Agentic AI at mid sized companies
Key Insights from Our Event with BVMW “AI in Practice”:
- Agentic AI at mid-sized companies succeeds where the foundation for AI (AI Foundation), impact measurement, and people enablement are considered from the start.
- Six breakout rooms delivered practical insights on Semantic Layer, AI competence, Marketing Mix Modeling, AI Search Visibility, Dashboards, and Agentic AI.
- With structured business context, we see a jump in AI system answer accuracy from around 21 to over 95 percent in our projects.
- Agentic AI is already operational at mid-sized companies, from the construction site to procurement.
What separates AI experiments from real business value at mid-sized companies: business context, clear ownership, and ongoing impact measurement, considered from the start. That is the central takeaway from six parallel breakout rooms we hosted on July 28, 2026, together with the German Association for Medium-Sized Businesses (BVMW) at our event “AI in Practice” at our Munich office. Instead of a lecture format, we deliberately set it up as a “barcamp-style” event, working directly with marketing, sales, and IT stakeholders on their open questions about Semantic Layer, AI competence, Marketing Mix Modeling, AI Search Visibility, Dashboards, and Agentic AI.
We’ve gathered the takeaways from these six rooms here as a snapshot of where mid-sized companies in Germany stand on AI in 2026, and where the biggest opportunities lie.
What Was the Event About?
“AI in Practice” was our second joint in-person event with BVMW, of which we are an official member. After a short keynote, attendees chose two of six breakout rooms before all takeaways were brought together in a joint closing round. This created a well-rounded picture of what’s actually on the minds of decision-makers at mid-sized companies when it comes to AI.
Why Impact Measurement Determines Success or Margin
Jörg Hopmann
Susanne Hopmann
Keynote, Part 1 · Jörg Hopmann, Managing Director, Hopmann Marketing Analytics
Jörg opened the evening with three adoption patterns that apply to almost every mid-sized company:
- The Cautious Ones: AI is already being used in the background, but official guardrails are still missing.
- The Full-Throttle Group: a lot of energy goes into AI, but more volume doesn’t automatically mean more impact.
- The Searchers: proceed in a structured way, but are still building AI knowledge without sufficient business context.
Using a twelve-month model calculation, Jörg showed how quickly rising AI costs can erode results without impact measurement in place.
Keynote, Part 2 · Susanne Hopmann, Managing Director, Hopmann Marketing Analytics
Susanne took over for the second part and introduced the Hopmann AI Foundation, our five-layer model spanning data quality, Semantic Layer, and Context Layer through to Agentic Marketing Intelligence. Using the model, she showed where many companies stand today and how the individual steps combine to create the greatest leverage.
Key Takeaway: Effective AI use doesn’t start with as much AI as possible, but with carefully prioritized use cases, ongoing impact measurement, and a solid data foundation.
The Jump from 21 to 95 Percent Answer Accuracy
Tobias Lanzl
Room 1 · Manager Data Analytics, Hopmann Marketing Analytics
Tobias showed in his breakout room why AI absolutely needs context to work precisely: the necessary domain knowledge exists in people’s heads, but in most systems it isn’t documented in a structured way. With a structured Semantic Layer, meaning consistently defined metrics and calculation logic, we see a jump in answer accuracy from around 21 to over 95 percent in our projects. When AI makes mistakes, they can’t simply be ignored: corrections and insights need to flow consistently back into this context so they don’t repeat. For this to work, writing down knowledge once isn’t enough, maintaining the Semantic Layer has to become a fixed part of the team’s daily workflow.
Key Takeaway: The return on an AI project isn’t decided by the model, it’s decided by whether knowledge is documented in a structured way and error corrections consistently flow back into the Semantic Layer and Context Layer.
Why Tool Access Alone Doesn’t Create AI Competence
Dr. Özgün Köksal
Room 2 · Senior Data Analytics Specialist, Hopmann Marketing Analytics
What quickly became clear in Özgün’s room: many employees are genuinely open to AI, but there’s often a lack of clear guidance, hands-on training, and dedicated learning time. Tool access alone isn’t enough. The central challenge is identifying truly useful AI use cases, starting from real, recurring tasks and concrete pain points, not from the technology itself. This works best when domain expertise and AI knowledge come together in hands-on workshops to develop and prioritize use cases. As next steps, Özgün recommends role-specific training, clear guidelines, dedicated learning time, and regular knowledge-sharing formats within teams. We go deeper on this approach in our post Building AI Competence Within a Team.
Key Takeaway: AI competence is a leadership topic. The lever isn’t tool access, it’s identifying the right use cases from real, everyday tasks.
When Marketing Mix Modeling Really Pays Off
Dr. Simon Hannemann
Room 3 · Manager Marketing Science, Hopmann Marketing Analytics
Simon’s breakout room on Marketing Mix Modeling (MMM) gave a solid introduction to the topic and to how MMM can measure impact more effectively and support better, data-driven budget allocation. Using concrete examples, Simon made it tangible where the method works best: with sufficient transaction volume and short to medium purchase cycles. With long B2B sales cycles spanning several years, other measurement methods often get you there faster, though that needs to be assessed case by case. The group agreed on one thing: measurability of marketing activities is becoming increasingly important for sound budget decisions.
