the instructor will cover:
- How Copilot Chat works inside Word, Excel, Outlook, and PowerPoint
- GCC limitations and the workarounds used in this module
- Introduction to Copilot Pages for team collaboration
Live demo — DOJ Priorities Review scenario:
The instructor will walk through three rapid demos using the fictional DOJ Priorities Review scenario (September 2026), showing how a senior DAG advisor would use Copilot to build real work products.
Demo 1 — Word
In Word, the instructor will upload DOJ-After-Action-Report.md and draft a read-ahead document opening section: "Draft the Purpose and Background sections for a DOJ Priorities Review read-ahead. The review is on 14–15 September 2026 at the Robert F. Kennedy Building. Draw on the attached document for context. Use formal government writing style."
Demo 2 — Excel
The instructor will attach DOJ-Budget-Data.csv in Copilot Chat and prompt: "Summarize the total budget allocation and spend by program office. Which offices are at risk of exceeding their FY2026 budget before year-end?"
Then show how to take that analysis and manually build a column chart in Excel — the GCC workaround for chart generation.
Demo 3 — Outlook
In Outlook, the instructor will draft an invitation from the DAG's office: "Draft a formal email from the Office of the Deputy Attorney General to all Component Directors inviting them to the DOJ Priorities Review on 14–15 September 2026. Include an RSVP deadline of 5 September. Formal government style."
Then refine with a follow-up: "Make the tone slightly more collegial — these are senior leaders the DAG works with regularly. Add a line about a hosted working dinner on the evening of September 14."
Key takeaway: Copilot does not replace your expertise — it removes the blank-page problem and compresses drafting time so you can focus on judgment, review, and refinement.
- Open Q&A
- How to access the full hands-on workshop (4-hour instructor-led lab)
- Where to find your Component's AI Acceptable Use Policy
- Recommended first prompt to try on your own today
Start with these three prompts on your own in Copilot Chat:
1. Summarize something you're already working on
"I am a [your role] at [your component]. Summarize the key points of the attached document in 5 bullet points for a [senior official/committee/stakeholder]. Use formal language."
2. Draft a response or memo
"Draft a [one-paragraph / one-page] response to [describe the request or topic]. My audience is [describe]. Keep the tone [formal / professional / direct]."
3. Test hallucination prevention
"[Ask about a policy or topic you know well]. Use only publicly available DOJ sources. Cite each claim. Flag anything you cannot verify."
Then check whether Copilot's answer matches what you already know.
This webinar is a condensed overview. The full 4-hour hands-on workshop covers the same material through guided lab exercises where participants build real work products in their own GCC environment:
| Webinar | Full Workshop | |
|---|---|---|
| Format | Instructor demo, follow-along | Hands-on lab exercises |
| Duration | 1 hour | 4 hours |
| Prompts | Instructor-led | Participant-written |
| Work products built | None (demo only) | Read-ahead document, budget analysis, Outlook invitation, briefing deck, Copilot Page |
| Hallucination training | Demonstrated | Practiced across 5 techniques |
| Prompt Gallery | Shown | Explored and saved |
Contact your facilitator or training coordinator to schedule the full workshop for your team.
Course: MS-4023 — Transform Ideas into Action with Copilot Chat (DOJ Edition)
Classification: UNCLASSIFIED // FOR OFFICIAL USE ONLY
All scenarios and data used in this session are fictional and for training purposes only.