Build An AI-Powered Knowledge Base using n8n + Notion + ChatGPT Template with Sam Calhoun
A starter workflow anyone can set up to make onboarding, brand books, and FAQs actually useful.
n8n intimidates many marketers I talk to who are less technically inclined. They know the power of the tool, and they may have even seen one of those videos on social that says “I replaced my entire team with n8n agents, and you can too with this one workflow” but when they set it up, nothing worked.
Today, Sam Calhoun busts the myth that n8n is hard. We built an n8n + Notion + ChatGPT workflow in 15 minutes.
This week’s Stack:
1 video: Build your first n8n chatbot in 10 minutes
1 prompt: Generate landing page A/B test variations
1 tool: Free Photoshop alternative with AI upgrades
3 resources: Branching in ChatGPT, Claude’s Notion bot, Google’s image prompts
3 jobs: Fresh AI-first roles from Grammarly, Notion, and OpenAI
Let’s dive in.
Workflow Walkthrough: Ask Your Notion Database Any Question With This Easy n8n Workflow
Meet Sam Calhoun, AI workflow consultant and co-founder of The Forge.
Sam shows us why automation isn’t reserved for engineers. In this episode, he live-builds a Notion + OpenAI chatbot inside n8n that answers company questions from your knowledge base. No coding, no complexity—this is the perfect starter workflow anyone can set up in under 10 minutes.
Here’s what we cover:
– Step-by-step setup of Notion + OpenAI inside n8n
– How to connect the same workflow to Airtable or Google Drive
– Why short-term memory windows keep bots useful (and affordable)
– Practical marketing use cases like brand books, onboarding, and messaging guides
– Sam’s lean AI stack for solo operators
Grab the leave-behind:
JSON code we used for this walkthrough here
Direct link to the OpenAI Key Page here
Direct link to the Notion integration page here
Prompt of the Week: Landing Page A/B Testing Variations
If you’ve been on my team before (or will be in the future 😊 ) , you’re going to like this week’s prompt because I’m notorious for saying “let’s test it.” And one of my favorite things to test landing page copy. But, too often, I see either a sloppy job or it takes too long to do it right.
The prompt below arms you with ready-to-test landing page variations without spending hours writing from scratch. Plug in your current section, and you’ll get three UI-ready versions built on different frameworks, emotional appeals, and lengths, perfect for head-to-head testing.
You are a high-performing conversion copywriter. For the landing page section below, create three distinct variations suitable for A/B testing. Paste the current section where indicated.
[PASTE CURRENT LANDING PAGE SECTION]
Requirements (apply to every variation):
• Produce a complete UI-ready content block: Headline (one line), Subhead (one sentence), 1–3 body paragraphs (short, medium, or long per variation instructions below), Primary CTA (text + 1-line microcopy under CTA), Secondary CTA (optional), 1–2 supporting bullets (benefit-focused), image caption or creative direction, and suggested trust signal (testimonial/stat, logo or number).
• Use a different copywriting framework for each variation (see frameworks list below).
• Use a different primary emotional appeal for each variation.
• Vary length/depth across variations: Short (snackable), Medium (scannable), Long (detail + proof).
• Change the primary benefit focus for each variation.
• Include a one-paragraph strategic rationale and a clear hypothesis for the test: what you expect to change (CTR, lead form submits, demo requests, time on page, bounce rate) and why, with a suggested KPI and target direction (↑ or ↓).
• Provide a suggested A/B test config: traffic split (e.g., 50/50/50 for three-way), sample size sanity check (ballpark), and recommended success metric.
• Keep voice aligned with the original brand tone (if none given, neutral professional B2B). Avoid jargon and excessive hype.
