Date: 14.10.2025 (2025-10-14)
Alexander Anserud (14.10.2025) • Klicka founder • Why did they stop it?
Notes:
• They are not building Klicka anymore
• Why: they spoke to a lot of RPA developers; a lot of people did the same thing; tried to find a niche; then they focussed on legacy SW; horizontal business idea; technology was not there yet when they did it; they had several paying customers; insurance, banks etc.;
• LLMs are still bad atm.
• Then they realised that people hyperfocus on one vertical
• e.g. focus on form filling on websites and be really good with that
• Check out https://www.askui.com/
• They also built their own models - but they were not good enough
• Check whitepapers
• Check out Strawberry -> new browser for AI
• Check out Comet -> how do we differentiate
• They explored a lot of different methods - not only Computer Vision
• A lot of support was required
• They were committed and finetune models
• Everyone is pivoting to foundation models
• Main challenges:
• Latency of models is really high (seems highest problem)
• Idea: automate stuff in the background (Strawberry is doing that)
• Accuracy: they tried a lot of things but they were not able to fix it
• Privacy: spyware on the computer feels wrong
• His cofounder was one of the first people in Europe that built an open source model
• They did a lot of experiments with users -> users need to click and screenrecording and talk
• Check Autotab
• They could not get the tech to work -> investors wanted to see it
• Citizen developer -> they wanted to build for them
• Check Compute Use
• Overall: horizontal is not possible -> hyper verticalise
• They discussed hyper verticalise within SAP
• RAG architecture with a lot of context to the system -> planning of agent should be based on RAG (e.g. for a swedish legacy system)
• They wanted to do tenant computer -> low latency -> build small macros easily
• They spoke with hundreds of people in many companies -> people do not want to automate due to being replaced
• They were focussing on on device to differentiate from LLM manufacturers -> on premise -> focus on banks, insurances, hospitals -> even hosting on MSFT Azure was sometimes not possible
• His new company?
• 500 users on the platform -> was able to sell it
• Websearch
• Innovation platform: quickly discover what people want -> then build a product, researcher or PMs
• Fetch research papers in domains, McKinsey reports -> build very well researched data base
• Check out Bluebook (YC company)