Key Takeaway: MMM is experiencing a genuine comeback as a marketing analytics method. The right question before adopting MMM isn’t “Do we want this?”, it’s “Does this fit our business model?”
How Visible Is Your Brand in ChatGPT & Co.?
Constanze Bogner
Room 4 · Manager Marketing Technology, Hopmann Marketing Analytics
Constanze’s breakout room, especially diverse with attendees from B2B, B2C, and startups, focused on Generative Engine Optimization (GEO): the discipline that secures visibility where customers increasingly ask ChatGPT, Perplexity, and other LLMs for recommendations instead of running a classic search. The discussion centered on evaluating suitable AI visibility tools, building meaningful prompt sets to analyze your own visibility, and the experience from a project where we deliberately built our own audit pipeline instead of using a standard tool. Attendees also discussed whether good classic SEO makes GEO unnecessary, how prompt language and location affect results, and which other sources like LinkedIn, YouTube, and others influence the outcome.
Key Takeaway: Good SEO is no substitute for GEO: AI search systems follow their own evaluation logic, and those who want to be visible there need their own criteria and their own monitoring.
Dashboard or AI: Who Do We Trust with the Number?
Dr. Ioanna Kakoulidou
Room 5 · Senior Data Analytics Specialist, Hopmann Marketing Analytics
Trust was the big theme in Ioanna’s room: AI makes working with data faster and more convenient, but a well-phrased answer can still be wrong. That’s why data quality, critical thinking, and keeping a human in the loop remain the load-bearing pillars. Technology is only part of the task: companies also need to clarify what customers actually allow, whether the cost of AI is justified by the value it creates, and what knowledge and context AI needs to deliver reliable answers. Asked whether dashboards are still needed, the group arrived at a clear answer: getting numbers onto a dashboard is the easy part, the real opportunity lies in rethinking how people will access, understand, and act on information in the future.
Key Takeaway: Dashboards aren’t going away, their role is growing into one of several trustworthy gateways to AI-powered insights.
From the Construction Site to Procurement: Agentic AI in Practice
Jörg Hopmann
Room 6 · Managing Director, Hopmann Marketing Analytics
In Jörg’s room, attendees first defined together what Agentic AI actually means before the discussion moved into concrete examples: automated quote and invoice generation in skilled trades along with automated project management directly on the construction site, agentic specification creation and supplier search for procurement optimization in industry, and conversational AI applied directly to marketing data. The group also touched on the difference in AI adoption speed between China and Germany, which was very interesting. Attendees traded their own practical tips. What stood out most was the mix of AI-savvy specialists and management-level decision-makers in the room.
Key Takeaway: Agentic AI is already operational at mid-sized companies, and the edge goes to those where domain expertise and AI competence sit at the same table.
What Event Is Coming Up Next?
After the event is before the event. That’s why we’re already inviting you, together with BVMW, to the next date: From AI Investments to Measurable Value Creation. Join us for a panel plus further interactive sessions with fresh perspectives on AI in marketing. All upcoming dates are updated continuously on our events page.
Thank you to all attendees for the open questions and active participation across six breakout rooms, and to Ingrid Janssen from BVMW for the smooth organization. A special thank-you also goes to our six breakout room leads for their engaging sessions, and to the entire Hopmann team, who supported everything behind the scenes.
Head of Marketing, Hopmann Marketing Analytics
Leads marketing at Hopmann, with a background as an industrial engineer and nearly 20 years of marketing experience at international B2B and B2C companies.
What our customers often want to know.
FAQs About Agentic AI for Small and Medium-Sized Businesses
What is Agentic AI, and why is this topic now relevant to SMEs?
Agentic AI refers to AI systems that independently plan and execute multi-step tasks, rather than simply responding to individual inputs. For SMEs, the economic benefit lies in the ability to automate repetitive processes without having to build their own AI team.
How does a well-maintained business context affect the accuracy of AI responses?
With a structured semantic layer and context layer, response accuracy increases in practice from around 21 percent to over 95 percent. If the context is actively maintained, this level remains stable.
At what point does marketing mix modeling become economically viable for a company?
MMM is particularly suitable when there is sufficient transaction volume and short- to medium-length purchase cycles. For long B2B sales cycles, other measurement methods often yield faster, more reliable results.
What distinguishes GEO (Generative Engine Optimization) from traditional SEO?
Traditional SEO optimizes for a ranking of links, while GEO optimizes for being cited in a single AI-generated response. Both disciplines follow their own evaluation logics and measurement approaches.
How do Medium-sized businesses specifically benefit from impact measurement in AI projects?
Impact measurement reveals which AI use cases actually contribute to revenue or efficiency. This allows budgets to be allocated strategically where the impact is greatest, rather than investing broadly.
When is the next AI event hosted by Hopmann and the BVMW?
The next joint event is the AI Evening for Marketing Managers in October. All other dates are regularly updated on our events page.