Frameworks (choose three, one per variation):
1. PAS — Problem, Agitate, Solution
2. AIDA — Attention, Interest, Desire, Action
3. FAB — Features → Advantages → Benefits
4. BAB — Before → After → Bridge
5. 4Ps — Promise, Picture, Proof, Push
Emotional appeals to test (choose one per variation):
• FOMO / Urgency / Scarcity
• Social proof / Status / Trust (results-driven)
• Relief / Simplicity / Ease (low friction)
• Curiosity / Surprise / Novelty
• Pride / Aspiration
Length profile (use exactly one per variation):
• Short — headline + one sentence subhead + 1 short paragraph + CTA (ideal for high-traffic hero tests)
• Medium — headline + subhead + 2 short paragraphs + 2 bullets + CTA + microcopy
• Long — headline + subhead + 3 paragraphs with proof + 3 bullets + CTA + trust panel copy
Variation mapping (use this mapping unless user specifies otherwise):
• Variation 1 — FOMO / Urgency; Framework: PAS; Length: Short; Primary benefit: Time-sensitive advantage (get it before/limited seats)
• Variation 2 — Results + Social Proof; Framework: AIDA or FAB; Length: Long; Primary benefit: Specific measurable outcome (X% saves / Y ROI)
• Variation 3 — Ease / Quick Wins; Framework: BAB or FAB; Length: Medium; Primary benefit: Low effort, fast wins
Output format:
• Clearly label each variation (Variation 1 / 2 / 3).
• For each variation include: Framework used, Emotional appeal, Length profile, Primary benefit focus, Full copy block (ready to paste into a hero or mid-page section), Strategic rationale (2–4 sentences), Test hypothesis (1–2 lines: metric + expected direction), and Suggested A/B config (traffic split, primary KPI).
• Use plain text only, no HTML. Keep each variation self-contained and under 300 words.
Example hypothesis language:
“Hypothesis: Variation 2 (results + social proof) will increase demo requests by ≥15% (primary KPI: demo form submits) because it replaces abstract claims with a clear, numeric outcome and customer proof.”
I’ve added this to my to my Library. Check it out for 65+ more prompts.
The Ultimate ChatGPT Prompt Library for Marketing Leaders
I remember the first time I used ChatGPT for marketing. It was late, I was up against a deadline, and I needed a competitive analysis that would have taken me and my PMM a full day (or more) to pull …
Tool of the Week: Free Photoshop Alternative
Photopea is a browser-based design tool that mirrors Photoshop’s core features without the price tag. It’s the best one I’ve seen.
You get pro-level editing for free (though it is ad-supported so you get pop-ups), with an optional $6/month upgrade for AI-powered features that streamline design tasks like background removal and smart object editing. It’s the fastest way to make quick design edits without waiting on your creative team.
Try it: photopea.com
AI Resource Roundup
New ChatGPT “Branch in New Chat” Button: A recent update lets you spin off a brand-new chat from any point in an existing conversation. This makes it possible to create parallel threads for campaign assets, persona-specific messaging, or A/B testing prompts without losing your original context.
Claude Artifact Demo: Notion + OpenAI Knowledge Bot: Anthropic showcased how Claude can be wired into a Notion knowledge base and surfaced via a clean chat UI. It’s a glimpse into what lightweight, no-code deployments look like when paired with real company data instead of generic training sets.
11 Prompts for Google’s New Image Model: Charlie Hills shared a practical list of prompt templates covering everything from photorealistic product shots to comic panels, inpainting, and text rendering. These patterns cut down on trial-and-error and make it easier to generate production-ready visuals across multiple formats.
Hot AI Jobs
Ok, the 3 jobs that I have for you this week are all KILLER!
Account-Based Strategist at Grammarly
– Location: Hybrid (United States)
– Pay: $110K–$150K
Performance Marketing Manager – Digital Events at Notion
– Location: Not listed
– Pay: Not listed
Product Marketing Manager, ChatGPT Business at OpenAI
– Location: San Francisco, CA (Remote OK)
– Pay: $196K–$245K
Well, that’s all for this week. Now, back to scrolling LinkedIn, where I’m getting some serious FOMO seeing all of the posts from INBOUND (especially around AI).
And with Dreamforce around the corner, I know it’s only going to get worse. While everyone else is out networking in San Francisco, I’ll be at home working and putting the finishing touches on next week’s newsletter.
Until next time
– Brandon