Date: 28.10.2025 (2025-10-28)
Julian Hürlimann (28.10.2025) • https://www.linkedin.com/in/julian-h%C3%BCrlimann-61a4b813b/ • Pitch the UI as API slides • Works as Netlight • Netlight is an international digital consulting firm, helping leading companies to succeed in the digital landscape, from advice to implementation.
His Experience & Current Practices
• What is their focus as a company? What is a typical customer problem?
• Typical projects:
• End to end data, tech, product
• Technology and industry agnostic
• They become a role in the team of the customer -> they are part of the team -> joint effort
• Longer time duration of their projects: 9 to 12 months
• Legacy systems:
• Legacy systems are everywhere -> SAP, internal custom SW -> every customer has that
• Challenge is in the implementation -> solutions:
• The closer you get to the data the better
• API best, database is second, deterministic part, RPA is one of the last solutions - when there is no other solution
• RPA is high maintenance
• Deterministic script: MoveIt -> regel basiert
• How do we execute it from a process Sicht
• He uses console in browser and lets a script run -> for non super repetitive or internal API call to get all things from a list -> for one time things
• UI vs. DOM vs. network requests
• Network requests would be best - less anfällig than DOM
• He can do in Chrome -> Dev Tools -> Network tab - > download all requests
• Download as HHR file -> give to LLM to create Javascript snippet
• If API changes -> has to be updated
• Challenge:
• How can one do it automated
• Network request to intercept
• How can we build an interceptor
• Solution to get data from legacy software
• Try to not introduce new stuff - most of the times it does not make sense to introduce sth new
• Most of the time a few integrations are necessary
• Solutions: MoveIt or other tools are used
• Focussed on repetitive processes
• Define data sources, trigger and then define what should happen to that data (folder, etc.) -> file transfer automation solution
• No self healing etc.
• Legacy systems: does he deal with them (accounting tools, banking interfaces, HRIS, ERP)? How painful are integrations?
• See above.
• How much focus on automation and agents already?
• Agents:
• They are quite ahead in comparison
• Many companies struggle with basics -> platform for MCP servers etc.
• Many companies have POC but no real production agents yet
• Successful agents?
• Personal automation
• ChatGPT that has the exactly the same access like me
• A lot can be done via prompting
• Quality measurement:
• Anthropic has enforced sources
• Show user where information was queried and how etc. -> show original data -> put as few responsibility to the LLM - as much as possible deterministic
• Customer Support
• Important with selection of use case -> hallucinations - either it does not matter or human in the loop; simple to review is important - otherwise people miss it
• Power Automate?
• Emil is really good https://www.linkedin.com/in/emiloldenburg/
• Do they work in finance / accounting area?
•
• Hypothesis: general tools like n8n are too complex to set up for many processes -> important to specialise?
•
• Day-to-day reality: What are the most admin-heavy processes in his finance role (invoices, reconciliations, reporting, payroll, compliance, audits)?
•
• Is he automating today? If yes, which tools (Excel macros, Zapier, RPA, ERP features)?
• If yes: who does it? Which tools? What works? What does not work?
•
• If no, why not (cost, complexity, lack of time, no clear ROI)?
•
• More processes to automate:
•
• Other automations from other people / departments:
•
• Successes & frustrations: if he had a “magic wand,” what would he automate first?
•
• Ask if he can introduce you to 3 peers in similar roles (finance leads in other startups).
•
• Other:
•