About Portfolio Cases Services Blog Contact 🎙 Talk to AI
EN DE RU
🎙 Talk to AI
November 19, 2026 · 5 min read

Your customers ask ChatGPT for a tradesman. Does it name your business

I’m Denis Shokhirev, an Agentic AI Systems Architect based in Freiburg im Breisgau, Germany. At DennisCraft AI Studio I ship production AI systems on Claude, Supabase, n8n, Doppler, and self-hosted Postgres. The moment that matters is usually boring: a property manager has a burst pipe on Sunday, opens ChatGPT, asks for a reliable local plumber, and your company is either in that answer or it isn’t. This is not mainly an SEO problem Most owners still think in terms of rankings, ads, and refer

Denis Shokhirev
Denis Shokhirev
Agentic AI Systems Architect
Telegram LinkedIn

I’m Denis Shokhirev, an Agentic AI Systems Architect based in Freiburg im Breisgau, Germany. At DennisCraft AI Studio I ship production AI systems on Claude, Supabase, n8n, Doppler, and self-hosted Postgres. The moment that matters is usually boring: a property manager has a burst pipe on Sunday, opens ChatGPT, asks for a reliable local plumber, and your company is either in that answer or it isn’t.

This is not mainly an SEO problem

Most owners still think in terms of rankings, ads, and referrals. Those still matter. But there is now a second layer between demand and your phone: language models summarize the public web and turn it into a short recommendation. If your business is hard to parse, inconsistent across sources, or described in vague marketing language, the model has less reason to mention you.

That doesn’t mean ChatGPT has a secret “best plumber” database. It means models tend to repeat businesses with clear, repeated, verifiable signals: what you do, where you work, when you answer, and why someone should trust you. If those signals are scattered or contradictory, you lose before the customer even reaches your website.

How customers actually ask

They rarely prompt with your brand name. They ask: “Who installs heat pumps near Freiburg?” “Which electrician handles emergency work in Offenburg?” “What locksmith has good reviews and answers fast?” If your homepage says “custom solutions for every need,” that is weak material for an AI-generated recommendation. If you have specific service pages, service areas, opening hours, pricing guidance, and detailed reviews, the model has something usable.

What the model can actually read

For non-technical owners, the simplest framing is this: the model is a summarizer that needs stable facts. OpenAI’s Cookbook examples on retrieval and structured outputs make the same operational point in a different context: cleaner documents and clearer structure lead to more reliable model behavior. Source: OpenAI Cookbook, 2024, https://cookbook.openai.com/. OWASP also keeps warning that poor data handling and weak validation create bad downstream behavior in AI systems and classic software alike. Source: OWASP Top 10 for LLM Applications, 2025, https://owasp.org/www-project-top-10-for-large-language-model-applications/.

SignalWhat the model infersWhat the owner should fix
Service categoryWhat you actually doUse plain descriptions like emergency plumbing, gate maintenance, heat pump installation
Service areaWhere you workList cities, districts, and response radius clearly
Trust markersWhy you are recommendableAdd detailed reviews, before/after examples, and response time expectations
AvailabilityWhether you can be contacted nowShow hours, emergency number, callback process, and lead form

This is uncomfortable but true: the better-known competitor is not always the one the model names. Sometimes it names the business with the cleanest public facts.

The 14-day fix most owners can act on

You do not need an expensive AI build to address this. In many cases, 14 days is enough for a first pass if someone on your team controls the website, your company profile, and customer reviews. That number is useful because it turns a fuzzy concern into a concrete operational sprint.

1. Replace brand copy with customer language

Write “emergency electrician in Freiburg,” “commercial door repair,” or “heat pump maintenance for apartment buildings.” Owners often underrate how much vague wording costs them. Customers do not ask a model for “quality-oriented full-service excellence.” They ask for a problem solved now.

2. Split services and locations into separate pages

A single “services” page is weaker than pages that match intent directly. “Emergency drain cleaning Freiburg” gives both search engines and AI systems far more context than a generic list with no location or scenario.

3. Remove contradictions

Phone number, hours, service area, legal entity name, and service categories should match across your site, Google Business Profile, and directories. In production systems, inconsistency is a trust failure. Public AI recommendations are no different.

