Business Process Automation
We set up processes where you can see who holds each request and where it is stuck, before anyone complains.
Benefits you can expect
You know in advance how long a process takes and who owns each step. Work carries on when people are away or leave, because the steps live in the system. Managers see each team’s workload and move people in time. New staff learn faster because the system prompts the next action. Growing volumes need fewer new hires. We map processes, set up the platform, link it to your systems and support it after launch.
What’s included
- Process mapping and redesign
- Routes, rules and deadlines
- Bots for systems without APIs
- AI steps under human control
Problems we solve
We learn about delays from complaints
Each step has a deadline, and the system passes an overdue request to a manager.
Each approval goes by separate email
Approvals follow one route in the system, and independent ones run in parallel.
People sort routine requests by hand
Rules handle routine requests, and people take only the exceptions.
Requests land in the wrong department
The system assigns each request by topic and type, so no one has to forward it.
Everyone decides the same case differently
Decision criteria are set in the system, so the same case always gets the same answer.
Homemade apps only their authors understand
Apps are built on one platform under an architect’s rules, and each one is documented.
When you need it, and when you don’t
You need it
- the same process runs more than 100 times a month
- a regulator requires a log of every approval
- the IT team can’t keep up with change requests
You don’t
- a process runs a few times a month: a request template and a shared spreadsheet are enough
- no one makes decisions, and data only moves between systems: systems integration is enough
Our approach
- 01
We cut needless steps before automating
A process moved into a system unchanged stays the same mess, only faster. So for each step we first ask whether it is needed.
- 02
Bots only where there is no API
A bot clicks through an old system in place of a person and breaks with each update. So we first look for a direct connection via API, and every bot gets an owner from day one.
- 03
AI prepares, people decide
AI reads a request and checks it against documents; an employee approves or rejects it. If AI is wrong or unavailable, the request goes to a person’s queue and is not lost.
Project stages
- 2–3 weeks
Discovery
We talk to process owners and find exceptions and workarounds.
- 2–3 weeks
Architecture
Platform and licence choice, process maps and decision rules.
- 2–3 months
Pilot
One process, most often requests or contract approval.
- 3–4 months
AI layer
If needed: AI is tested on your exceptions before and after launch.
- ongoing
Improvement
We find where the process slows down and test each change before release.
Platforms & standards
Scriptum · UnityBase · Camunda 8 · Mendix · OutSystems
UiPath · Microsoft Power Automate · Automation Anywhere · SS&C Blue Prism
OpenAI · Anthropic Claude · LangGraph · Microsoft Copilot Studio · MCP
BPMN 2.0 · DMN · ISO/IEC 27001 · ISO/IEC 42001 · OWASP Top 10 for LLM · EU AI Act
Questions
How do we choose between Scriptum, Camunda and UiPath?
We use Scriptum for applications with forms and processes, Camunda 8 for processes developers run in code, UiPath for bots. Camunda 8 on your own servers needs a commercial licence in production.
How do we measure the return on automation?
We track completion time, errors per thousand requests and the cost of one request, licences included. We judge the return no earlier than six months after launch.
Does AI‑driven automation fall under the EU AI Act?
It depends on the purpose: triaging and routing requests is usually not high‑risk. Candidate screening or credit scoring is, and those obligations apply from 2 December 2027.
Further reading
Let’s map your process
Thirty minutes with a process architect to see if automation is worth it and where to start.