Innocap · Enterprise Enablement Program
Working Smarter with AI
Prompting & Responsible Use
A two-part program to build practical AI literacy — from your first prompt to responsible, governed use across every team. Explained here, then shown live in Claude Desktop.
Live
Demos in Claude Desktop
Press → to begin · O for overview · F fullscreen
Today · 2 Hours
How we'll spend our time together
0:00 – 0:15
Why AI, why now
Innocap's AI vision & roadmap · what AI really is
0:15 – 0:45
AI foundations
Generative AI & LLMs · capabilities vs. limitations · enterprise use cases · risks
0:45 – 1:20
Prompting & live demos
Prompt structuring · live walkthroughs in Claude Desktop (with a taste of Claude Code) · role-based examples
1:20 – 1:50
Responsible AI & governance
Policies · data protection · security · ethics · financial-services compliance
1:50 – 2:00
Wrap-up & assignments
Your take-home use cases · Q&A
Part One
01
AI & Prompting Fundamentals
Building practical AI literacy — and understanding how AI can support your day-to-day work.
Session 1
Why AI matters at Innocap
⚡
Do more, faster
Automate the repetitive — drafting, summarizing, research — so people focus on judgment and clients.
🎯
Better decisions
Synthesize documents, data and context in seconds to support sharper, evidence-based decisions.
🛡️
Do it responsibly
Adopt AI with the guardrails a regulated financial-services firm requires — from day one.
The goal isn't to replace people. It's to give every employee a capable assistant — and the confidence to use it well.
Vision & Roadmap
Innocap's AI journey
Now
Foundation & literacy
Enable every employee with safe tools (Copilot, Claude), training, and clear usage policies.
Next
Role-based adoption
Embed AI into everyday workflows across each function with tailored use cases.
Then
Governed scale
Integrated, monitored AI within a governance framework aligned to financial-services regulation.
Foundations
So… what is AI?
Artificial Intelligence is software that performs tasks we'd normally associate with human intelligence — recognizing patterns, understanding language, reasoning, and generating content.
Machine Learning
Learns from data
Improves with examples
Predicts & generates
📊
Machine Learning
Systems that learn from data
🕸️
Deep Learning
Neural networks at scale
✨
Generative AI
Creates new content
Foundations
Generative AI & Large Language Models
LLMs (the engines behind Copilot & Claude) are trained on vast amounts of text to predict the most likely next words — which lets them read, write, summarize, translate and reason over language.
📚
Trained
On huge volumes of text & code
🔮
Predicts
The next token, one step at a time
💬
Converses
Follows instructions in plain language
🧩
Adapts
To your context & examples
Think of an LLM as an extremely well-read assistant that has read almost everything — but remembers patterns, not facts. That distinction explains both its power and its pitfalls.
Capabilities vs. Limitations
What AI can — and can't — do
✓ Great at
✓Drafting, rewriting & summarizing text
✓Explaining complex topics simply
✓Brainstorming & structuring ideas
✓Extracting insights from documents you provide
✓Translation, tone & formatting
✓Coding, formulas & automation help
✗ Weak at / risky
✗Guaranteeing factual accuracy
✗Knowing real-time or private events
✗Precise math & complex calculations unaided
✗Making final judgment / accountable decisions
✗Understanding true intent or emotion
✗Keeping confidential data safe if you paste it
Rule of thumb: AI is a co-pilot, not an autopilot. You stay in the driver's seat and verify.
Real-world value
Enterprise use cases
✍️
Content & comms
Emails, reports, presentations, meeting notes.
🔎
Research & analysis
Summarize long documents, compare, extract key points.
📋
Process & ops
Draft procedures, checklists, and standard responses.
📈
Decision support
Structure options, surface pros & cons, prepare briefings.
🛰️
Monitoring
Track regulatory & market updates into digestible summaries.
⚙️
Productivity & code
Automate spreadsheets, scripts & technical documentation.
Know the risks
Three risks to always keep in mind
🌀
Hallucinations
Confident, but wrong
AI can invent facts, sources or figures that look plausible. Always verify anything that matters.
⚖️
Bias
Inherited from data
Models can reflect bias in their training data. Review outputs for fairness, especially about people.
🔒
Data privacy
What you paste, you share
Never enter confidential client or personal data into unapproved tools. Use sanctioned, enterprise tools only.
Remember: these risks are manageable — Session 2 is dedicated to using AI responsibly and within policy.
Live demo
🎬 Let me show you — live in Claude Desktop
I'll take a few real tasks and demonstrate how a good prompt turns a blank page into a useful draft — mostly in Claude Desktop, with a quick look at Claude Code.
📧
Draft an email
From 3 bullet points to a polished message
📄
Summarize a doc
10 pages → 5 key takeaways
🧮
Explain a report
Turn data into plain-language insight
👀 Watch along: follow the demo now — you'll try these yourself on your own tasks after the session.
