Skip to main content

# SERVICE 01 / 04

AI that ships to production, not just to a demo

Your AI pilot nailed the demo. Six months later it's still in staging — no MLOps, no monitoring, no owner. We ship agents, RAG pipelines and MCP integrations into live production, typically in under 90 days, with EU AI Act classification built into delivery.

<90d

to production

15–20%

productivity lift

6 wks

to first release

Built-in

EU AI Act ready

# THE-PROBLEM

Why this matters

AI pilots die in staging. Nobody built the MLOps pipeline, nobody set up monitoring, and nobody planned how the thing creates value once the demo is over. Only about 26% of companies have moved AI projects beyond a pilot.

According to McKinsey's 2024 State of AI report, only 26% of companies have moved AI projects beyond pilot stage to full production. Our <90-day delivery model exists because the other 74% aren't stuck on technology — they're stuck on operationalization.

// SOUND FAMILIAR?

  1. The POC still hasn't left staging, six months on.
  2. No CI/CD for models — retraining is manual, if it happens.
  3. EU AI Act deadline looming with no classification on file.
  4. AI can't reach real data — the APIs were never built.

# IS-IT-FOR-YOU

Where we fit — and where we don't.

// best fit for

  • You have a specific AI use case, not "we need an AI strategy"
  • Your team can point us at the data the AI needs to reason over
  • EU AI Act compliance is part of the ask, not an afterthought
  • You want the people building it on every call — not a delivery manager

// not ideal for

  • Greenfield "AI transformation" programs without a concrete first use case
  • Research-grade projects with no production target
  • Fortune 500 multi-year rollouts — we're a two-person team, not a consultancy of scale

# HOW-WE-WORK

Four steps, no ceremony.

  1. Scope & classify

    Use case, data reality and EU AI Act risk tier — settled in week one.

  2. Build the pipeline

    Agents, RAG and MCP integrations wired to your real systems.

  3. Production engineering

    MLOps, monitoring, drift detection, human oversight — the boring parts that make AI work.

  4. Launch & iterate

    Ship to your infrastructure, document it, train your team, iterate on real usage.

// TYPICAL 90-DAY DELIVERY

scope & classify
wk 1

build the pipeline
wk 2–5

production engineering
wk 6–9

launch & iterate
wk 10–12

First named release usually ships around week 6 — mid-way through production engineering, not at the end.

// typical anchors: architecture audit from €5,900 net (fixed) · fixed-scope builds from €12,000 net · fractional lead from €1,200 net/day

# THE-STACK

Production-tested tools, not a wish list.

LangChainOpenAIAnthropicRAGMCPMLOpsPythonVector stores

# DEFINITION

What is applied ai & custom solutions consulting?

Applied AI consulting is the practice of building production-ready AI systems — agents, RAG pipelines, and MCP integrations — for enterprise use cases, with a focus on measurable business outcomes and regulatory compliance. We deliver this for mid-market companies in DACH with EU AI Act risk classification and documentation built into every project.

# FAQ

Common questions.

  • How long does it take to get an AI agent into production?
    Most engagements reach production in under 90 days. Scope and EU AI Act classification are settled in week one, a first release is usually live around week six, and the remaining time goes to monitoring, hardening and hand-off.
  • Do you handle EU AI Act compliance?
    Yes — risk classification and the required documentation are built into delivery, not bolted on after. Juri handles compliance, Oleks builds the technical controls.
  • What is MCP and why does it matter?
    MCP is a standard way for AI agents to reach your real systems and data safely — the piece that's usually missing when a pilot can't touch production.

// FROM THE JOURNAL

The long-form version.

AI agents in your browser, your editor, your stack — powerful, and exactly why you need rules.

A hands-on hardening guide for AI agents across three surfaces — browser, coding agent, MCP servers. Commands, config, and a shareable rules card.

Your AI agent doesn't just read your data — it acts on it. Here's what it can reach.

Browser agents, coding agents and MCP connectors don't just read data — they act. What each can reach, and how to harden it before something goes wrong.

You installed an AI desktop app. Here's what actually happens to your data.

A desktop AI app can move client data across borders and accounts nobody logged. What that means for confidential professions — and a 20-minute self-check.

Moving past chatbots: deterministic, schema-validated AI agents with LangGraph and MCP

Why RAG chatbots stall in production, and how to build deterministic, schema-validated AI agents on AWS with LangGraph state machines and MCP.

The EU AI Act takes effect August 2, 2026. Here's what mid-market DACH companies actually need to ship.

The EU AI Act applies from August 2, 2026, with fines up to 7% of global turnover. Five practical steps to get compliant in time — not the 100-page version.

Why AI pilots fail before production — and the 90-day fix

Over 70% of enterprise AI pilots never reach production. Here are the 5 reasons they stall — and the 90-day framework that actually ships them.

Ready to put applied AI into production?

A 30-min call. We'll bring an architecture sketch and a rough number.

 book-call

// or write: hello@saloid.com · gräfelfing · de