DATA STORIES | AI & FINANCE | KNIME ANALYTICS PLATFORM
Bridging Finance, AI, and KNIME for Automation and Smarter Decision Making
My Data Guest — An Interview with Martin Dieste
I recently had the pleasure of interviewing Martin Dieste as part of the My Data Guest series. A business consultant, digital enabler, AI expert, and founder of Finnovation Flows, Martin helps businesses optimize financial processes through automation and low-code solutions. Martin is also a KNIME Contributor of the Month (Nov ’24) and KNIME Certified Trainer.
We discussed how KNIME streamlines finance workflows, from ERP migrations to financial modeling, and how GenAI can enhance decision-making. Martin also shared insights on developing KNIME extensions, the role of AI in finance, and his approach to bridging the gap between technical tools and business needs.
Rosaria: You are a finance, AI, and KNIME expert: With which role do you identify the most?
Martin: That’s a tough one! I’d say I’m definitely a finance expert, given my 15 years in finance, accounting, and controlling. With KNIME, I’m on a strong path, and AI is a broad field. There are researchers pushing boundaries in model development, and I wouldn’t call myself an expert there. Instead, I’d say I’m an AI enthusiast, early adopter, and a “use case chaser.”
Rosaria: Tell us more about what you do for work.
Martin: Right now, I run my own consulting business. I started last summer, and I’m still refining my approach, but I enjoy the journey. My focus is on finance and low-code tools, especially KNIME. I help businesses become more efficient by enabling them to use tools that don’t require extensive IT involvement.
Rosaria: What kind of work do you do with companies?
Martin: It’s quite varied. I support ERP migrations using KNIME for ETL, ensuring the right data moves between systems. I’ve also built KNIME-based time series data apps for clients to evaluate performance with KPIs. A big part of my work is replacing what I call “poorly aged spreadsheet empires” with more scalable solutions.
Rosaria: Any AI-related projects so far?
Martin: The most common AI use case I see is Retrieval-Augmented Generation (RAG). It’s a quick win because once it’s set up, it immediately saves employees time by streamlining knowledge retrieval. Some clients are exploring AI agents, but adoption is slower due to concerns around regulation, especially in Europe.
Rosaria: Let’s talk about your achievements with KNIME. You’re everywhere: Game of Nodes, YouTube videos, Medium articles, you’re a certified trainer, deliver training, etc. How hard was it to become a KNIME Certified Trainer?
Martin: I would say it wasn’t too hard. The certification path builds up your knowledge step by step. Last year, I had extra time early in the year, so I went for it, completing levels 1–3 before moving on to trainer certification. The process was well-structured, with opportunities for feedback. My final teaching submission was a session on automating report generation and distribution with KNIME’s Reporting and Email Processing extensions, where I presented to KNIME employees who provided feedback. The feedback was so positive that I later published the session on my YouTube channel.
Rosaria: Would you recommend becoming a KNIME Certified Trainer?
Martin: Absolutely! If you are into KNIME and you enjoy sharing knowledge, I’d rate it a 5/5 recommendation.
Rosaria: There are rumors you’re working on a KNIME course. What will it cover, and when can people register?
Martin: You heard correctly! The course will focus on using KNIME for finance, covering budgeting, forecasting, management reporting, and financial modeling. It’s still a work in progress, but my goal is to provide practical KNIME applications for financial planning and analysis while also covering key financial concepts to give learners a well-rounded understanding.
Beyond just learning how to use KNIME, I want participants to understand the why behind financial processes: how different techniques apply to real-world scenarios and how automation can improve efficiency. Many finance teams still rely heavily on spreadsheets, so I’ll also address how KNIME can streamline complex workflows, integrate multiple data sources, and enhance decision-making. The course will bridge the gap between technical tools and financial expertise, helping professionals optimize their processes with a low-code approach.
Rosaria: What other KNIME achievements did you have last year?
Martin: It was a big year! I became KNIME L-4 certified, was a runner-up in Game of Nodes, launched my own GenAI-focused KNIME extension using Python, and got recognized as Contributor of the Month. I also completed all 30 Just KNIME It! challenges.
