Achieving a 70% cost reduction for KF.no by using AI to migrate their system to a new Low Code platform.
Aventude created an agentic-human loop; leveraging AI powered automation and digitization to turn a prohibitively complex digital transformation, into one that could be managed in a matter of days.
Organizations are often ready to invest heavily in digital transformation and modernization to create advanced, user-friendly platforms; but transitioning the existing forms, content and processes to the new platform can often create so severe a bottleneck that it calls into question the financial viability of the transition.
KF.no (Kommuneforlaget) is a leading Norwegian technology and publishing provider that delivers digital solutions, tools, and platforms to support municipalities across Norway. Serving over 300 municipalities, KF.no plays a critical role in enabling local governments to streamline their processes, manage citizen interactions, and conduct large-scale surveys and consultations. With a focus on user-friendly digital services, compliance, and efficiency, KF.no helps public sector organizations modernize and improve the way they deliver value to citizens.
This was the situation facing KF.no. A technology provider that runs the platform over 300 Norwegian municipalities use in conducting citizen surveys. These surveys are complex: varied in format, multi-language, multi-page and featuring complex branching conditional criteria.
Transitioning to a Low-Code Platform
KF.no adopted a cutting-edge Low-Code platform, Compose Glow, for their new system, which allows for the creation of highly complex interactive surveys easily using a sleek drag and drop interface, enables intelligent workflows and features built in data analysis tools. It also provides a better experience for the audience taking the survey, with a more modern UX and better usability features.
The bottle neck however, is moving the hundreds of highly complex existing surveys created in PDF, XML and HTML formats. Manually migrating the old surveys to new platform which has its own proprietary format is a task so complex and time consuming, it would have taken over half a year and perhaps proved to be financially infeasible.
Accelerating The Transition Using an Agentic-Human Loop
As a solution, Aventude proposed an Agentic-Human Loop, creating a double cycle where the AI migrated the Survey in its base form, its work was checked by KF's consultants, and then a Low Code, Compose Glow version was created by the Gen AI as well, to be checked by a testing automation AI agent before proceeding to a final human check.
Here is a detailed breakdown of how the five-step agent-human process was implemented.
- Extraction Agent – This agent is given the XML and PDFs to extract the forms definitions, validation parameters, translative parameters, etc. The validations have if-else conditions based on the survey answer. This is translated to the show-hide logic and other JavaScript logics in the new platform.
- Error Correction and Tool Calling – When the extract agent cannot scan or interpret the logic it makes a tool calling which alerts a human to the exact section where the agent failed.
- New Platform Agent – This is agent takes the curated definitions and transforms them into new platform definitions.
- An Automation Agent powered by Browser Use – runs all the combinations of the answers, checking all the possibilities of the workflow and delivers the final verdict on the migration.
- At the very last step an experienced human consultant will validate the overall migration process, checking the language and different flows.
Reaping the Commercial Benefits
This method cut down the migration time and effort by more than 70% making the transition financially easy on KF.no. Coupled with the quicker adoption of the new platform; born from reducing the time it takes to deploy and use it to its fullest extent, brings added commercial benefits to the client; enabling them to reap the rewards of their investment sooner, and fast track their digital transformation process.
Want to learn more about the products we have been creating for our clients?
Next Case