AI Tools Presentation — Using Claude & AI Agents as a Team
Audience: Mixed team — developers, PMs, designers, and other roles
Goal: Show practical, real-world use cases that make AI tools feel approachable
1. Let’s Address the Elephant in the Room
A lot of people feel uncertain — or even anxious — about AI tools. That’s completely normal, and it’s the right instinct.
The concern usually isn’t “will this replace me?” The real concern is: “If I use it wrong, will I break something?”
The honest answer: AI tools are designed to assist you, not act autonomously. You are always in the driver’s seat. Think of Claude less like a self-driving car and more like a very capable co-pilot who never gets tired and never judges you for asking a “dumb” question.
2. The Tools We’re Talking About
Claude (claude.ai)
Anthropic’s AI assistant — available as a web app, desktop app, and mobile app. You interact with it in plain language. No special commands or syntax needed.
Best for: Writing, research, analysis, brainstorming, summarising long documents, drafting emails, explaining complex topics, and Q&A.
Cowork Mode (Claude Desktop)
A desktop version of Claude that can access files on your computer and automate multi-step tasks — creating documents, reorganising files, filling in reports, etc.
Best for: File-heavy workflows, document generation, repetitive formatting tasks, and anything where you’d normally copy-paste between apps for 20 minutes.
Other AI Tools in the Ecosystem
| Tool | What it does | Best for |
|---|---|---|
| GitHub Copilot | Autocompletes code as you type in your IDE | Developers writing code |
| Cursor | A full code editor with an embedded AI assistant | Developers who want deeper AI integration |
| Perplexity | AI-powered web search with cited sources | Research and fact-finding |
| Notion AI | AI built into Notion for summarising and writing | Teams already in Notion |
3. Practical Use Cases by Role
For Everyone on the Team
Drafting emails and messages
Instead of staring at a blank page, paste your rough bullet points and ask Claude:
“Turn these notes into a professional email to a client explaining the project delay.”
Summarising long documents
Got a 40-page spec, contract, or research paper? Paste it in and ask:
“Give me a 5-bullet summary of the key decisions in this document.”
Explaining unfamiliar topics
Nobody knows everything. Ask Claude to explain anything in plain language:
“Explain how OAuth works like I’ve never worked in tech before.”
Brainstorming
“We’re launching a new internal tool. Give me 10 ideas for what to name it.”
For Developers
Code review partner
Paste a function or PR diff and ask:
“Review this code for any security issues or edge cases I might have missed.”
Debugging help
Paste an error message and your code:
“I’m getting this error in production. Here’s the stack trace and the relevant function — what’s likely wrong?”
Writing tests
“Write unit tests for this function covering happy path and edge cases.”
Documentation
“Write a README section explaining how to set up this service locally.”
Using Claude Code (CLI)
Claude Code is a command-line tool that lets Claude read your whole codebase, run commands, and make changes autonomously. It’s like pair programming where Claude holds the keyboard when you ask it to.
Example workflow:
# In your terminal
claude
> Refactor the auth module to use our new JWT library. Run the tests after.
Claude will read the relevant files, make the changes, and run your test suite — pausing to ask if it’s unsure about anything.
For Project Managers & Leads
Meeting notes → action items
Paste raw meeting notes and ask:
“Extract all action items from these notes, with owner and due date where mentioned.”
Status update drafts
“Here’s what happened this week on the project. Write a stakeholder update in a professional but concise tone.”
Risk identification
“Here’s our project plan. What risks or dependencies am I not accounting for?”
Writing tickets / requirements
“Turn this rough idea into a well-structured Jira ticket with acceptance criteria.”
For Designers
UX copy
“Write 3 versions of the error message for when a user’s payment fails — one formal, one friendly, one very brief.”
User research synthesis
Paste interview transcripts:
“Identify the top 5 recurring themes from these user interview notes.”
Accessibility checks
“Review this UI copy for plain language and accessibility. Suggest improvements.”
4. Guardrails — Why It’s Safer Than You Think
One of the biggest fears is: “What if Claude does something I didn’t want?”
Here’s how the tools are built to protect you:
You always review before anything ships. Claude produces drafts, suggestions, and explanations. Nothing goes to a customer, a codebase, or a database without you approving it.
Claude tells you when it’s uncertain. It won’t confidently make something up without flagging it. You can always ask “how confident are you in this?” or “what are the risks here?”
Claude Code asks before making irreversible changes. It will pause and ask for confirmation before deleting files, running migrations, or doing anything that can’t be undone.
Your data stays in your control. Claude doesn’t train on your conversations by default in the Teams and Enterprise plans. Check your plan settings to confirm.
Start small. You don’t have to hand off a whole project. Start with one task — a single email draft, a code review, a meeting summary. Build trust through small wins.
5. Common Mistakes to Avoid
Treating the output as final. Always read what Claude produces. It’s a draft, not a finished product.
Being too vague. “Help me with this project” is hard to act on. “Summarise the risks in this project plan in 3 bullet points” gets a much better result.
Not giving context. Claude doesn’t know your codebase, your team, or your company. The more context you provide, the better the output. You can give it background upfront:
“I work at a fintech startup. Our API is in Python/FastAPI. Our users are small business owners.”
Giving it sensitive data carelessly. Don’t paste customer PII, production credentials, or confidential contracts into a public AI tool. Use an enterprise plan with data privacy guarantees for sensitive work.
6. How to Get Started — This Week
You don’t need training or permission to try this. Here’s a low-risk way to start:
- Pick one task you find tedious. An email you’ve been putting off, a document you need to summarise, a piece of code you’re not sure about.
- Open claude.ai and describe what you need in plain language — like you’re asking a smart colleague.
- Read the output critically. Edit it. Make it yours.
- Notice the time saved. That’s the whole pitch.
Once you’ve done it once, it gets easier. Within a week most people wonder how they worked without it.
7. Questions to Anticipate
“Will this replace my job?”
No. It will change what your job looks like — the repetitive, tedious parts will get easier, freeing you for the higher-value work only humans can do (relationships, judgment, creativity, accountability).
“What if it’s wrong?”
It will be sometimes. That’s why you always review it. AI tools are like a very fast intern: useful, but they need oversight. The danger isn’t that it gets things wrong — it’s using its output without checking.
“What’s the cost?”
Claude is free to try at claude.ai. Paid plans (Pro, Teams) unlock longer context windows and higher usage limits. Claude Code has a separate usage-based pricing model.
“Is my data safe?”
In the free tier, Anthropic may use conversations to improve the model. In Teams and Enterprise plans, your data is not used for training. For sensitive work, always use an appropriate plan and check the privacy policy.
8. Summary
| Claude (web/desktop) | Cowork Mode | GitHub Copilot / Cursor | |
|---|---|---|---|
| Who | Everyone | Everyone | Developers |
| How | Chat interface | Desktop app + files | IDE plugin / editor |
| Best for | Writing, research, Q&A | File tasks, document automation | Coding |
| Learning curve | Very low | Low | Low–Medium |
| Risk | Low (always review output) | Low (always review output) | Low (always review output) |
The common thread: you stay in control. These tools make you faster, not redundant.
Prepared for team presentation — May 2026