Choosing the idea, finding the waste, picking tools, equipping the team and building the product: what changes when AI is there from the start.

Most companies add AI to an existing business. Starting with it changes almost every early decision.

We asked Magic Safari founder and CEO Philip Joubert, who was also involved in SnapScan and Luno, what he’d tell any founder building this way.

Philip set 3 rules: keep the team tiny, build something that wasn’t possible 5 years ago, and make sure the next Claude or ChatGPT release doesn’t wipe it out. That last rule killed most of his ideas. He landed on luxury safaris, a global industry worth around R320 billion with almost nothing written about how it works behind the scenes, so the models have little to learn from.

Here’s how to test whether the next AI model will kill your startup idea.

Adapted from SpaceX, his “idiot index” compares how long each task takes today with how long it would take if all the information were already in front of you. Rank the tasks by that gap and redesign from the top. It shows where AI saves the most time instead of where it looks impressive.

Here’s how to find the waste in any process with an idiot index.

Magic Safari spent 2 weeks building custom AI agents for its data work, then matched about 90% of the result on the first try with an off-the-shelf tool and a few text files. Philip’s advice is to stay near the front of AI, not at it, and let the tools improve for you.

Here’s why you shouldn’t build your own AI agents yet.

Everyone at Magic Safari gets top-tier AI accounts, not just the developers, plus a shared company brain and read-only access to the business data. He sets the bar as if everyone uses AI fully, and still expects people to think: handing in AI slop counts as not doing the job.

Here’s why every employee should have your developers’ AI tools.

Inside the product, AI adds testing, upkeep and cost. Magic Safari’s AI search took around 20 seconds and sometimes gave odd results, so they replaced most of it with ordinary filters and a thin layer of AI that reads what the customer typed. Search became instant and almost free to run.

Here’s why you should use as little AI in your product as possible.

The pattern is consistent: AI does the heavy lifting behind the scenes, and the product customers see stays simple and fast. As Philip puts it, Magic Safari isn’t an AI company. It’s a safari company that uses a ton of AI.

Philip’s full keynote from The Open Letter Cape Town 2026 is available to members inside the Founder Collab, where he goes further than we could cover here:

You’ll also get access to 40+ masterclasses from SA founders and operators on sales, fundraising, UX, paid media and more inside The Founder Collab.

Get more SA tech and business news, tips and business-building workflows and subscribe to The Open Letter.