Most businesses do not fail because they lack ambition; they fail because their systems cannot keep pace. I map that journey as five levels of business system development. We are now at level 5: AI-assisted software development—where the constraint is no longer only budget and headcount, but how clearly you set up the work so humans and AI can ship together.
This article summarises the five levels, then pulls out practical themes from my talk on why AI-native delivery can be a competitive edge—including speed, documentation, integration, and cost.
Open the video on YouTube if the embed does not load in your browser.
The five levels of business system development
- Level 1 — No system: work lives in people's heads and ad hoc messages. Useful at very small scale; it does not scale.
- Level 2 — Paper systems: forms, spreadsheets, shared drives. It scales to a point, then breaks under volume, audit, or multi-site coordination.
- Level 3 — Software system: you buy a platform. The gain is standardisation; the trade-off is that the business often changes to match the software.
- Level 4 — Custom software: you get exactly what you want—workflows, reports, and integrations tailored to you. The trade-off is cost, time, and ongoing maintenance.
- Level 5 — AI-assisted software development: delivery and documentation move at a different pace. Well-run teams are seeing order-of-magnitude improvements in throughput when the project setup is right—requirements, governance, testing, and review still matter.
What you put into the AI is what you get back out again—but there is room to operate at radically higher leverage than a classic bespoke project.
Speed, reliability, and living knowledge
When AI is embedded in engineering practice, teams can ship large, working changes faster than manual-only workflows—with documentation that stays readable and truthful about behaviour. Context does not evaporate when someone changes role: the codebase and collateral carry more of the story.
In the talk I reference how—under disciplined setup—you can produce thousands of lines of code that work the first time, because the bottleneck shifts from typing to specifying and verifying.
Why integration is the hidden prize
AI has huge implications for business integration. Historically, connecting marketing, fulfilment, customer service, production, and management reporting was slow and expensive. AI-assisted development lowers the cost of glue code, interfaces, and explainable documentation—so high-level integration becomes realistic for more mid-sized firms, not only enterprises with very large IT functions.
That is how smaller companies can behave more like bigger ones: a truly scalable system that can grow with the business, at a fraction of the historic integration cost—without pretending that change management disappears.
What to do next
If you want a grounded view of where AI, automation, and better systems can move the needle for you, book a short session with Deep Retail or explore what we do. For more articles like this, browse What's in it for me.