Skip to content
Smedja

AI transparency

I build with AI and I sell AI. So you should know where it sits. This is the state on 30 September 2026.

On this site

Nothing you see or fill in here passes through an AI model. I wrote the texts, the form sends an ordinary email, the appointment planner reads my calendar. There is no chatbot. What I do measure: visitor statistics through Vercel Web Analytics, without cookies and without a personal profile, only totals per page.

In my emails to companies

If I send you a business email without us knowing each other, an AI model (Claude by Anthropic) has first given your profile a score from 1 to 10 on job title, industry and company size, with a short reason. That decides whether you get an email and which sequence. The emails themselves were written by me; the model writes nothing to you personally. They go out on a fixed schedule, not at the moment someone reads them. More on that is on the privacy page.

In the work I build

  • If there is AI in it, it says so. A draft is called a draft, a summary a summary, a signal a signal. The signals in the sales app on this site, for example, are plain rules, not AI.
  • A tool does not make decisions about people on its own. Someone always looks before anything goes out.
  • At delivery your team gets a short guide: what the tool does, what it does not, and what not to paste into it.
  • For hiring and credit decisions, stricter rules from the European AI Act apply. I discuss that on the first call.
  • The screens on this site come from real software. Where it says demo data, you see test data. There are no AI-generated images of products or people on this site, except inside the CreatorScope screens, where the thumbnail candidates are themselves made by AI.

How I use AI myself when building

AI helps write code, tests and documentation. I read and test everything before it runs for you, and the automated tests run again on every change. Which model is used for what is in the agreement per project.

Which models

Currently: models by Anthropic (Claude), through the API, under their commercial terms, which rule out training on API data (version of 17 June 2025, read on 30 September 2026). Other models only in consultation with you.

Questions

Email hello@smedja.nl.

Subprocessors

The parties that process data for me, what for, and where they are established. For each client project I draw up the same list again, with the region of every service added.

  • ServiceVercel Inc.
    What forHosting of smedja.nl and of apps I build, server logs, cookieless visitor statistics
    EstablishedUnited States
  • ServiceSupabase Inc.
    What forDatabases: appointments, messages, data of companies I approached
    EstablishedUnited States, with data centres chosen per project
  • ServiceResend Inc.
    What forSending emails from the site and from apps
    EstablishedUnited States
  • ServiceAnthropic PBC
    What forAI model (Claude) through the API, for scoring business profiles and for AI in tools I build
    EstablishedUnited States
  • ServiceGoogle LLC
    What forMy Google Calendar, where appointments end up
    EstablishedUnited States, Google Ireland for the EU
  • ServiceMicrosoft Corporation
    What forMy Outlook calendar and mailbox for appointments and emails
    EstablishedUnited States, Microsoft Ireland for the EU
  • ServiceApollo.io (ZenLeads Inc.)
    What forSource of business contact data for my own emails to companies
    EstablishedUnited States

Where exactly the servers your project uses are located, we agree per project and I write into the data processing agreement. On this page I promise no region.

Last updated: 30 September 2026