4. Add structured business data

This is technical under the hood, but the business effect is simple: machines understand your company facts more reliably.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "LocalBusiness",
  "name": "Example Heating GmbH",
  "telephone": "+49-761-123456",
  "areaServed": ["Freiburg im Breisgau", "Emmendingen"],
  "openingHours": "Mo-Fr 08:00-18:00",
  "address": {
    "@type": "PostalAddress",
    "addressLocality": "Freiburg im Breisgau",
    "addressCountry": "DE"
  }
}
</script>

Where AI helps, and where a standard tool is better

This is where owners should be skeptical. If your issue is simply missed phone calls after hours, a normal telephony setup or call routing feature may be the better answer. If you need appointment booking, standard scheduling software is often cheaper and safer than a custom AI system. If your team just needs a FAQ widget, do not buy an autonomous workflow.

AI earns its place when leads disappear between email, forms, WhatsApp, PDFs, and back-office handoffs; when staff retype the same data into multiple systems; or when inquiries need triage before a person takes over. In plain business language: use AI when it cuts response time, reduces manual admin, or prevents revenue leakage.

workflow:
  - incoming_form
  - classify_request
  - check_service_area
  - store_in_postgres
  - notify_dispatcher
  - ask_human_if_uncertain
  - send_confirmation

That kind of workflow does not replace your foreman or your office staff. It keeps the request from dying in an inbox over the weekend.

The honest risk: not every business should build this

For regulated European markets, there is also a compliance layer. If you connect AI to customer inquiries, you need to know what personal data leaves your systems, which processor handles it, and how you document that under GDPR. If a standard CRM, booking tool, or phone system solves 80% of the problem, that is often the right call.

My own rule of thumb is boring on purpose: first audit the public facts, then audit the lead flow, then decide whether automation is worth it. If you only get 20 to 30 inbound requests a month and almost all of them come through referrals, a custom AI project may not pay back. Fix your website structure, your company listings, and your intake process first.

#!/usr/bin/env bash
set -euo pipefail

urls=(
  "https://example.de/"
  "https://example.de/emergency"
  "https://example.de/contact"
)

for url in "${urls[@]}"; do
  echo "Checking $url"
  curl -s "$url" | grep -E "Freiburg|Phone|Opening Hours" || true
done

FAQ

What does this usually cost?

The cheapest useful step is usually not custom AI. It is fixing your public business data, service pages, and intake flow. Public market rates checked in Germany in 2026 show local SEO and listing management often starting in the low hundreds of euros per month, while custom process automation is materially higher.

How long does it take?

A first audit and the correction of core public signals can often be done in 14 days. If you have multiple branches, an outdated website, or conflicting directory entries, expect 4 to 6 weeks for a proper cleanup.

What happens when the system fails?

For internal workflows, uncertain cases should go to a human. For ChatGPT-style public recommendations, you do not control the final answer directly, so your job is to improve the quality and consistency of the public inputs.

Does customer data leave the building?

Sometimes yes, sometimes no. With the right architecture, lead storage can stay in Postgres or Supabase and only minimal data is sent onward, but once inquiry text is processed by an external AI provider, you need to treat that as a data-processing and compliance decision.

Who is liable if something goes wrong?

The business owner remains responsible for the process. AI does not remove liability for missed leads, bad routing, or poor handling of personal data. That is why boundaries, escalation, and logging matter.

If 10 of your ideal customers asked ChatGPT today for a tradesman in your area, would it mention your company, and if not, which public signal is missing: service detail, location clarity, proof of trust, or response availability? I’d genuinely like to know.
I run a free 30-min stack audit for DACH founders building AI in regulated markets. DM me on LinkedIn or write to @ger_dennis_ai.

For your business

I build agents like this for businesses in Germany and the EU.

Tell me in two sentences what eats your team’s time – reply within 24 hours.

Custom AI agents (German page) →
Continue reading
How not to drown in MCP chaos: 50+ remote MCP servers already exist, but without curation your production agent becomes a security hole
Your AI agent breaks on mobile UI: how to connect iOS and Android via MCP and run real workflows instead of screenshot demos
One stolen key put 5,000 developers at risk: what the Cursor breach and trojanized Claude Code extension change for AI IDE security
Monday morning at a medical practice: what an AI phone assistant takes over and what it does not
All articles →
Where this is applied
Services — what we build
Talk to the voice agent
Case studies
Ready to build?

Turn your process into an AI system

Production quality. DACH B2B focus.

Start a project → ← All articles