Your AI toolkit
Working effectively with Copilot & Claude
- Lives inside Microsoft 365 — Word, Excel, Outlook, Teams
- Great for docs, email, meeting recaps & spreadsheets
- Works with your work content in the Microsoft environment
- Best for: everyday productivity in tools you already use
- Strong reasoning, long documents & nuanced writing
- Excellent for analysis, drafting, and structured thinking
- Great for exploring ideas and complex, multi-step tasks
- Best for: deep drafting, research synthesis & problem-solving
Use approved tools only. Pick the right assistant for the task — and keep confidential data within sanctioned, enterprise-grade tools.
Prompting
Prompting fundamentals
A prompt is simply your instruction to the AI. Better instructions → better results. Four habits do most of the work:
🎭
Give a role
"Act as a compliance analyst…"
🎯
Be specific
Say exactly what you want done.
📎
Add context
Paste the relevant details & goals.
📐
Set the format
Bullets? Table? Length? Tone?
Iterate. Your first prompt is a starting point — refine with "make it shorter," "more formal," "add examples."
Prompt structuring
Anatomy of a great prompt
You are a senior operations analyst at a fund-services firm.
Summarize the attached client onboarding procedure
for a new team member who has no prior context. Focus on approval steps and hand-offs.
Give me a numbered checklist, max 10 steps, plain English.
Role — who the AI should be
Task — what to do
Context — the details it needs
Format — how the answer should look
R-T-C-F — Role, Task, Context, Format. Cover these four and your results improve dramatically.
See the difference
Weak prompt vs. strong prompt
Weak
"Write about our new process."
✗No audience
✗No length or format
✗No context → generic output
Strong
"As an ops lead, write a 150-word Teams announcement for staff explaining the new expense-approval process. Friendly, clear, with 3 key steps."
✓Role + audience + length
✓Tone + structure defined
Guided walkthrough
✍️ Let's build a prompt — together
I'll build a prompt live using R-T-C-F, and you can call out a task from your role for us to try.
// Fill in the blanks
You are
[role].
[what you want done]
Context:
[the key details]
Format:
[bullets / table / length / tone]
💡 Reminder: use realistic but non-confidential examples when you practice this yourself.
AI best practices
Ten habits of effective AI users
- Verify important facts, figures & sources
- Never paste confidential or client data into unapproved tools
- Be specific — give role, task, context & format
- Iterate — refine instead of accepting the first draft
- Give examples of what "good" looks like
- Break down big tasks into steps
- Keep the human accountable for the final output
- Ask it to explain its reasoning when it matters
- Stay in policy — use approved, enterprise tools
- Save good prompts — build your personal library
Tailored to you
4×
Role-based use cases
Examples and exercises tailored to how your team actually works.
👔
Track 1
Managers & Leaders
📅
Meeting prep
Agendas, briefing packs, and talking points in minutes.
🧭
Decision support
Structure options with pros, cons & trade-offs.
📣
Communications
Clear team updates, announcements & sensitive messages.
✅
Action planning
Turn notes into owners, deadlines & next steps.
Try: "Summarize these 4 project updates into a 1-page status for my exec meeting, flag risks in red."
📊
Track 2
Business Professionals & SMEs
📄
Document analysis
Summarize, compare & extract key clauses or data.
🔬
Research
Rapidly explore topics and organize findings.
📝
Report preparation
Draft, structure & refine reports and memos.
🛰️
Regulatory monitoring
Digest updates into clear, actionable summaries.
Try: "Compare these two vendor reports and list the 5 most material differences in a table."
🗂️
Track 3
Administrative & Operations Staff
✉️
Email drafting
Professional replies and outreach in the right tone.
🗒️
Meeting summaries
Notes → decisions, actions & owners.
📁
Document management
Organize, tag, rename & format consistently.
📋
Procedure creation
Turn a process into a clear step-by-step SOP.
Try: "Turn these rough meeting notes into a clean summary with decisions and action items with owners."
💻
Track 4
Technology & Data Teams
📘
Technical docs
Generate & explain documentation and READMEs.
🔎
Technical research
Compare approaches, libraries & patterns fast.
🤖
Automation
Scripts, formulas & boilerplate to save time.
🚀
IT productivity
Debug, refactor & draft tickets and runbooks.
Try: "Explain this SQL query in plain English, then suggest a faster version with comments."
Between sessions
📌 Your practical assignment
Before Session 2, apply AI to three real tasks from your week. Bring your prompts — the good, the bad, and the surprising.
1
Pick a repetitive task you do weekly
2
Write a prompt using R-T-C-F
3
Note what worked & what didn't
We'll review your learnings together at the start of Session 2.
Part Two · ~1 week later
02
Responsible AI Usage & Governance
Using AI safely, ethically and within policy — as a regulated financial-services firm.
Session 2
Responsible AI usage starts with three questions
❓
Should I use AI here?
Is this task appropriate, and is the tool approved for it?
🔐
What am I sharing?
Is any input confidential, personal, or client data?
🧑⚖️
Who's accountable?
I verify and own the final output — not the AI.
