Age of Excel coming to an end? Our conversation with Jan Zimoch, Co-Founder & CPO at Tracelight AI

Age of Excel coming to an end? Our conversation with Jan Zimoch, Co-Founder & CPO at Tracelight AI

Age of Excel coming to an end? A technological transition comparable to the dawn of the internet is happening before our very eyes. Within a few years, AI tools could make analysts’ work up to ten times more efficient. We speak with Jan Zimoch, Co-Founder & CPO of Tracelight AI, on how artificial intelligence will change financial decision-making and why Excel is becoming the “programming language” of the AI era.

Tracelight AI has just raised $3.6 million in seed funding to revolutionize the way financial models are created and verified. What key problem with traditional Excel modeling are you looking to solve?

We are currently undergoing a technological transition that happens only once every few decades. Artificial intelligence has the potential to change the way we work to the same extent that the internet once did – even if its development were to halt today. Within a few years, AI tools will enable people to work more than ten times more efficiently.

Given this context, the key problem we want to solve is the inefficiency of communication between humans and Excel. Traditional financial modeling in spreadsheets requires a tremendous amount of manual work—tedious data entry, formulas, and scenarios. This means that analysts spend 90% of their time on mechanical tasks instead of what really creates value: talking to clients, understanding business problems, and interpreting results.

In the age of AI, the role of humans should change from “Excel operators” to creators and decision-makers who delegate work to artificial intelligence and then evaluate its effects in a business context. Thanks to Tracelight, every analyst will be able to analyze up to 10 times more data and scenarios at the same time, significantly increasing productivity and the quality of decisions made.

Financial models often contain business logic that is critical to operations, hidden in complex spreadsheets. How does Tracelight AI “translate” this logic for both technical users and laypersons?

One of the greatest advantages of AI, and language models in particular, is the ability to tailor the message to the recipient. That is why Tracelight can “translate” complex financial logic into language that is understandable to both technical analysts and managers or customers without specialist knowledge.

The key aspect is that before we provide an answer, our system must thoroughly understand the financial model itself. It is precisely to this issue—the interpretation of spreadsheets—that we have devoted a significant portion of our product development efforts. Tracelight analyzes models in a similar way to humans: it examines formulas, tracks dependencies (precedents and dependents), and interprets values in the context of the assumptions and structure of the entire model.

This allows us not only to explain how a given business logic works, but also to present it in a form tailored to the knowledge and needs of a specific user – from simple, intuitive explanations to detailed, technical descriptions.

You are entering a niche but very important area – financial decision-making. How do you combine automation with the need for human oversight, especially in such a sensitive area?

This is a very important question, because in such a sensitive area as financial decisions, it is crucial to combine automation with full human supervision. Our goal is not to replace the analyst, but to create a system that allows them to easily control and evaluate the work of AI. That is why we place great importance on designing the interface in such a way that supervision and cooperation with AI are as natural and intuitive as possible.

It is also worth emphasizing that the construction of a financial model is not only the end result, but also a process—a path to a better understanding of the company and its dynamics. Therefore, the analyst must remain an active participant in this process, but no longer as a person who spends time manually “typing in” data, but rather as an operator and partner of AI.

Similarly to as when working with junior analysts or interns, we first check their work, correct mistakes, and teach them, and as trust grows, we entrust them with increasingly difficult tasks. It will be similar with AI: we always leave to the human the ability to pause, verify, and guide the system’s work. This is what ensures a balance between the speed of automation versus safety and accountability in financial decision-making.

Unlike many AI models that focus on generating text or images, your tool works on deeply structured financial data. What unique challenges and opportunities does this present for model training and accuracy?

AI has already revolutionized the work of programmers, because natural language and code are relatively “natural environments” for LLM models. Working with financial spreadsheets in Excel is a completely different challenge. Here we have two-dimensional data, numbers, formulas, formatting, and a structure that is often encoded only visually and does not resemble text at all.

