AI research and pilot proposal for a municipality
A research and advisory assignment for a municipal organisation. The question: which AI and automation applications genuinely save time and reduce errors in the technical preparation of civil engineering projects?
The answer is now on the table: a substantiated shortlist and a concrete pilot proposal, ready for decision-making.
The client and project details remain confidential here.
- Client
- Municipal organisation
- Field
- Civil engineering · project preparation
- Role
- Independent researcher and advisor
- Duration
- Roughly three months
- Tooling
- Revit · Dynamo · Python · Copilot
- Deliverable
- Shortlist + concrete pilot proposal
- Status
- Pilot proposal awaiting decision
The challenge
The team prepares construction projects in the same steps every time: surveying, 3D modelling, calculating and drawing. Much of that work is done by hand and repeated for every sub-project. Information from earlier work is scattered and rarely reused, so every round costs time again and errors creep into drawings.
What I did
- Interviews with every role involved, from modellers and structural engineers to technical management.
- Analysed the preparation process and planning: where does the time go, where do errors appear and what is on the critical path?
- Worked out five solution directions and scored them on value, feasibility, the team's own influence, complexity, risk and investment.
- Delivered a substantiated shortlist and a concrete pilot proposal, including safeguards for secure data handling, risks and predefined success criteria with a go/no-go.
The proposed pilot
Recurring structures are currently modelled and drawn by hand for every sub-project. The pilot flips that around: a script builds the 3D model automatically from a parameter table (Revit, Dynamo and Python), including the standard sections and views. A second script automatically checks the drawings for slips and inconsistencies.
When an assumption changes, only the table is updated and the model and drawings follow. AI helps write the scripts and can broaden the checks later; the scripts themselves run locally, without project data going to an AI service.
The approach in practice
The research followed the same five-step approach described on this site. This is what it looks like in a real assignment:
Explore
Interviews, process analysis and data verification: which data exists, where does it live and is it usable?
Idea generation & selection
Five directions worked out and weighed in a matrix, resulting in a substantiated shortlist.
Getting AI-ready
Safeguards defined for secure, local processing of project data and the development environment required.
Proof of concept
A concrete pilot proposal: automated modelling with automated drawing checks, tested on existing work first.
Embedding
Work instructions, version control and handover to the team as a fixed part of the proposal.
The honest outcome
The biggest gain here is not an AI tool, but automation. AI speeds up building it. Independent advice delivers the best solution, not the trendiest.
Status
The research is complete and the pilot proposal is awaiting the client's decision. This page will be updated once the pilot has been carried out and evaluated.