# EPSIA GmbH — AI enablement for SMEs

> Machine-readable full text for AI systems and LLM crawlers. Stand-alone,
> without navigation, advertising or layout. Short version: /en/llms.txt

## What EPSIA does

EPSIA is an AI enablement partner for German SMEs. We solve two
problems:

1. **Making existing software AI-ready.** Many small and medium-sized companies have
   codebases (PHP, Java, .NET, JavaScript) that have grown over the years. AI coding
   assistants such as Claude Code or GitHub Copilot only work reliably in such
   projects if the structure, tests, typing and documentation are sound.
   We measure this maturity with an AI-Readiness Score and raise it
   in a targeted way.

2. **New development AI-optimised from day 1.** When new software is created, we build
   it so that humans and AI can work on it productively over the long term.

## The AI-Readiness Score

The AI-Readiness Score is a number from 0 to 100 that indicates how well a
codebase is suited to working with AI tools. It is calculated from
8 weighted dimensions.

### The 8 dimensions and their weighting

1. **Architecture (20%)** — Modularity, clear boundaries, coupling. The cleaner the
   structure, the better an AI can make local changes without triggering
   unexpected side effects.
2. **Code Quality (15%)** — Readability, consistency, complexity, duplicates.
   Understandable code is easier for both humans and models to change correctly.
3. **Typing (15%)** — Static types and signatures. Types are guardrails:
   they give the AI verifiable context and catch errors early.
4. **Testing (15%)** — Coverage and expressiveness of the automated tests. Tests
   are the safety net that makes AI changes immediately verifiable.
5. **Documentation (15%)** — README, inline docs, context files (e.g. CLAUDE.md).
   Good documentation gives the AI the “why” that is not apparent from the code
   alone.
6. **DevOps (10%)** — CI/CD, automation, reproducibility. Automated
   pipelines catch faulty AI contributions before they reach production.
7. **Stack (5%)** — Modernity and adoption of the technologies used. For
   common, up-to-date technologies a model has more reliable training knowledge.
8. **Security (5%)** — Known vulnerabilities, handling of secrets, dependencies.
   Assesses the baseline security level of the codebase.

The sum of the weights adds up to 100%.

### How the analysis becomes a score

Each dimension is rated on a scale from 0 to 5. The overall value is derived
from two tracks:

- **Track A — Automated metrics (40%).** Objectively measurable indicators:
  test coverage, degree of typing, results of linter and security scans
  (including npm audit), presence of CI/CD, documentation density. These are
  translated into sub-scores via stored thresholds.
- **Track B — AI-driven analysis (60%).** An AI model (Claude Code) reads the
  codebase and assesses what metrics alone cannot capture: Is the architecture
  really sensibly structured? Are the tests meaningful or just a façade?
  Does the documentation explain the “why”? Every assessment is backed by concrete
  evidence (file:line), strengths, weaknesses and recommended actions.

Formula per dimension: `combined_score = metric_score × 0.4 + ai_score × 0.6`

Overall score: `AI-Readiness Score = Sum over all dimensions (combined_score / 5 × weight × 100)`

### Score categories

- **0–39 — Critical:** AI assistants deliver unreliable results here.
  Fundamental modernisation recommended.
- **40–59 — Has room to grow:** AI is usable in places, but with risk. Targeted
  measures raise the score significantly.
- **60–79 — Solid:** AI tools can be used productively. Fine-tuning brings
  further efficiency.
- **80–100 — AI-Ready:** Humans and AI work together smoothly.

## Services and Prices

### 01 — Quick Check
- Price: from €990
- Duration: 1 day
- Includes: Automated code analysis, scorecard with radar chart of all
  8 dimensions, 30-minute results call.
- For whom: A first objective assessment without a large budget.

### 02 — AI-Readiness Assessment
- Price: €4,500–8,500
- Duration: 3–5 days
- Includes: Full 8-dimension analysis, hands-on AI test with 3 real
  tasks in the actual codebase, prioritised modernisation roadmap,
  effort estimate.

### 03 — AI-Ready Modernisation (core offering)
- Price: €15,000–120,000
- Duration: 4–24 weeks
- Includes: Refactoring, building up tests and typing, CI/CD pipelines,
  context files (CLAUDE.md) — the codebase becomes measurably AI-ready.

### 04 — AI Enablement Training
- Price: €3,500–6,000
- Duration: 2–3 days
- Includes: Hands-on training so that developer teams can use AI coding assistants
  productively and safely.

### 05 — AI-Ready Retainer
- Price: €1,500–4,000 per month
- Term: from 6 months
- Includes: Ongoing support so that the AI-Readiness Score achieved is
  maintained and continues to rise.

### 06 — AI-Assisted New Development
- Price: €25,000–200,000
- Duration: 2–6 months
- Includes: Building software with AI-optimised architecture from day 1.

## Frequently Asked Questions

**Is our code modified or uploaded?**
During the Quick Check and Assessment the code is only ever read, never
modified. There is no telemetry and no cloud uploads; the analysis
logs only paths and line numbers, no code snippets.

**Which technologies are supported?**
The approach is language-agnostic. Common focuses among German SMEs are PHP, Java,
.NET, JavaScript/TypeScript.

**Is it eligible for funding?**
The Quick Check is deliberately kept below the €1,000 threshold. Larger projects are
often eligible for funding depending on the federal state and programme.

**What sets EPSIA apart from a classic software agency?**
The focus is exclusively on making software fit for working with AI
— with a measurable, verifiable score instead of gut feeling.

## Contact and company

- Company: EPSIA GmbH
- Managing Director: Daniel Erbert
- Address: Siegfriedstraße 152, 10365 Berlin, Germany
- Phone: +49 1573 7731059
- Email: contact@epsia.io
- Web: https://www.epsia.io
- Commercial register: Amtsgericht Berlin-Charlottenburg, HRB 194525
- VAT ID: DE317515217