Responsible use is a habit, not a checkbox. These questions take five seconds and prevent most problems.
Governance
Corporate AI usage & acceptable use
✓ Acceptable use
- Approved, enterprise-grade tools only
- Non-confidential or properly authorized content
- Human review of every output that's used
- Purposes aligned with your role & policy
✗ Not acceptable
- Pasting client, personal or confidential data into public tools
- Using unapproved / "shadow" AI apps
- Publishing AI output unchecked
- Using AI to make final regulated decisions alone
Data protection
Handling sensitive & client data
🚦 Classify before you paste
🟢Public / general — usually fine in approved tools
🟡Internal — approved tools, with care
🔴Confidential / client / personal — do not input
🧰 Safer patterns
- Anonymize — remove names, IDs, account numbers
- Generalize — use "a client" instead of the real one
- Use sample/dummy data for testing prompts
- When in doubt, ask — don't paste
Golden rule: if you wouldn't email it to an outside party, don't paste it into an AI tool.
Restrictions on inputs
What should never go into a prompt
🧾
Client & investor data
Portfolios, positions, holdings, account details.
🪪
Personal data (PII)
Names + identifiers, SSNs, contact details.
🔑
Credentials & secrets
Passwords, API keys, internal system access.
📑
Contracts & MNPI
Confidential agreements, material non-public info.
🏦
Regulated records
Anything subject to confidentiality obligations.
🏢
Trade secrets / IP
Proprietary models, code & strategies.
Security
Security risks to understand
💧
Data leakage
Sensitive info entered into a tool may be stored, processed, or exposed. Once it's out, you can't pull it back.
Defend: approved tools + data classification + anonymization.
🎯
Prompt injection
Malicious text hidden in a document or web page can trick AI into ignoring rules or leaking data.
Defend: be cautious with untrusted content; never let AI act on secrets automatically.
Ethics
Ethical considerations
⚖️
Bias
Watch for skewed or unfair outputs, especially about people.
🤝
Fairness
Ensure AI-assisted decisions treat people equitably.
🔍
Transparency
Be clear when AI meaningfully contributed to work.
💡
Explainability
Be able to justify outcomes — don't defer to a black box.
Ethics isn't abstract in financial services — fairness and explainability are expectations, and increasingly, requirements.
Governance
How AI is governed at Innocap
📜
Policies
Clear rules on what's allowed, which tools, and for what.
👀
Oversight
Human-in-the-loop review and accountability for outputs.
🔁
Monitoring
Ongoing review of usage, risks and effectiveness.
Governance exists to enable AI safely — not to block it. It gives everyone confidence to use these tools.
Compliance
AI regulations & compliance
🇪🇺
Emerging AI laws
Rules like the EU AI Act introduce risk-based obligations for AI use.
🔏
Data protection law
GDPR & local privacy rules govern personal data in AI inputs/outputs.
🏛️
Sector rules
Financial-services regulators expect control, records & accountability.
You don't need to be a lawyer — you need to follow policy, protect data, and keep a human accountable. Compliance handles the rest.
Our industry
Financial-services considerations
🔐
Confidentiality is core
We hold sensitive client and investor information — protecting it is non-negotiable.
🧾
Records & auditability
Decisions and communications may need to be explained and evidenced.
⚠️
Higher stakes on errors
A hallucinated figure in our context can have real financial & reputational impact.
🤝
Trust is the product
Responsible AI use protects the trust clients place in us.
Interactive
🧩 Scenario discussions — what would you do?
Scenario AThe helpful shortcut
A colleague pastes a client report into a free web AI to summarize it fast. Right or wrong? Why?
Scenario BThe confident answer
AI gives a precise regulatory figure with a citation. You're on deadline. What do you do next?
Scenario CThe hidden instruction
A document you summarize contains hidden text telling the AI to "ignore rules." What's the risk?
We'll discuss each in small groups — there's a right principle behind every scenario.
Bringing it home
Innocap use cases & your learnings
🏢 Innocap-specific opportunities
- Faster, safer document & report drafting
- Summarizing regulatory & operational updates
- Standardizing procedures across teams
- Boosting everyday productivity — within policy
🔄 Review of your assignments
- What tasks did you apply AI to?
- Which prompts worked — and which flopped?
- Where did you have to verify or correct it?
- What will you keep using next week?
Remember these
Five things to take away
1
AI is a co-pilot
It assists — you decide and stay accountable.
2
Prompt with R-T-C-F
Role, Task, Context, Format — then iterate.
3
Always verify
Especially facts, figures and sources.
4
Protect data
Approved tools only — never paste confidential info.
5
Use it responsibly
Follow policy, respect ethics, keep a human in the loop.
Thank you
Let's build an AI-confident Innocap
Start small, stay curious, stay safe. Pick one task this week and let AI help — the right way.
💬
Questions?
Now's the time — let's talk.
📚
Resources
Policies & prompt library to follow.
🚀
Next steps
Apply it — bring learnings back.
Innocap · AI & Prompting Training Program