How Gumloop's founder built a real AI automation business

Executive overview

Most "AI automation" content on social media is marketing fiction — selling hope, not results. Genuine value from AI comes from applying it to problems you already understand deeply, not from replacing understanding with shortcuts.

Max Brodeur-Urbas built Gumloop (4 million workflows/day, clients including Instacart, Shopify, DoorDash) by failing fast, talking to users obsessively, and treating AI as an accelerant rather than a substitute.

Apply AI to what you understand deeply — automation without understanding produces slop.

The "wantrepreneur" anti-pattern

• Social media AI gurus sell a fantasy: effortless income, one-click businesses, 50 AI agents running your company
• If a $30,000-weekend workflow existed, no one would share it for free on Twitter
• "Course bros" monetise hope, not results — they found their own print-money machine selling the dream
• Vulnerable audiences in hype cycles (crypto, NFTs, AI) are the targets, not the beneficiaries
• The pattern never works for the buyer; it always works for the seller

Why Max left big tech early

• Realised the "work at big tech first, then start a company" plan leads to golden handcuffs, not a startup
• Learned nothing novel at Microsoft that he uses at Gumloop
• The only value: a logo on a resume that signals basic competence
• Years with no obligations (21–23) are finite — spending them on tickets and nine-to-fives wastes them

The deportation that forced focus

• Got turned around at the US-Canada border and banned for five years (no wrongdoing — just suspicious travel pattern)
• No fallback plan meant he had to build something serious
• Spent the next six months in a Vancouver apartment shipping MVPs as fast as possible

How to validate ideas fast

• Build quickly, then immediately hunt for reasons the idea won't work — not reasons it will
• If you can't find a reason it won't work, pursue it
• Waiting months hoping users validate your idea wastes time; early Max lost three months this way
• Failed roughly once a week: VR game moderation, bot detection, anti-scam platforms — each taught him faster

How Gumloop was actually born

• Saw Auto-GPT explode on Twitter; joined its Discord and found non-technical users asking what a terminal was
• Built a simple UI (Agent Hub) to help them — expected nothing, got traction
• Realised the agents were unreliable; users secretly wanted predictability, not agents
• Pivoted to a step-by-step automation framework — reliability over autonomy
• Discovered 80% of the audience was non-technical (ops, HR, business admins) and redesigned accordingly

YC and early growth

• Got into YC five months before the batch; spent the wait shipping
• First paid customer: $20/month on Stripe — still a user
• Missed YC's networking events because he was stuck in Canada; turned out to be an advantage
• The people building something real are not at the cocktail parties
• Investors came inbound once the product proved itself — building something exceptional makes fundraising simpler

The right way to use AI

• Use AI to accelerate things you already understand, not to skip understanding them
• Automating something you don't understand produces unpredictable, brittle output
• "Vibe coding" without fundamentals creates malware — it will come back to bite you
• AI as a learning tool (teacher + accelerant) compounds; AI as a replacement stagnates
• A split is emerging: those who use AI to understand faster will become exceptional; everyone else will produce slop

Hiring and team culture

• Almost every hire came through the network or from existing customers
• Customers from Instacart, Webflow, and Shopify quit their jobs to join — already bought in
• Hiring filter: would I want to spend 24/7 with this person?
• No mandatory hours; everyone shows up because they're excited about the mission
• You can't beg people to join — build something worth joining

The founder mindset

• There are always a hundred reasons not to start: moat questions, big-company competition
• Obsessing over those questions guarantees you never build anything
• Blind confidence — believing you can do it — is what separates founders from people who stay pawns
• The answer to "how did you get there?" is almost always: tried, failed, tried again