Addressing the challenge of data retrieval
In today's data-driven world, organisations face significant inefficiencies and security risks associated with manual data retrieval. AgentFormers addresses these challenges by providing a seamless, scalable solution that enables users to securely automate routine tasks, freeing up organisations to focus on more valuable work and improving the service experience for all users while saving time.
Research and development: From concept to innovation
AgentFormers' breakthrough lies in combining deterministic methods, such as state machines, with creative AI prompting. This hybrid model minimises language model hallucinations while retaining effective text processing capabilities. This innovation primarily lies in the field of computer science and is supported by research in human-computer interaction (HCI) and socially interactive agents (SIA).
The uniqueness stems from its hybrid AI model, which addresses the critical issue of AI hallucinations. It combines Reliable State Machines for process control with creative LLMs for specific language tasks, ensuring trustworthy results. The core technology uses retrieval-augmented generation (RAG) to ground all answers in one’s own secure data rather than the open internet.
Unlike other tools, this no-code platform gives administrators full control to build custom agents, set granular security permissions and deploy them via flexible interfaces ranging from simple text chat to interactive 3D avatars.
The core research was conducted at the University of Helsinki's Department of Computer Science. Applications were then tested in collaboration with other university units, including HR and legal department.
Market opportunities and applications: SaaS for knowledge-intensive sectors
This technology is scalable and has potential for widespread adoption. For example, it can automate internal reporting and policy lookups for departments such as HR and Legal, as well as create intelligent customer service avatars for sectors such as pharmacies and libraries. These avatars can provide instant, accurate answers from internal knowledge bases. AgentFormers solution is intended to be a B2B SaaS platform.
Its target markets include large, knowledge-intensive organisations such as universities and healthcare providers, which are represented by the initial client base. The no-code tool enables these organisations to easily build and deploy their own secure AI agents, significantly improving internal efficiency and the overall customer experience. With the ability to customise agents for specific requirements, this innovation is well-placed to transform operations across various sectors.
Impact and benefits: democratisation of AI
AgentFormers' broader impact is the democratisation of AI for organisations lacking dedicated AI teams. The no-code platform enables companies to develop robust custom solutions, improve service quality and boost productivity by automating knowledge work. It also improves service quality by offering instant 24/7 support while ensuring data security by processing information in-house. Proof-of-concept projects are underway with key partners, including units from the University of Helsinki and industry collaborators.
Future and path to commercialisation
The immediate next step involves completing production pilots with university and industrial partners, with key findings expected in autumn 2025. This validation stage is critical, as the technology is actively tested with real users and data. The aim is to leverage the insights gained from these pilots, alongside positive feedback, to explore opportunities for further development and growth, with one possibility being the establishing of a spinout company. Securing seed funding from investors will be essential for scaling operations and launching the B2B SaaS platform. This project has secured funding from Business Finland for commercialisation preparation, thereby facilitating a successful transition to market.
Summary
AgentFormers is a no-code platform that automates routine tasks securely, transforming data systems. It eliminates the inefficiencies of manual data retrieval, allowing users to concentrate on more valuable work. Developed at the University of Helsinki, AgentFormers combines deterministic methods with creative AI prompting to minimise AI hallucinations. Designed for knowledge-intensive sectors, this Software as a Service (SaaS) solution enhances efficiency and customer experience. Through ongoing proof-of-concept projects and commercialisation, AgentFormers aims to make AI more accessible to organisations without dedicated resources.
Join the collaboration
We are seeking potential seed-stage investors with expertise in B2B SaaS or enterprise AI. We are also looking for new strategic industry partners, especially in healthcare and professional services, to expand our pilot programs and accelerate the go-to-market strategy of our technology.
Contact us
Sasu Tarkoma
Project Lead
Jussi Wright
Commercial Champion