Presentation

███████╗██╗███╗   ██╗███╗   ██╗
██╔════╝██║████╗  ██║████╗  ██║
█████╗  ██║██╔██╗ ██║██╔██╗ ██║
██╔══╝  ██║██║╚██╗██║██║╚██╗██║
██║     ██║██║ ╚████║██║ ╚████║
╚═╝     ╚═╝╚═╝  ╚═══╝╚═╝  ╚═══╝

AI with agency & urgency

↓

Executive Summary

01
The shift is here. AI is at >80% human quality in any discipline.
02
This is not optional. AI usage is expected at FINN. Agency and curiosity now matter more than expertise.
03
The reality check: AI won't do the hard parts — you still prioritize, decide, and own tradeoffs.
04
Start now. Pick one idea this week. Be specific. Accept imperfection.
0%
human-level quality across disciplines

The Reality Check

AI will not take your job. It will replace parts of it — the repetitive, the tedious, the "I could do this but it takes 3 hours" parts.

✓ AI can

  • ✓Write production-ready software
  • ✓Search and synthesize information
  • ✓Execute repetitive tasks at scale
  • ✓Explore possibilities you wouldn't have time to try

✗ AI can't

  • ✗Make difficult decisions
  • ✗Know your context without you providing it
  • ✗Compensate for lack of planning
  • ✗Own outcomes
"If you work them and if you learn them, you'll be, no exaggeration, 0x as productive."

Success Factors in the AI Age

What separates those who thrive from those who struggle.

[?]

Curiosity

Dive into areas outside your expertise. AI will help you. A PM can prototype code. A designer can query databases.

The barrier to entry has collapsed.

[►]

Agency

You are in the driver's seat. AI proposes; you decide. Don't accept outputs blindly.

You can literally just do things.

[■]

Knowing When to Stop

AI is a slot machine for productivity. The hardest skill: recognizing when to start fresh with a clearer prompt.

Sometimes delete everything.

[~]

Accepting Imperfection

You won't get 100%. That's fine. The goal isn't perfection; it's progress.

Be specific. Start small. Improve over time.

A map for AI tools

Quick vocabulary, then what to try.

Three Things to Know

[◉]LLM

Large Language Model — the AI "brain" that reads and generates text

Think: a very fast reader and writer

[▣]Context

What the AI can "see" — your documents, code, instructions

More context = smarter responses

[⚙]Tools

Actions AI can take: search files, run code, browse the web, send emails

Skills, plugins, integrations — all just "tools"

What Should I Try? (January 2026)

[!]
Where to start?
You probably already use ChatGPT or Claude. Good. Next step: try a coding agent like Cursor (we have licenses) — even if you don't code. They're powerful for any complex task.
Quick Start Guides→

Step-by-step setup for each tool

AI Can Learn Any Job

Give AI the right context — documents, examples, instructions — and it adapts to your domain. People are already using it for work far outside its original purpose.

Real examples from the wild

[1]
Lawyer

Analyze contracts, draft legal documents, summarize case law

Source: Anthropic legal team

[2]
Doctor / Health Coach

Interpret lab results, create health dashboards, explain test benefits

Source: Real user stories

[3]
Hiring Manager

Write job descriptions, create interview rubrics, design hiring plans

Source: Clay team

[4]
Consultant

Analyze data, generate reports, provide strategic recommendations

Source: Common enterprise use

The pattern is simple:
Context (what you give it) + Instructions (what you want) = Domain expertise

This is why "prompt engineering" matters. It's really just: be specific about what you need, and give the AI enough context to do a good job.

Go deeper

Everyone should be using Claude Code more
Lenny Rachitsky

Great overview of non-coding use cases

↗
How AI Is Transforming Work at Anthropic
Anthropic

Engineers becoming "full-stack" across disciplines

↗

Real Examples

What people are actually building with AI tools.

External

2 afternoons
Peter Steinberger

Rebuilt a fitness tracking app (originally maintained by 100 engineers)

— steipete.me
99%
Claude Code

Built 99% using Claude Code itself — recursive AI development

— Anthropic
25% + 70%
Duolingo

25% faster onboarding, 70% more PRs merged

— faros.ai

FINN-Specificinternal

spaces.finn.com

Built using AI-assisted development

by Maxi G.
Slack CLI

Full command-line tool in <10 hours with Claude Code

by Max W.
<10h
Data Team

Using Cursor with dbt repo to query warehouse without raw SQL

by Analytics
CV Screener

AI pre-screens applications and drafts recruiter summaries

by Jan Heiselbetz

What This Means for You

This is not a "nice to have." Not using AI is like not using email in 2005 — technically possible, but you're handicapping yourself.

A PM can prototype. A designer can analyze data. An ops manager can automate workflows. The barriers have fallen.

AI lets us do in hours what used to take weeks. Every improvement shipped faster is value delivered sooner.

Practical Guide

How to actually use AI effectively, step by step.

1

Think First

Spend 2 minutes clarifying what you want. The clearer your request, the better the result.

❌ Don't
"Summarize this meeting"
✓ Do
"Turn a messy meeting + Slack thread into a clean recap and action plan, then draft the follow-up email and tasks."
2

Be Specific

Vague requests produce vague results.

❌ Don't
"Make a recap from this meeting and Slack thread"
✓ Do
"From the notes below, write a 6-bullet recap for the team, then list owners + due dates for each action item."
3

Tell It What NOT to Do

Constraints help more than you think.

❌ Don't
"Create a recap and tasks from this meeting"
✓ Do
"Create a recap and tasks. Do not invent decisions or owners; mark unknowns as [TBD]."
4

Explain Why

Context matters. "This is for leadership" leads to a different result than "this is just for my team."

❌ Don't
"Draft a follow-up email"
✓ Do
"Draft a follow-up email so stakeholders who missed the meeting can approve next steps today."
5

Clean Slate When Stuck

If AI keeps failing, start fresh. Copy only essential context. Performance improves dramatically.

❌ Don't
Keep pushing the same broken conversation
✓ Do
New chat, paste only the meeting notes + Slack thread
6

Experiment

You don't need every tool. But try them. If you're not experimenting, you're losing out.

❌ Don't
Stick with what you know
✓ Do
Try a workflow tool that can create tasks for you
./appendices

Deep Dives

For those who want more.

click to expand
AI Companion(Best on desktop)