The “3 AI Terms a Week Keep the Buzzwords at Bay” edition, Part 3

Welcome back to our summer series - where we break down three AI terms a week, minus the buzzwordzzzzzzzzzz fatigueeeeeee.

This week we’re digging into how AI reads, remembers, and sometimes... forgets. The curse of creative minds with short attention spans like mine.

1️⃣ Token

🧠 What it is: A token is a tiny chunk of text - a few characters, a word, or part of a word. AI models read and generate text one token at a time.

🚫 What it’s not: It’s not a full word or sentence. “Marketing” is one token, but so is “Mar” or “ing” depending on the model.

🛠️ Real example: ChatGPT can process around 4,000 to 32,000 tokens at once (depending on the model) - roughly 3 to 25 pages of text.

📈 Why this matters to us: Token limits shape how much the AI can handle in one go - and also affect cost. Most tools charge by token. If you're embedding AI in your business, understanding tokens = understanding scope and spend.

2️⃣ Context Window

🧠 What it is: The context window is how much the model can “see” or “remember” at once - measured in tokens. It’s like the AI’s working memory.

🚫 What it’s not: It’s not permanent memory. Once a conversation or document goes beyond this limit, the AI model starts forgetting the earlier parts.

🛠️ Real example: An AI assistant summarizing a 50-page board report may miss key points from the beginning if its context window only fits 30 pages.

📈 Why this matters to us: When inputs get too long, earlier content can get dropped - leading to answers that miss the mark. Some newer models now support much larger windows, and as costs fall, they’ll become more common in everyday tools.

3️⃣ Context Management

🧠 What it is: Context management is the practice of giving AI just the right amount of relevant input - not too little, not too much - so it can respond accurately and efficiently.

🚫 What it’s not: It’s not dumping everything into the prompt and hoping for the best. If the input is too long or unfocused, the model may miss the point or skip important parts.

🛠️ Real example: A leadership team prepping for an investor Q&A uses AI to draft talking points. Instead of pasting a 50-page earnings report, they feed the summary section plus key analyst questions - keeping it sharp, within the model’s limits.

📈 Why this matters to us: The quality of AI output depends on how well we manage context. Teach teams to focus inputs, trim excess, and guide the model clearly - especially when accuracy matters. It's about the right questions.

🧠 Bringing it All Together - Why This Matters for Leaders Like Us

Tokens, context windows, and context management might sound technical, I know.

But understanding these basics helps us get better results, ask smarter questions, and avoid the most common AI frustrations I have witnessed.

I'll see you next week, and will bring more terms with me.

—

Hey, I’m Marco - I write, speak and train execs on #marketing and #AI. Enjoyed this? 👉 Connect or follow me - Marco Andre