AI Agents in Practice: Platform, Sovereignty and a Real Use Case
This episode was recorded in German. YouTube may offer a dubbed audio track.
Show Notes
A client wanted to know where their AI agent actually runs. The question sounds simple, and it leads straight to the decisions companies have to make right now. For our event “AI in B2B Commerce” on 17 November 2026 in Munich, we built an agent that turns a data sheet into a purchasable product in the shop in just under a minute, complete with item, price, promotion and landing page. Matthias Dietrich and Philipp Krueger talk about it.
At the centre is the question of sovereignty. Anyone building agents today quickly ends up placing their business logic with the model provider, and later finds it hard to get out again. We show the layer model that brings order to this: workflow engine, runtime, model access and, across all of them, identity, permissions and costs. We also explain how the agent came about, why a human approves every step and why a run costs only a few cents.
Where should your business logic live? What happens when you want to switch models? How do you bring sales, engineering and finance into one shared process chain? And why do we measure AI against the mistakes people make every day?
More on the event “AI in B2B Commerce” 2026 in Munich (in German): foobar.agency/de/events/ki-im-b2b-handel-muenchen-2026
What Works Tomorrow. The podcast for decision-makers in retail and industry. By Matthias Dietrich and Philipp Krueger, Co-CEOs of foobar Agency.
#AI #AIAgents #Sovereignty #B2BCommerce #WhatWorksTomorrow
About the Podcast
What Works Tomorrow
What Works Tomorrow is the podcast for leaders in retail and industry. Philipp and Matthias talk with decision-makers about how companies use data, AI and commerce to reach their…
All EpisodesThe Hosts
Matthias Dietrich
Co-CEO
Philipp Krueger
Co-CEO
Subscribe
Chapters
- 00:00Welcome and the ROI of AI agents
- 02:55Where does the agent run? A client's question
- 04:16Platforms, clouds and vendors
- 06:23Sovereignty when switching models
- 11:33The layer model
- 14:14The use case: from data sheet to product
- 16:59Human in the loop
- 20:46What a run costs
- 23:10Business units in one process chain
- 25:08Errors and adoption
- 28:39Zwei Kurven und unser wichtigster Rat
Related Articles
Agentic Commerce: What Retailers and Manufacturers Should Prepare Now
A year ago, shopping was supposed to move into ChatGPT. What became of agentic commerce, where revenue and savings lie today and which five steps pay off now.
AI in B2B Sales: How Assisted Selling Multiplies Your Sales Efficiency – Without Replacing Your Sales Reps
Discover how Assisted Selling empowers your sales team—with data-driven recommendations, not automation. Achieve up to 400% revenue growth.
Benchmarking Open-Source LLMs
Learn how to benchmark open-source large language models and select the right one for your needs.