AI Use Cases

Save

In the analyzed podcast episodes, one of the most frequent patterns, above all in these departments: customer service; product and digital offering; IT and software development.

It starts with a knowledge base built from existing content: the company's own website, internal process documents or staff training material (AF-165.1, AF-141.1, AF-155.2). The bot answers from it with source references and states up front that an AI is answering (AF-115.2, AF-080.1). A fixed rule determines when it hands over to a person, and sensitive matters such as account blocking or fraud inquiries follow predefined paths (AF-145.2, AF-183.1, AF-080.1).

The bot exposes the quality of the company's own content without any filter: at BARMER a large share of the remaining project time went into outdated pages and links (AF-036.1), and the town of Burgwedel first removed duplicate and contradictory information from its website (AF-165.1). DKB has a second model check every answer before it is sent and, according to its Chief Growth Officer, reaches a resolution rate of over 80 percent (AF-156.1). The complex and sensitive cases stay with employees (AF-114.1).

Status: confirmed, 4 new supporting episodes

Supported by these conversations, 17 of them from the last three months

• #238: How a 700-Person Bank Is Using AI to Build Apps, Agents, and Digital EmployeesThe Artificial Intelligence Show · 10 Sept 2026 · 39 min · EN · 4 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• Fintechs, AI and Digital Banking: How the Financial Industry Is ChangingDas Ohr am Netz · 10 Sept 2026 · 30 min · DE · 2 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• AI as Co-CEO: How AI Integration in the Company Succeeds, with Christian Steiger of LexwareOMR Education · 31 Aug 2026 · 52 min · DE · 4 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• #95 AI Native FINN co-founder and CTO Andreas Stryz on AI, Scrum, 98% and AI-reviewed codeTech and Tales · 28 Aug 2026 · 42 min · DE · 3 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• AI at OTTOInside IT. Der deutschsprachige CIO Podcast. · 10 Aug 2026 · 42 min · DE · 1 use case
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)

Save

In the analyzed podcast episodes, one of the most frequent patterns, above all in these departments: IT and software development; research and development.

The typical sequence is this: first the requirement is described precisely, then the agent builds the code, and at the end a person reviews the pull request. mobile.de works this way, with Claude Code also writing the tests (AF-083.1), and so does Parloa, with the fixed steps specify, research, implement, validate, ship (AF-054.1). Beforehand, the context had to become readable for the AI: at mobile.de an indexed codebase (AF-083.2) as well as a style guide and security guidelines (AF-083.1), at Lauda all software projects on one platform (AF-177.6), at Upvest a technically enforced four-eyes principle (AF-089.1).

The work shifts from writing to describing and reviewing, and the person stays in charge of approval: at Thermondo no code goes into production without human review (AF-030.1), Upvest enforces an independent human review for every pull request (AF-089.1), and Parloa lets only low-risk changes through automatically (AF-054.2). According to the CIO of Goldman Sachs, around 70 percent of developer time on complex cloud migrations goes into the plan rather than the implementation (AF-122.1).

Status: confirmed, 1 new supporting episode

Supported by these conversations, 16 of them from the last three months

• #240: How David’s Bridal Is Rebuilding a 76-Year-Old Business Around AIThe Artificial Intelligence Show · 17 Sept 2026 · 36 min · EN · 4 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• Tobi Lütke: AI Agents, Better Decisions, and the Future of WorkThe Knowledge Project · 15 Sept 2026 · 66 min · EN · 1 use case
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• When AI Is Allowed to Code, Are Developers Left with Only Reading? - Benjamin HummelTestautomatisierung, KI & Agilität · 15 Sept 2026 · 35 min · DE · 2 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• Fintechs, AI and Digital Banking: How the Financial Industry Is ChangingDas Ohr am Netz · 10 Sept 2026 · 30 min · DE · 2 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• #95 AI Native FINN co-founder and CTO Andreas Stryz on AI, Scrum, 98% and AI-reviewed codeTech and Tales · 28 Aug 2026 · 42 min · DE · 3 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)

Save

In the analyzed podcast episodes, one of the most frequent patterns, above all in these departments: marketing and communications; product and digital offering.

It starts with a real image: a photo of the item on a mannequin bust, the first sample from production or a motif from the image archive (AF-136.1, AF-062.3, AF-158.1). Where standard models miss the company's own image style or alter the product, an image model is trained on it (AF-130.1, AF-162.1). The AI generates many variants, and the test in the online shop decides which image stays; About You tests 20 variants per item this way and two image models against each other, and Bonprix plans the same (AF-136.1, AF-052.5, AF-062.2). The photo shoot remains for mood, faces and skin (AF-049.3, AF-003.2).

The bottleneck moves from the photo shoot to selection and to checking whether the image matches the real item (AF-151.1). At About You the image AI initially invented extra buttons, and returns rose. According to the co-founder, only an automated check before going live brought returns back in line with those for the studio images (AF-052.4). According to the co-founder, the company reached cost savings of over 90 percent only in the third test round, after 15 months (AF-052.1).

Status: confirmed, 2 new supporting episodes

Supported by these conversations, 11 of them from the last three months

• Guardrails for AI don't slow you down, they focus you, with Valerie S., L'OréalMY DATA IS BETTER THAN YOURS · 17 Sept 2026 · 46 min · DE · 3 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• How HOLY built its own creator platform - Daniel Setzermann #26TokenMade - AI Use Cases · 15 Sept 2026 · 28 min · DE · 2 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• Eduard Dörrenberg: Wolff GPT, how a family business introduced AI faster than large corporationsBIG BANG Podcast · 14 Sept 2026 · 43 min · DE · 3 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• #85 Pragmatic AI Change Management in Mid-Sized Companies, with Matthias Holder, Head of AI @ Uzin Utz SEDigital Success | Erfolgreich den Wandel meistern · 17 Jul 2026 · 51 min · DE · 4 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)
• What Marketers Can Learn From Heinemann's AI StrategyMarketing mit KI · 15 Jul 2026 · 43 min · DE · 2 use cases
  Spotify (listen with a subscription)Apple Podcasts (listen with a subscription)