How Telos Builds Software
The layered architecture Telos uses to build customer applications — from platform through domain, tools, and workflows — and how the blueprint and AI-assisted development drive the process.
How Telos Builds Software
Telos builds software in layers, each one grounded in the work done before it. The approach is deliberate: technical decisions are made early by senior engineers, domain knowledge is captured precisely, and AI is embedded throughout — not bolted on at the end.
The blueprint (see TEL201) drives every layer. It is maintained and enriched throughout development, becoming an increasingly complete record of the software and the domain it serves.
Bottom-up, not top-down
Telos builds from the bottom up. Many tools (such as Lovable and similar AI-first builders) start from the top — describe a user experience, generate a UI, and hope the back end follows. Telos does the opposite: architecture and domain first, user experience last.
This matters because a well-modelled domain produces better software than a well-designed interface. Once the domain is correct, UI can be generated, iterated on, and changed cheaply. The reverse is not true — a beautiful interface built on a weak domain creates technical debt that compounds over time.
Design through AI mockups
Once the domain and tools layers are in place, Telos uses AI image generation to produce UI mockups for decision-making — not as design deliverables. The App Director generates multiple options based on the established style guide and domain language. There is no lorem ipsum; the mockups use real terminology and reflect the actual data model. The customer reviews the options and picks a direction. That decision then flows directly into development with full context. This makes design fast and removes the need for a dedicated UX designer in most cases.
The layers
1. Platform Layer
Architected by senior Telos engineers before any customer-specific work begins. This layer covers:
- Technology stack selection
- Application architecture
- Security standards
- Performance design
Customers do not make these decisions. Telos owns the technical platform and ensures it is sound before building anything on top of it.
2. Domain Layer
Built directly from the blueprint. This layer translates the customer's business domain into the application — the language, data structures, entities, and relationships that reflect how the business actually works. Getting this right is critical: the domain layer is the foundation everything else sits on.
3. AI Tools and UI Tools
Built on top of the domain layer. These are the discrete, functional capabilities the system exposes — the things a user or an AI agent can do. For every action a user can perform, the AI can perform the same action, and vice versa. This symmetry is intentional: it means users and AI can work together within the same system without duplication or gaps.
4. Workflows
Built last, and in two tracks:
User workflows — the experience a human user has when working in the system. These include dashboards, job-based flows, and daily interaction patterns. They are designed carefully because they represent how people spend their time. Leaving these until the tools layer is complete ensures they are built on solid, tested capabilities rather than assumptions.
AI workflows — every point in the system where the AI interacts with a user or performs an action autonomously. These are designed alongside user workflows to ensure the AI operates coherently within the same environment.
AI in development
Development is heavily AI-assisted throughout. The blueprint is the primary input to the AI development process — it provides domain context, stories, and patterns that allow AI tools to generate accurate, relevant code. As development progresses, the blueprint is updated to reflect what has been built, creating a feedback loop that keeps it current and useful.
This is why the blueprint is treated as a living document, not a one-time deliverable. By the time the software is live, the blueprint is a rich, detailed record of how the system works — and the foundation for maintaining and evolving it over time.
Skills at runtime
Once the application is live, the customer's skill book (see TEL202) is loaded into the embedded AI agent. The skills tell the agent how to operate the software, interact with users, and apply the business's domain knowledge — turning a well-built application into an intelligent, business-aware system.