Startup Profile
pi_optimal

Introduce yourselves

Tell us in three sentences who you are

pi_optimal develops workflow intelligence systems that reliably automate operational decision-making while learning from every interaction. Our processes grow alongside their users and continuously improve at taking the right actions at the right time. pi_optimal was founded in November 2024.

How did the idea come about and what challenges had to be overcome?

The idea emerged from combining hands-on experience at a digital marketing agency with research work at the Max Planck Institute for Intelligent Systems. In the agency, we experienced daily how much time is spent on repetitive decision-making processes and what happens when those decisions are suboptimal: budgets are wasted, teams lack time for strategic work, and overall performance suffers. After initial positive tests in research environments, we are now focused on transferring both automation and learning approaches into real-world applications.

How does Archiplan work and how does AI support this project?

Users describe their workflows in natural language, they simply state what they want to achieve. pi_optimal orchestrates these workflows and executes them automatically. We follow a human-in-the-loop approach: humans remain involved at critical moments and retain control. Using reinforcement learning methods, our systems learn from every interaction, continuously improving decision quality and evolving together with their users.

In which areas is your technology currently being applied?

We currently focus on digital campaign management. Our system takes over the operational layer, the daily optimizations that would otherwise require significant manual effort. This includes pacing and budget control, continuous keyword management, and performance optimization based on real-time data. Our goal is to achieve operational excellence, which often fails today not due to a lack of expertise, but simply because there is not enough time to optimally manage every campaign on a daily basis.

Why did you choose marketing as your initial target market?

There are two main reasons: we know the industry, and we know the pain points. Through our agency experience, we understand how much time is spent on manual optimizations and how frustrating it is when results still fluctuate despite high effort. At the same time, digital campaign management presents an interesting challenge: traditional machine learning approaches often lack sufficient data, while purely LLM-based solutions are not yet reliable enough at scale. Our approach combines both worlds, allowing us to demonstrate value quickly.

What is your most important goal for this year, and where do you see pi_optimal in five years?

This year, our primary goal is to prove that our approach works in practice. Specifically, this means turning pilot customers into long-term partners and delivering measurable results, less manual effort, more stable performance, and better campaign outcomes.
In five years, we aim to have demonstrated that our learning workflow approach works beyond campaign management. Our long-term goal is to enable better decision-making in other domains as well, where decisions today are still too often based on intuition and too little on data-driven learning.