I want to describe a real moment.
Two summers ago, I sat at my mom's dining room table and tried to show her ChatGPT. She was going to use it to write product descriptions for her Shopify store. I had, in my head, rehearsed how this would go. I would explain prompts. She would write a great prompt. She would be delighted.
Here's what actually happened. She opened ChatGPT. The input box said “Send a message…”. She stared at it for about forty seconds, turned to me, and asked — completely seriously — “What do I type?”
I had no answer that didn't feel condescending. “Just type what you want” is what you say to someone who already has a mental model of what the machine is. My mom didn't. To her, ChatGPT was a blank page asking her to be a writer.
The industry's blind spot
The AI industry has an enormous blind spot about who its software is for. Every new tool assumes:
- You know what a prompt is.
- You know that the phrasing matters.
- You know what “tokens” and “temperature” might do.
- You're willing to iterate.
- You've read at least one Twitter thread about best prompting practices.
None of these assumptions describe a dental receptionist, a florist, a plumber, a 62-year-old woman running a 3D printer business out of a shop in Israel. And there are vastly more florists than there are prompt engineers.
“Non-technical” is not the right word
The industry often lumps these users under “non-technical.” It's the wrong word. My mom is not non-technical. She configures a Shopify store, manages a WhatsApp business account, tracks inventory in a spreadsheet, handles digital invoicing for the Israeli tax authority. That's plenty of technical work.
What she lacks is AI literacy. Specifically: the folk theory that AI tools require her to “ask correctly.” Her mental model of software is “click a button, get a result.” Her mental model of a chat interface is WhatsApp — where you write like a human because there's a human on the other end.
AI literacy is a skill, not a binary. Most of the people who will run the world economy in the next ten years don't have it and won't develop it, and they shouldn't have to.
Five design principles we use at Sarra
1. The interface mirrors WhatsApp, not ChatGPT.
Because that's the interaction pattern our users already trust. Send a voice note; get a video back. No “/help” commands. No slash menus. Just a conversation with something that responds like a competent employee.
2. Defaults replace decisions.
Every decision point in the flow has an opinionated default. Vertical 9:16? Default. Hook-first script structure? Default. English caption style? Default. If a user doesn't know what to choose, they don't have to. The right answer is already selected.
3. Error messages are conversational.
“Sorry, I couldn't read that link — can you try a different one?” Not “Error 422: Unprocessable entity.” The difference is a worldview.
4. One happy path, visible.
Power users can find everything. Novice users see one big button. The advanced panel is literally hidden behind a disclosure labeled “Advanced” that exists precisely so novice users know it isn't for them today.
5. Results are shown, not explained.
We don't write documentation for what an avatar is. We show you three avatars that look like real people and let you pick the one that feels right. “What is an AI avatar?” is a question novice users ask after they've used one, not before.
The market is the inverse of the hype
There's a lot of noise about AI power users and prompt experts. The actual market for AI is overwhelmingly people who will never touch a prompt, never write a custom GPT, never read an OpenAI changelog. They outnumber power users by about 100 to 1.
The tools that serve this group don't have to be less sophisticated. The underlying AI can be the best in class. What has to change is the thinking on top — the product layer — so the sophistication is hidden inside instead of exposed as configuration.
That's what we're building at Sarra. If you're one of the overwhelming majority — someone great at your business, mildly allergic to AI software, with no patience for learning yet another tool — we built this for you. My mom will vouch.
— Idan
Read this next
- What can Sarra actually do? The full tour, end to end. — the hub post: what the engine actually does, in plain language for non-technical readers.
- Which AI influencer should you create?
- How to talk to Sarra: a prompt guide for people who hate prompt engineering.