In brief
To introduce artificial intelligence into an SME it makes sense to start from repetitive processes where information arrives in writing (orders, documents, emails, requests) and to connect the AI to the software already in use, with a person checking uncertain cases. Liquid does this with a 30-day Sprint at €3,000 or with individual projects from €1,000, using models that are not trained on the company's data.
Why ChatGPT subscriptions aren't enough
Many companies have already tried AI: a few ChatGPT or Copilot subscriptions, a chatbot on the website, a course. Everyone uses it their own way and the way the company works stays the same. The reason is simple: the AI isn't connected to the data and software where the work happens.
The value comes when AI enters a specific process: it reads the order and writes it into the management system, reads the delivery note and records it, drafts the quote from the customer's request.
Where AI works today in an SME
- Reading orders received on WhatsApp, email or forms and entering them in the management system
- Extracting data from delivery notes, invoices, order confirmations and supplier PDFs
- Sorting incoming requests and giving a first reply
- Searching procedures, price lists and internal documents
- Drafts of quotes, reports and messages for review
- Sales analysis to spot inactive customers and opportunities
Where it isn't needed
Where data is already in order, a classic integration between two programs is enough, cheaper and more predictable. Where a decision needs judgement and experience, AI can prepare the information but the choice stays with a person. Saying so at the start avoids expensive and disappointing projects.
How we introduce it
- Analysis: meetings with the owner and managers, a map of repetitive activities with hours and costs
- Priorities: two or three interventions ranked by impact, feasibility and risk
- Project: development on your data, connected to the software in use, with human control over uncertain cases
- Adoption: training the people involved, rules for data use, measuring results
Data, security and the AI Act
We use the models through APIs under terms that exclude training on the company's data, with accounts registered to you. We define which data is processed and by whom, and we prepare the documentation needed for GDPR and the AI Act, including training for the staff who use the systems.
Team training
We don't run generic courses. We train people on the tools introduced in their own company: how to use them, how to check them, what to do when the system flags a doubt.
What it costs
| Project | Price | Includes |
|---|---|---|
| AI Automation Sprint | €3,000 | Per month: process mapping, priorities and new automations in production. |
| Single automation | from €1,000 | A process already identified, connected to the software in use. |
| Custom software | from €1,000 | A new tool when what you need doesn't exist: CRM, management system, portal. |
Indicative prices. The final figure is in the written proposal.