If you run a trades business or any small operation, you've probably heard people throw around "AI this" and "AI that" without much clarity on what it actually means for your day-to-day.
Here's a mental model that cuts through the noise: Software 1.0, 2.0, and 3.0.
Andrej Karpathy (OpenAI co-founder, former Tesla AI director) laid this out recently, and it's the clearest way to understand where your business sits — and where the leverage is.
Software 1.0: You Write the Rules
This is traditional software. You (or a developer) write explicit logic:
- If the customer owes more than $500, send a reminder.
- If it's a Tuesday, schedule the recurring routes.
Most small business software — QuickBooks, Jobber, ServiceTitan — is Software 1.0. It does exactly what it's told, nothing more, nothing less.
Good for: Predictable, rule-based tasks. Invoicing, scheduling, inventory counts.
Bad for: Anything that requires judgment, nuance, or adapting to edge cases.
Software 2.0: You Feed Data, the System Learns
Instead of writing rules, you provide examples. The system finds patterns on its own.
- You label 1,000 photos of HVAC units as "needs replacement" or "still good" → the model learns to classify new photos.
- You feed 10,000 past job estimates → the model suggests pricing for new jobs.
Most small businesses aren't running Software 2.0 directly. But you benefit from it every day: Google Maps predicting traffic, your phone's keyboard predicting your next word, Stripe flagging fraudulent transactions.
Good for: Pattern recognition at scale. Image classification, demand forecasting, anomaly detection.
Bad for: Situations where you can't get enough training data, or where explainability matters (you need to know why a decision was made).
Software 3.0: You Describe What You Want
This is the new one, and it's where most SMB owners are already operating without realizing it.
Software 3.0 means you give a natural-language prompt — a description of what you want — and a large language model generates the output.
- "Write a follow-up email to a customer who ghosted after their estimate."
- "Turn these job notes into a professional invoice."
- "Draft a Google Business Profile response to this 1-star review — keep it professional but firm."
If you've ever prompted ChatGPT, you're already doing Software 3.0.
The question is: are you doing it deliberately, or are you winging it every time?
How SMBs Should Think About This Framework
Here's the practical breakdown:
| Task Type | Software Version | Example |
|---|---|---|
| Routed scheduling | 1.0 | If/then rules in your CRM |
| Demand prediction | 2.0 | Seasonal volume forecasts |
| Customer communication | 3.0 | AI-drafted emails and responses |
| Quote generation | 3.0 | Prompt-driven estimates from job notes |
| Review management | 3.0 | Drafting personalized replies |
| Compliance tracking | 1.0 | Checklists and rule-based alerts |
Most small businesses have Software 1.0 tools (their CRM, invoicing, scheduling) and are starting to dabble in Software 3.0 (ChatGPT for emails and marketing copy).
The gap? They're not systematic about it.
3 Steps to Do Software 3.0 Deliberately
1. Build a Prompt Library
Stop rewriting prompts from scratch. Create a shared document — Google Doc, Notion page, even a paper binder — with your best prompts for recurring tasks:
- Customer follow-ups
- Estimate descriptions
- Review responses
- Job posting descriptions
- Social media captions
Each prompt should include context: who the customer is, what tone to use, what the desired outcome is.
2. Add a "Review and Ship" Step
Software 3.0 outputs are drafts, not final products. Always:
- Read the output before sending
- Edit for accuracy (AI doesn't know your specific pricing or policies)
- Add personal touches where they matter
This takes 30 seconds. Skipping it is how you end up sending a customer an email that references a service you don't offer.
3. Close the Loop
Track what works. Which prompts produced output you could use almost verbatim? Which ones needed heavy editing? Over time, you refine your prompts and your process.
This is the "loop engineering" concept applied to your business: prompt → review → ship → learn → improve the prompt.
The Bottom Line
Software 3.0 isn't some far-off future. If you've used ChatGPT to write a customer email, you're already there. The competitive advantage isn't in using AI — it's in using it deliberately, systematically, and with a review process that catches mistakes before they reach customers.
Start tonight: write down the three prompts you use most often. Put them in a doc. That's your first prompt library entry, and it's the difference between winging Software 3.0 and running it like an operation.
Top comments (0)