What an SME can build with AI in one day

What can a small or medium-sized company realistically achieve in a one-day AI workshop? In the case of Dockyard Ventures, the answer was four tested AI skills and two clickable prototypes: a new website and a mobile time-tracking application.

That sounds like a story about speed. The more important result was different. The team translated its own process knowledge, quality standards and everyday friction into tools it could use, test and improve.

The workshop in Bolzano on 27 August 2026 also marked the conclusion of our joint MMIND Erasmus+ project. By the end of the day, the enthusiasm in the room was unmistakable. It did not come from a spectacular AI demonstration. It came from seeing familiar problems turned into something tangible that belonged to the team.

What did the team build in one day?

• four reusable AI skills for venture initiation, presentations, knowledge management and content
• a clickable prototype for a new company website
• a clickable prototype for mobile time tracking
• a clear owner and next real-world use for each result

The skills were built with the team and tested on practical tasks. The two applications were then developed in vibe-coding sessions with Lovable. They are workshop prototypes, not approved production systems.

Why the workshop did not begin with an AI tool

Many AI workshops begin with features. A chatbot summarises a document, generates an image or writes a post. It can be impressive. On Monday, however, the team often returns to exactly the same work.

We reversed that sequence. The first question was not, “What can Claude do?” or “What can we build with Lovable?”

It was: Which recurring work costs you time today, and what does a good result look like?

The answers were specific. Workshop presentations often start with notes, sticky notes or a voice briefing. Knowledge is spread across Notion, SharePoint, personal collections, screenshots and voice memos. Published content must sound like Dockyard, use real observations and protect confidential client details.

The applications also had a business starting point. The existing website is difficult to update. Monthly time reporting involves late entries, checks, exports and a separate spreadsheet.

These are not abstract “AI use cases”. They are recognised points of friction in the business.

Which four AI skills did Dockyard build?

An AI skill, in this context, is not a very long prompt. It is a small reusable work system containing the trigger, required material, sequence, quality rules, boundaries and expected output.

Venture initiation

This skill captures Dockyard's method for moving from a search field through target customer, value proposition and business model to a prototype and validation plan. It must distinguish clearly between decisions made by the team and suggestions generated by AI.

Presentations

Notes, voice briefings or workshop materials become a first PowerPoint draft in the correct template. The skill carries a quality standard as well as a sequence: a clear story, correct design, no duplicate slides or technical overflow, and language that fits the person presenting.

Knowledge management

Loose notes, links, screenshots and voice memos become standardised knowledge cards with source, date, tags and approval status. Dockyard's own gold standard takes priority. External research must not quietly overwrite the firm's method.

Content

A real observation, approved photograph or event becomes a publication-ready draft. The skill checks tone, facts, confidentiality and relevance. Nothing is published automatically.

When is a skill enough, and when do you need an app?

A skill is suitable when one person works with AI on files and follows a recurring method. It can prepare work, apply quality rules and produce an output that is easy to review.

An application becomes useful when several people need to see the same data, roles and permissions matter, or the process must also work outside the AI tool.

This distinction avoids two common mistakes: turning every problem into a software project, or forcing every team process into a chat window.

Which two prototypes were developed?

The website prototype makes the offer understandable on mobile devices and should allow the team to update content without an agency. It already shows a basic structure covering value, services, method, team and contact. Positioning, final copy, privacy, redirects and analytics still need to be reviewed before publication.

The second prototype supports mobile time tracking. Time is assigned to a client, project and task, marked as billable or non-billable, and prepared for monthly review. The complex variable-compensation calculation was deliberately excluded. Its rules must first be verified with anonymised data and test cases.

Vibe coding was therefore not the beginning of the thinking. It was the translation of a well-defined problem into a visible interface the team could discuss.

How can another SME repeat this workshop?

• Bring a real workflow and the material used to complete it.
• Identify where the work stalls and define what a good result looks like.
• Only then use AI to structure the process, rules and open questions.
• Build a small skill or prototype and test it with a real or anonymised case.
• Give every result an owner, a storage location and a first practical use.

The visible workshop day requires preparation. Without real documents, working access and suitable test cases, too much of the day would be lost to searching and setup.

What is not finished after one day?

A clickable prototype is not a production application. Roles, authentication, privacy, real data, integrations and testing belong in controlled follow-up steps.

A skill only proves its value in the next real assignment. It must produce consistently useful preparation, recognise its limits and make human review easier. Professional judgement remains with the team.

The purpose of the day is not to promise six finished products. It is to replace abstract AI discussion with working artefacts the team can examine and improve.

Frequently asked questions about the AI workshop

What is an AI skill?

An AI skill packages a recurring workflow with context, sources, rules, boundaries and an output format. The team does not have to explain the task from scratch every time.

What did vibe coding mean in this workshop?

Here, vibe coding meant creating a clickable software prototype from natural-language requirements. The processes defined by the team were translated into visible interfaces with Lovable.

Can every SME build an app in one day?

A well-prepared team can develop and test a tightly scoped prototype in one day. Turning it into a production application depends on data, security, integrations, business rules and further testing.

What is a sensible first step?

Do not begin with a list of AI tools. Choose one recurring workflow that costs time and describe what a good result looks like. That makes it possible to decide whether the right answer is a skill, an app or a simpler process change.

Would you like to try this with your team?

The AI workshop can be adapted to other companies and teams. The outputs will differ because each company works with its own processes, documents and quality rules.

Tell me which recurring workflow costs your team the most time today. We can then explore what your team could usefully build and test in one day.