That is why we dedicated a large part of our work to building a so-called middle layer, which allows the AI agent to understand the spreadsheet in a way similar to humans. Our system can analyze the structure, read formatting, comments, and section names, as well as track dependencies in the computational graph, execute formulas, and check their correctness.

This opens up enormous possibilities. On the one hand, it required the creation of a dedicated approach to training and interpreting data, but on the other, it offers the chance for precision and accuracy in an area that has been extremely difficult for AI until now. Thanks to this, we can transfer the efficiency known from tools for programmers to the world of finance – and thus dramatically increase the productivity of analysts.

From a CPO perspective, how do you ensure that product development is closely linked to both user experience and advanced AI capabilities?

As CPO, I make sure that Tracelight’s development is firmly rooted in the real needs of financial analysts while leveraging the full potential of artificial intelligence. From the very beginning, our goal was to create a tool that analysts would want to use – one that would allow them to focus on solving business problems instead of mechanical work in Excel.

That is why we place great importance on cooperation with users. I regularly conduct interviews and tests with financial analysts, whose opinions and feedback are a key element of our product development process. An additional advantage is the experience of Pete, our CEO, who spent five years at McKinsey and is very familiar with the daily practice of working in Excel.

At the same time, we want Tracelight to fit naturally into the existing workflow of analysts. That is why we decided to make the technology available as an Excel plugin – so that users can take advantage of new AI capabilities without having to change their familiar work environment.

Your founding team has experience at companies like McKinsey, Jane Street, and fast-growing startups. How has this mix of strategic, technical, and entrepreneurial expertise influenced Tracelight’s product vision?

Our experiences at McKinsey, Jane Street, and fast-growing startups shaped Tracelight’s vision at the intersection of strategy, technology, and entrepreneurship. Having worked at McKinsey for years, Pete understands the challenges of financial analysts and the importance of the modeling process itself as a way to understand the business.

Alek and I, on the other hand, started using AI tools supporting programming back in 2023. At this point, most of our code is created with the help of AI, and I act as a verifier and proofreader. Of course, I am still responsible for writing the most important parts of the code, but I automate the less important ones, increasing my productivity up to four times. This experience has shown us how radically we can increase work efficiency by delegating repetitive tasks to a machine while maintaining full control over key elements.

Excel is essentially the most popular programming language, which has changed very little in almost 40 years. While talking with Pete, we saw enormous potential in transferring AI experience to the world of financial modeling—and we’re not talking about a 20% improvement in efficiency, but a tenfold transformation. Thanks to this, something that used to take an analyst a week will soon be possible to accomplish in an hour.

Looking 3-5 years ahead, how do you think AI will change the way companies conduct financial planning and analysis – and what role can Tracelight play in this?

In a 3-5 year perspective, AI will radically change the way companies conduct financial planning and analysis. With tools such as Tracelight, analysts will be able to delegate repetitive, mechanical tasks and focus on understanding the unique context of projects or customer needs.

At the same time, the quality of financial models will improve significantly – they will be less prone to errors, and analysts will be able to analyze many more scenarios and data, enabling them to make better-informed business decisions. Tracelight acts as a catalyst here: by integrating AI into the modeling process, it enables both increased efficiency and human control over key decisions.


Jan Zimoch – Co-Founder & CPO at Tracelight AI, which he founded in 2024 together with Aleksander Misztal and Peter Fuller. A graduate of Cambridge University, where he completed the prestigious Machine Learning program, he previously studied at the University of Bristol and as a scholarship holder at Stonyhurst College.

Jan gained experience as a researcher at a financial institution and was one of the first members of the 11x team, a startup developing innovative technologies in the field of AI agentic flows. His passion focuses on combining the latest advances in artificial intelligence with practical product solutions that shape the future of technology.

Read also: Technology alone will not suffice. An interview with Martin Pauli, Global Circular Economy Services Leader, Arup

Last Updated on September 2, 2025 by Anastazja Lach

TAGS