AI in 2026

From Understanding to Action

January 2026

Today's Journey

10 min What is AI, really?
10 min The tools landscape
10 min How AI connects to everything
30 min Live demos & practical examples

AI Is Already Here

800M

Weekly ChatGPT users

34%

of US adults use AI weekly

88%

of companies using AI in 2025

Your team is probably already using it. Let's use it well.

What is a Large Language Model?

Think of it as a pattern completion engine trained on most of the internet.

"Imagine you shout a question, and everyone in the world hears it. They all start answering — but one word at a time. You hear everyone's first word and pick the most popular one. Then everyone hears that word and suggests the next word... and so on."

It's not thinking. It's predicting what word comes next, informed by patterns from billions of examples.

Training vs Context

Training Data

Everything the model learned before you started chatting.

Like everything you vaguely remember from your life — you absorbed patterns, but can't quote specific pages from books you read years ago.

Fuzzy. Pattern-based. Can't be updated mid-conversation.

Context Window

What you put in this conversation — your prompts, pasted docs, previous messages.

Like documents open on your desk right now — you can quote them, reference specific parts, reason about details.

Precise. Can be referenced directly. Limited by size.

This is why giving good context matters — it's the difference between vague recall and working with real data.

AI Limitations

Hallucinations

AI can confidently state things that are completely false. It doesn't "know" — it predicts plausible-sounding text.

Always verify important facts.

No Real Knowledge

It doesn't have access to the internet in real-time (usually). Its knowledge has a cutoff date.

Check dates on time-sensitive info.

Confidentiality

What you type may be used to train future models (depending on the tool & settings).

Don't paste sensitive data carelessly.

The Tools

What's out there and when to use what

The Big 3 Chat Tools

Tool Best For Strengths Price
ChatGPT General tasks, creative writing Most natural text, huge ecosystem Free / $20/mo
Claude Writing, coding, research Fewer hallucinations, follows instructions well Free / $20/mo
Gemini Research, Google integration Largest context window, real-time info Free / $20/mo

The best tool depends on the task. Most people just pick one and stick with it — that's fine too.

The AI Ecosystem

Beyond chat, there are specialized AI tools for almost everything:

Images & Design

  • Midjourney
  • DALL-E
  • Canva AI
  • Adobe Firefly

Writing & Docs

  • Notion AI
  • Google Docs AI
  • Grammarly
  • Jasper

Presentations

  • Gamma
  • Beautiful.ai
  • Tome
  • Google Slides AI

Video & Audio

  • Descript
  • ElevenLabs
  • Runway
  • Otter.ai

Most of these use the same underlying AI models (GPT, Claude, etc.) with specialized interfaces.

Technical Concepts

A few basics that unlock AI's full potential

Tokens: How AI Counts

AI doesn't see words — it sees tokens.

A token is roughly 3/4 of a word. "Futurestay" might be 3 tokens.

"Hello, how are you today?" = 7 tokens

"Antidisestablishmentarianism" = 6 tokens

Why It Matters

  • Context limits are measured in tokens
  • API pricing is per token
  • Longer prompts = more tokens = higher cost
  • Free tiers have daily token limits

Rule of thumb: 1,000 tokens ≈ 750 words

What is an API?

An API (Application Programming Interface) is how software talks to other software.

Think of it like a restaurant menu. You (the app) tell the waiter (the API) what you want, and the kitchen (the service) makes it and sends it back.

Examples You Already Use

  • Stripe API → processes payments
  • HubSpot API → manages contacts
  • Google Maps API → shows maps
  • OpenAI API → sends text to ChatGPT

MCP: USB-C for AI

MCP (Model Context Protocol) is a universal way to connect AI to any service.

Before USB-C, you needed different cables for every device. MCP is like USB-C for AI — one standard connection to everything.

MCP vs API

API MCP
Static endpoints Dynamic — AI can discover what's available
Request/response Persistent, real-time connection
You write the code AI figures out how to use it

MCP still uses APIs underneath — it's a smarter layer on top that makes AI integration easier.

How AI Connects to Your Tools

You
Claude Code
MCP
HubSpot / Slack / Files

With the right setup, you can ask Claude: "Find all contacts in HubSpot who haven't been contacted in 30 days and draft follow-up emails" — and it can actually do it.

Practical Applications

How this applies to your actual work

AI for Every Role at Futurestay

Customer Support

  • Draft responses
  • Summarize threads
  • Translate messages
  • Find similar tickets

Marketing

  • Plan campaigns
  • Write social posts
  • Generate ad copy
  • Landing page copy
  • Email sequences

Product

  • Write specs & PRDs
  • Summarize research
  • Competitive analysis
  • Build simple utilities

Engineering

  • Write & debug code
  • Code review assist
  • Documentation
  • Test generation

Leadership

  • Summarize docs
  • Meeting prep
  • Draft comms
  • Data analysis

Getting Better Results

The 3 C's of Prompting

  • Context — Give background info
  • Clear ask — Be specific about what you want
  • Constraints — Set boundaries (length, tone, format)

Example

Bad: "Write an email about our new feature"

Good: "You're a customer success rep at Futurestay. Write a 3-paragraph email to property managers announcing our new automated pricing feature. Tone: professional but friendly. Include one specific benefit and a call to action."

Live Demo

Claude Code + Real Futurestay Use Cases

HubSpot + Claude Code

What We'll Do

  • Connect to HubSpot via API
  • Pull recent contacts or deals
  • Ask Claude to analyze patterns
  • Generate follow-up suggestions

This shows how AI can work with your actual business data.

Content Creation Workflow

What We'll Do

  • Take a product update or feature
  • Generate a blog post draft
  • Create social media versions
  • Draft customer email announcement

One input → multiple outputs in different formats and tones.

Support Response Assistant

What We'll Do

  • Paste a real (anonymized) support ticket
  • Generate a professional response
  • Adjust tone based on context
  • Create template for similar issues

Turn a 10-minute email into a 1-minute review.

Getting Started Today

Free Options (Start Here)

  • ChatGPT — General purpose, great for most tasks
  • Claude — Excellent for writing & detailed work
  • Gemini — Good for research & Google integration

Power User Tools

  • Claude Code — Terminal-based, connects to everything
  • Paid tiers ($20/mo) — Faster, more capable
  • API access — Build custom integrations

Learn More

Documentation & Guides

Claude Docs OpenAI Docs Gemini Docs

Prompt Engineering

Prompting Guide Learn Prompting

Stay Updated

AI Explained (YouTube)

Key Takeaways

  • Context is everything — The more relevant info you give, the better the output
  • Verify important facts — AI can hallucinate confidently
  • Start simple — Use free tools before investing in paid
  • Iterate — First output is rarely final; refine your prompts
  • Be specific — Vague asks get vague answers

Questions?

Let's discuss how AI can help your specific work

Feel free to reach out anytime to explore use cases together