Rosaria: Let’s talk about the extensions you developed. What have you built, and what’s next?
Martin: I mostly focused on AI. One of my components added structured output prompting to enrich the current AI extension, ensuring models return responses in a predefined format, reducing the risk of hallucination. Then, I experimented with the development of Python-based KNIME nodes, focusing on adding nodes for AI with vision capabilities. I’ve also been experimenting with AI agents, developing web scraping nodes to retrieve online content efficiently.
Now, I’m working on a finance extension. Some financial calculations, like loan amortization schedules, can be done in KNIME but they’re not straightforward. My extension aims to simplify tasks like that.
Rosaria: How do you decide when to create a KNIME component versus a node?
Martin: If I can build something reliable using only existing KNIME nodes and scripts, I create a component. But if an external package is needed, I build a node to make installation seamless. My AI web scraping node is a good example — doing it manually in KNIME required complex workarounds, so I built a dedicated node instead.
Rosaria: Let’s talk about AI. How do you use AI in your projects?
Martin: AI is embedded in my workflows. Even for this interview, my GPU is handling noise removal and background adjustments. I think AI is also touching up my face on StreamYard because I usually don’t look like this.
In my KNIME node development, I use AI to get assistance with code structuring. When it comes to AI research, I leverage tools that go beyond the model’s training data, like OpenAI’s web search capabilities.
Rosaria: You’re speaking at the KNIME Spring Summit in Berlin. If people want to see the “real” Martin without AI filters, they should come!
Martin: Exactly! If you want to chat about nodes, extensions, or AI in person, meet me there.
Rosaria: Let’s now talk about AI agents. They are becoming increasingly popular but definitions seem a bit fuzzy at times. How would you define them?
Martin: AI agents are systems that have a varying degree of autonomy to solve tasks by expanding their knowledge and capabilities beyond their training data. This is achieved by providing them with access to external tools and data sources. For example, a retrieval-augmented generation (RAG) agent connects to vector stores containing domain-specific knowledge, while a web scraping agent retrieves live information from the Internet. Similarly, an AI agent integrated with a database can query structured data based on user input. The key is equipping the agent with the right interfaces to dynamically interact with different sources and execute tasks efficiently.
Rosaria: Have you tried DeepSeek?
Martin: Yes, I tested the 7-billion-parameter version locally. The reasoning capabilities impressed me, but I noticed some censorship concerns. With the right prompting, you can work around certain restrictions, but the potential for curated information is something to watch.
Rosaria: AI has risks, like data protection, reliability, hallucinations. How do you manage them?
Martin: Sensitive data should stay on premise. If I need to use external APIs, anonymization is essential, and the KNIME Presidio extension helps by detecting and masking sensitive information like names, credit card numbers, and phone numbers before sending data for processing. Once the response is received, the masked data can be easily re-identified.
To minimize hallucinations, I rely on retrieval-augmented generation (RAG), ensuring models reference verified sources. Additionally, I use the KNIME Giskard extension to evaluate model quality, helping detect biases and risks before deployment.
Rosaria: You mentioned that much of your content comes from experimenting. What’s the value of trying things out and sharing your findings?
Martin: I enjoy pushing KNIME’s boundaries to spark lightbulb moments — showing that it’s not just for data science but also for finance, AI, and automation. Many don’t realize KNIME’s full potential until they see real use cases like financial modeling or AI-driven workflows.
Even in my past work with spreadsheets, I liked testing new possibilities. Now, I do the same with KNIME, sharing what I learn to help others explore and innovate. Writing and making videos also help me structure complex ideas in a way that’s easier to understand.
Rosaria: Let’s talk about the KNIME Team plan. Have you tried it?
Martin: Yes! I tested it during the free trial in December and was impressed. It’s a great way to bring KNIME’s cloud capabilities to companies that aren’t ready for Business Hub. Many of my recent videos focus on the KNIME Team plan. It lowers the barrier to cloud-based execution and automation.
Rosaria: Before we say goodbye, how can people connect with you?
Martin: LinkedIn is the best way. You can also visit Finnovation Flows, or email me at martin@finnovationflows.com.
Watch the full interview with Martin Dieste on KNIMETV:
