
Lovable sandbox for public administration: how ~80 officers learned to prototype
In spring 2026, the State Administration School (VAS) invited me to organise a training cycle for public administration officers on AI prototyping with Lovable. This is the story of how, in two 4-hour sessions, civil servants turned their everyday problems into their first working digital prototypes.

Context: why this project happened
Latvia's public administration — state institutions, agencies, municipalities — employs thousands of people who deal with document flows, manual processes and complex bureaucracy every day. They know their work environment better than any outside consultant. But they lack the tools to translate that knowledge into concrete digital solutions.
In early 2026, Elvis Dibaņins from the State Administration School (VAS) reached out. They were preparing the EU Recovery Fund project "Public Administration Digital Academy" and looking for a practical format where staff could themselves try out AI tools.
The goal was clear: not to theorise about "digital transformation", but to give a real experimental experience. An environment where a civil servant is allowed to make mistakes, try things out, prototype — without the risk of being told off for "doing it the wrong way".
That's how this project's working name was born — Lovable sandbox.
Client: VAS and the public administration audience
Client: the State Administration School, which has been developing public sector professionals in Latvia since 2010.
Audience: around 80 public administration officers from various institutions — experts, project and programme managers, department heads, IT and data specialists, administrative and communications staff, customer-facing officers.
Digital level of the audience: very mixed. Some participants work with Excel and Word every day. Others run IT projects. This range was one of the challenges.
Context: EU Recovery Fund project Nr.2.3.2.2.i.0/1/23/I/VARAM/001 — i.e. a publicly funded training cycle with a requirement for tangible, measurable outcomes.
Challenge: two simultaneous problems
When training the public sector, two problems usually go together:
First — differences in digital level. In one group, a person who has used a single Excel formula in their career sits next to an IT manager with 15 years of experience. A classic "beginner course" goes too slowly for the first and too fast for the second.
Second — security concerns. The public sector works with sensitive data. Many staff are cautious with AI tools because they don't know where the data goes and how it is used. Without solving this, no serious prototype gets built.
VAS' brief to me was to find a format that solves both problems at once.
Outcomes: what participants got
Outcomes
Two 4-hour sessions (15.04 and 28.04) with around 80 participants.
Every participant finished by formulating a concrete first step to take back to their institution. But the numbers aren't the story. The story is what participants themselves said.
What participants said during the sessions
Comments from the Zoom chat
For the first time I see that I myself can build a working thing, not just describe it in a spec.
Lovable surprised me — I thought only developers could use it.
Now I understand what 'prototyping' means — until now it was just a buzzword to me.
These days I built a simple staff training form myself — something we should have had a year ago.
The main lesson — you're allowed to make mistakes. In our bureaucracy that isn't obvious by default.
The first session gave the idea. The second — the courage to finish it.
I'm genuinely grateful for this kind of feedback — I heard it more than once across both sessions.
Client testimonial
After the cycle I asked Elvis Dibaņins — the project organiser on the VAS side — to share his view of the course and the sandbox format:
Why it worked
After these two sessions I've identified three things that made this project distinct from classic IT training:
First — built-in freedom in time. We didn't pressure participants that they had to finish a specific task at a specific moment. They could experiment, return to basics, try their idea several times. That gives the kind of psychological safety that some workplaces simply don't have.
Second — using a real AI tool, not demoing one. Many AI trainings consist of slides and screenshots. Here participants worked in Lovable themselves from minute one. The AI tool isn't an object of observation — it's a working tool in their hands.
Third — a complete learning loop. The first session set the foundations. Between sessions participants worked on their own. The second session was reflection and going deeper — what worked, what didn't, what to do next. This loop is more effective than classic one-off lectures.
Next step
If you're looking for a training cycle for your team or company — about AI tools, prototyping, or general digital-thinking development — get in touch.
Free 30-min consult — to discuss whether your situation is a fit. Or write directly to info@fullmarketing.me with a short brief — audience, format, goals.
Egils Boitmanis is a business, AI and strategic partner with 20+ years of experience in marketing and business development. Former co-founder of Infinitum 8 (the largest digital agency in the Baltics). Google Certified Trainer (Zurich, Dublin, Kraków, Prague, Vilnius). Lovable Vibe Coding instructor. Founder of AI Circle at the University of Latvia. All numbers in this article and Elvis Dibaņins' testimonial are published with VAS' permission.
Frequently asked questions
Want a similar result?
Start with a free 30-min business diagnostic — we'll see if we're a fit.
Related articles
Working with MiniAkselerators: AI lectures and a masterclass
In spring 2026 I worked with Mikus Losāns and MiniAkselerators in two formats: AI lectures inside the University of Latvia digital marketing course, and a masterclass on how to build a profitable product with AI. This is the story of the collaboration, the audience and the potential.
Ivars's story: 8 months, +33% measurable results
An anonymized client story: how the owner of a 22-person manufacturing company rebuilt his strategy in eight months, implemented AI and recovered 12 hours per week.
