The one thing most people get wrong

Here's the most important thing you will learn today:

During training, the model sees millions — sometimes billions — of examples of text and learns one thing: given everything I've read so far, what word comes next?

That's it. That's the core mechanism. Everything else — the ability to write emails, answer questions, code software, analyse documents — is a consequence of doing that one thing extremely well, at enormous scale.

How AI actually reads your words

Here's something that surprises most people: AI doesn't read text. It reads numbers. Before your words reach the model, they go through a transformation pipeline. Understanding this pipeline explains a lot about why AI behaves the way it does.

The three stages that turned a text predictor into an assistant

The model that comes out of the tokenisation/embedding/prediction pipeline is not the ChatGPT or Claude you use. It's a raw text completion machine — useful, but not helpful. Three training stages transform it into an assistant.

Why all of this matters for how you prompt

You now understand the full pipeline. Here's how each stage explains something you'll experience every day:

The three tools you'll use in this course

Start with ChatGPT or Claude — both free tiers are enough for Weeks 1 and 2.

Check your understanding

Why most prompts fail

Here's what most people type: "Write me a marketing email."

Here's what AI actually needs to write something useful:

• Who is writing it? (Your company, your expertise, your tone)
• Who is receiving it? (What do they care about? What stage of the funnel?)
• What is the goal? (Click a link? Book a call? Buy something?)
• What constraints apply? (Length, formality, number of CTAs)
• What should the output look like? (Subject + body? Just body? HTML?)

Without those answers, AI guesses. Guesses are generic. Generic wastes your time. The 5-Part Framework eliminates guessing.

The 5-Part Framework — every part earns its place

Real examples — see the difference yourself

Each example below shows the same goal tackled first with a weak prompt, then with the 5-Part Framework. Read both carefully. You'll start seeing exactly what's missing.

Build your own prompt live

Now it's your turn — spot the mistakes

Below are 5 real prompts that professionals actually use. Each one has problems. Your job: identify what's missing or broken in each prompt, then submit your answers. You'll see how your thinking compares across all 5 areas.

Select all the problems you can spot in each prompt. There may be more than one correct answer per question.

The refinement loop

Your first prompt won't be perfect. That's expected — and fine. Here's how to improve it systematically:

1. Run it with real input

   Don't test with a made-up example. Use something from actual work this week.

2. Rate the output 1–5

   Ask yourself honestly: would I use this as-is, or does it need major rework?

3. Diagnose which part failed

   Too generic? → ROLE. Wrong topic? → CONTEXT. Wrong format? → OUTPUT. Too long/wrong tone? → CONSTRAINTS.

4. Change one thing at a time

   Fix only the section that's wrong. This way you know exactly what made the difference.

5. Save your best version

   When you hit a 4 or 5, document it as a Prompt Card. That's your template for every future use of that task.

Step 1 — Choose your task

Pick something you do at least once a week. The best candidates are tasks that are:

• Repetitive (same structure every time, different content)
• Writing-heavy (drafts, summaries, responses, reports)
• Time-consuming but not deeply strategic (you shouldn't need to think hard about format)

Step 2 — Write your prompt

Use what you learned in Lesson 2. Here's a template you can copy and fill in:

Step 3 — Test three times

Don't just test with a perfect, easy example. Use real inputs from your work:

1. Easy test

   A simple, typical version of the task. This confirms the prompt works at all.

2. Real test

   An actual example from last week's work. Does it produce something you'd use?

3. Edge case test

   An unusual version — a difficult customer, a complex situation, an edge case. Does it break?

For each test, note: what score (1–5) would you give the output? What would you change?

Step 4 — Document it as a "Prompt Card"

A Prompt Card turns your prompt into a reusable tool. Here's the format:

Week 1 completion checklist

• ✓
  Signed up for ChatGPT or Claude (free tier)
• ✓
  Identified my repeating task
• ✓
  Written my 5-Part prompt
• ✓
  Tested it 3 times with real inputs
• ✓
  Refined based on test results
• ✓
  Saved as a Prompt Card
• ✓
  Used it in real work at least once

Why single prompts aren't enough

Imagine you want AI to handle customer feedback end-to-end. A single prompt might say: "Analyse this feedback and tell me what to do about it."

The output will be okay. But compare it to a system that:

1. First reads and classifies the feedback by type (bug, feature request, complaint, praise)
2. Then extracts the specific pain point or request from each item
3. Then scores urgency and business impact
4. Then drafts a response for each item that matches the category
5. Then formats everything into a summary report for your team

Same input. Radically more useful output. That's the power of workflows.

The core pattern: Input → Process → Output

Each arrow is a handoff. The output of Step 1 becomes the input of Step 2. Each step uses its own focused prompt — which means each step can be tested and refined independently.

The four workflow patterns you'll use most

Pattern 1: Analyse → Summarise

Used for: customer feedback, meeting notes, reports, support tickets, survey results.

Pattern 2: Generate → Filter → Rank

Used for: content ideas, product features, lead prioritisation, strategy options.

Pattern 3: Classify → Draft Response

Used for: support triage, enquiry routing, lead qualification, content tagging.

How to map your own workflow

Before you touch LangFlow, map your process on paper. Answer these questions:

1. What is the input?

   What data, text, or information starts the process? (e.g. a customer email, a form submission, a meeting transcript)

2. What is the final output?

   What should exist at the end that didn't exist before? (e.g. a report, a draft email, a scored lead record)

3. What transformations happen in between?

   Work backwards from the output. What needs to happen to get from input to output? Write each step.

4. Where does AI help?

   Mark each step: AI (language-based, pattern-recognition) or Human (judgment, approval, context).

Flowise vs LangFlow — which one is right for you?

Both tools do the same core job: let you build AI workflows visually. The differences come down to setup, interface, and where they shine. Read both cards, then commit to one for this course.

Setting up Flowise

1. Go to flowiseai.com

   Click "Get Started". You can use their cloud version (no setup needed) or self-host. For this course, use the cloud — no install required.

2. Create your first Chatflow

   Click "Add New" → "Chatflow". Give it a name like "Feedback Classifier". You'll see a blank canvas.

3. Explore the component panel

   Click the "+" button on the canvas to browse components. You'll see categories: Chat Models, Chains, Tools, Memory. Spend 5 minutes clicking through — don't build yet.

4. Add your API key

   Go to Settings (top right) → API Keys → Add your OpenAI or Claude API key. This is what powers your AI nodes.

Setting up LangFlow

1. Go to langflow.org

   Sign up for a free account. No credit card needed. The cloud version is ready immediately.

2. Create a new project

   Click "New Flow" and choose "Blank Flow" — not a template. We want you to build from scratch so you understand every piece.

3. Explore the canvas

   Click around. Add a node, drag it, delete it. Get comfortable before building anything real — 5 minutes of exploration saves 30 minutes of confusion later.

The three node types — same in both tools

Start with exactly one of each. Prove it works. Then add more AI nodes for additional steps.

Build your first test flow — Feedback Classifier

We'll build the same flow in concept for both tools. Follow the steps for whichever you chose.

1. Add an Input node

   Flowise: Add a "Chat Input" component from the Inputs category.
   LangFlow: Drag "Text Input" from the left panel onto the canvas.

2. Add an AI node with your prompt

   Flowise: Add "ChatOpenAI" (or Claude) and a "Prompt" component. In the prompt, add your system instruction.
   LangFlow: Add an "OpenAI" LLM node. Paste your prompt into the system message field.

   Use this prompt in either tool:

1. Connect the nodes

   Drag from the output port of the Input node to the input port of the AI node. Then connect AI node → Output. In Flowise the connections are called "handles" — small coloured dots on each component.

2. Run it

   Flowise: Click the chat icon at the bottom right. Type in feedback and press Send.
   LangFlow: Click the ▶ Run button. Paste test feedback into the Input field and run.

Your build checklist

• ✓
  Process map drawn on paper (from Lesson 2.1)
• ✓
  Flowise or LangFlow account created (pick one)
• ✓
  API key added to your chosen tool's settings
• ✓
  Input node added and named clearly
• ✓
  AI node(s) added — one per processing step
• ✓
  5-Part prompt written in each AI node (with {input} placeholder)
• ✓
  Output node connected at the end
• ✓
  Tested with 3 real inputs from your work
• ✓
  Output quality is at least 4/5
• ✓
  Flow exported or saved (JSON export or screenshot)

Testing your workflow

Test with three inputs — always from real data, not hypotheticals:

What makes something a "tool"?

The difference between a workflow and a tool is simple: a tool has an interface that hides the complexity underneath. Instead of "open LangFlow, load the flow, paste input, run, copy output" — your user gets a clean box, they type, they click, they get their result.

Three paths to deployment — pick one

Path A: Share from Flowise or LangFlow (do this first)

If you used Flowise:

1. Open your chatflow

   Log in to Flowise and open the workflow you built in Week 2.

2. Go to Configuration → Share Chatbot

   Toggle "Share Chatbot" to on. Flowise generates a clean chat interface — no extra work needed.

3. Copy the link and test it

   Open the link in an incognito window. Use it as a new user would. Does it respond correctly?

4. Write your user guide

   One paragraph: what should the user type? What will they receive? Any tips for best results?

If you used LangFlow:

1. Open your workflow

   Log in to LangFlow and find the flow you built.

2. Click Share → toggle Public

   Top right of the canvas. LangFlow generates a shareable form interface automatically.

3. Copy the link and test it

   Share it with a colleague and ask them to test it. Do they get the same quality output?

4. Write your user guide

   Same as above — one paragraph explaining input, output, and tips.

What is an API? (Really)

Getting your OpenAI API key

1. Go to platform.openai.com

   Create an account if you don't have one. This is separate from your ChatGPT account.

2. Go to API Keys

   In the top-right menu, find "API Keys". Click "Create new secret key".

3. Copy your key immediately

   You only see it once. Copy it and save it somewhere safe — a password manager, a private note, or an environment variable.

4. Add billing (small amount)

   Go to Billing and add a card. Add £5–10 as a credit. This is more than enough for the entire course.

5. Set a spending limit

   In Billing → Usage limits, set a monthly cap of £10. This ensures you can't accidentally overspend.

Understanding cost

You pay per "token" — roughly 4 characters of text. Here's what that means in practice:

A £5 credit will handle roughly 1,000–5,000 typical requests. For learning and testing, it's essentially free.

Path B — Building a simple HTML form

If you want your own interface beyond LangFlow's public share, here's the simplest possible HTML tool you can build, deploy, and share:

Replace YOUR_API_KEY with your OpenAI key, and YOUR PROMPT HERE with your Week 1 system prompt. Save the file. You're done with the code.

Deploying to Netlify (free, 2 minutes)

1. Go to netlify.com

   Sign up for a free account.

2. Drag your HTML file

   On the dashboard, there's a drop zone. Drag your index.html file onto it.

3. Get your live URL

   Netlify generates a random URL like amazing-pancake-abc123.netlify.app. You can keep this or set a custom name.

4. Test it

   Open the URL in an incognito window. Use it as if you're a first-time user. Does it work? Is the output useful?

How automation actually works

Every automation is built from three parts:

The trigger is what starts everything. No trigger = nothing happens. Common triggers:

• An email arrives in a specific inbox
• A form is submitted on your website
• A new row is added to a spreadsheet
• A message is posted in a Slack channel
• A file is uploaded to a folder
• A scheduled time (every Monday 9am, every day at 6pm)

Zapier vs n8n — which should you use?

Plan your automation before you build it

Answer these four questions on paper before touching Zapier:

1. What is my trigger?

   Be specific: which inbox, which form, which Slack channel, which time?

2. What data does the trigger capture?

   What information is available when the trigger fires? (e.g. sender name, email body, form fields)

3. What should AI do with that data?

   Paste your prompt from Week 1 here. Make sure it has a {placeholder} for incoming data.

4. Where should the result go?

   Which spreadsheet column? Which Slack channel? Which email address?

Setting up Zapier

1. Create your account

   Go to zapier.com. Sign up with your Google or work email. The free plan gives you 100 tasks per month — more than enough to test.

2. Click "Create Zap"

   This opens the Zap editor. You'll see a canvas with "Trigger" and "Action" already waiting for you.

3. Name your Zap

   Give it a descriptive name at the top: e.g. "Lead Qualifier — Web Form → Sheets". This helps when you have 20 Zaps later.

Step-by-step: building your Zap

1. Set your trigger

   Click on "Trigger". Search for your trigger app (Gmail, Google Forms, Typeform, Slack, etc.). Choose the specific event — "New Email", "New Form Response", "New Message". Connect and sign in to the app. Configure which folder/form/channel to watch.

2. Test the trigger

   Click "Test trigger". Zapier will pull in a recent real example from your connected app. Check that the data looks right — you'll see all the fields available (sender, subject, body, etc.).

3. Add an Action step: OpenAI

   Click the "+" below your trigger. Search for "OpenAI" (or "Claude" if available). Choose "Send prompt". Connect with your API key. In the Prompt field, paste your system prompt from Week 1. Then click into the User Message field and select your trigger's data field (e.g. "Email Body" or "Form Response").

4. Test the AI step

   Click "Test step". Zapier sends the sample trigger data to OpenAI and shows you the result. Check quality. If it's wrong, adjust your prompt in the Action field and re-test.

5. Add a second Action: the destination

   Click "+" again. Choose where the result goes: Google Sheets (add row), Gmail (send email), Slack (send message). Map the OpenAI output to the right field in your destination. Test this step too.

6. Turn it on

   Toggle the Zap to "On". Now it runs automatically whenever your trigger fires.

Testing checklist

• ✓
  Trigger fires with a real test input
• ✓
  AI step receives the correct data
• ✓
  AI output quality is 4/5 or better
• ✓
  Result reaches the correct destination
• ✓
  Safeguard (human review) is in place
• ✓
  Zap is turned ON
• ✓
  You've seen it trigger automatically once from a real event

Monitoring in Week 1

Check your Zapier dashboard daily for the first week. Look at:

Measuring real impact

This is what justifies the work — to you, to your business, to your team. Track it simply:

Write your Runbook

A Runbook is a one-page document that means anyone on your team can understand, use, and troubleshoot your automation without asking you. It's your gift to Future You and Future Colleague.

What to automate next

Now you know how it works. Identify your next two automations:

Final checklist — the full course

• ✓
  Week 1: 5-Part Prompt Card saved and used in real work
• ✓
  Week 2: LangFlow workflow built, tested, exported
• ✓
  Week 3: AI tool deployed with shareable link
• ✓
  Week 4: Automation running live in Zapier
• ✓
  Runbook written and shared with team
• ✓
  Impact measured (hours saved, cost known)
• ✓
  Next 2 automation candidates identified