In short
- Training your team in AI improves productivity and quality, and reduces the risks of unguided use: exposed confidential data, unchecked errors and generic copy.
- Training by department pays off most: sales, export, customer service, administration and management have different use cases.
- Claude and ChatGPT offer business plans that, according to their providers, do not train models on your content by default.
- Article 4 of the EU AI Act requires companies that use AI to take measures to support the AI literacy of their staff.
- Vycte's training courses can be subsidised through FUNDAE using each company's training credit.
Why train your team in artificial intelligence
In most SMEs, AI is already in use even if nobody decided it: a sales rep drafting emails with ChatGPT, an administrator summarising contracts or a technician translating a product sheet. The problem is not that it is used, but that everyone does it their own way, with no method and no rules on which data can be shared.
Good artificial intelligence training for companies turns that scattered use into an organised advantage. There are three reasons to do it now:
- Productivity: research, writing, translation and analysis tasks that used to take hours are done in minutes when the team knows how to give context, ask for the right format and iterate.
- Quality: a trained team gets less generic copy, spots errors and hallucinations, and uses AI as an expert first draft, not as the final answer.
- Controlled risk: pasting a confidential quote into a personal account, sending an unchecked translated email or trusting a made-up figure are common mistakes of unguided use. Training reduces them with simple rules.
For a company that exports or wants to win B2B clients, the impact shows mostly in preparation work: researching markets, qualifying accounts, adapting materials to each language and following up. That is exactly where small teams lose the most time.
What artificial intelligence training for companies covers, by department
A generic prompting course is forgotten within two weeks. What sticks is what each person applies to their work the next day. That is why we organise content by area, with examples from each company's sector.
| Department | What they learn to do with AI | Practical outcome |
|---|---|---|
| Sales and prospecting | Define the ideal customer, research accounts before reaching out, write email and LinkedIn sequences by segment, prepare meetings and summarise calls in the CRM | More qualified meetings with the same team and more relevant messages |
| Export | Research markets and competitors by country, adapt and translate product sheets and catalogues, prepare trade fairs (target visitor list, sales pitch, pre- and post-fair emails) and write quotes in the buyer's language | Materials ready for each market and trade fairs prepared with a full agenda |
| Customer service | Answer frequent questions from a reviewed knowledge base, classify tickets, draft replies in several languages and spot sales opportunities | Faster, more consistent replies without losing the brand's tone |
| Administration | Summarise contracts and long documents, extract data from invoices and delivery notes, prepare spreadsheets and reports, and draft supplier communications | Less repetitive work and fewer transcription errors |
| Management | Analyse scenarios and sales data, prepare presentations, stress-test decisions with arguments for and against, and set the company's AI usage policy | Better-documented decisions and AI adoption with clear rules |
The sales part connects directly with our B2B client acquisition with AI service and with the guide to AI lead generation. The export part ties in with the daily work of our outsourced export department and with preparing international trade fairs to win clients.
Claude and ChatGPT at work: practical differences
We use both in our training, because many companies already have one of them and because each stands out in different tasks. Rather than choosing «the best», it pays to know which plan to buy, how to set it up and what to use each one for.
| Aspect | Claude (Anthropic) | ChatGPT (OpenAI) |
|---|---|---|
| Business plans | Team and Enterprise | Business and Enterprise |
| Use of your data for training | Anthropic states that, by default, it does not use inputs or outputs from its commercial products to train models | OpenAI states that, by default, it does not train on inputs or outputs from ChatGPT Business, Enterprise and the API |
| Team management | Central billing, SSO and admin controls; more options in Enterprise | Workspace with central administration; more controls in Enterprise |
| Where it tends to perform best in our use | Careful writing, analysis of long documents and reasoning tasks with plenty of context | Very broad ecosystem, varied day-to-day tasks and teams already using it |
| Organising work | Projects with shared instructions and reference documents | Projects and custom GPTs for repetitive tasks |
Two caveats. First, features change every few months, so check the official pages of Anthropic and OpenAI before buying. Second, the data protection the providers describe applies to business plans, not to free or individual personal accounts, which have different terms. If your team uses personal accounts for work, that is the first thing to fix.
Good practice and risks: confidential data, GDPR and human review
Training is not only about getting the most out of the tool, but about knowing where the limits are. These are the rules we put in writing at every company we train:
- Company accounts, not personal ones: work is done on the corporate plan, with privacy settings reviewed by the company.
- Classify information: what can be entered freely (public copy, product sheets), what must be anonymised (client data, client-specific prices) and what is never entered (passwords, sensitive data, information under confidentiality).
- GDPR: if you enter personal data, you are processing it. Review with your data protection adviser the provider's data processing agreement, the legal basis and the information given to data subjects.
- Mandatory human review: no client email, quote, translation or report goes out without a person reading it. AI can invent facts with total confidence.
- Check the figures: tariffs, regulations, market data and prices are always verified at the official source.
- Transparency: if an AI assistant talks to clients, the client must know they are talking to an AI.
AI literacy under the EU AI Act (Article 4)
Regulation (EU) 2024/1689 on artificial intelligence sets out an AI literacy obligation in Article 4. According to the European Commission, it has applied since 2 February 2025, and the text in force as of October 2026, amended by the Digital Omnibus package, keeps the requirement for providers and deployers of AI systems to take measures to support the AI literacy of their staff.
In practice, a company that uses ChatGPT or Claude to write, translate or analyse is a «deployer». The text itself states that it does not require a specific level of literacy for each individual, and the Commission explains in its AI literacy questions and answers that there is no one-size-fits-all format or mandatory certificate, but staff should be aware of risks such as hallucinations. The prudent approach is to train the team and keep a record of that training. You can read the text on the Commission's AI Act Service Desk.
Vycte's sample AI training programme
Every programme is adapted to the company's sector and goals, but this is the indicative structure we start from. Modules can be delivered together or separately, depending on which areas take part.
| Module | Content | Indicative length |
|---|---|---|
| 1. Fundamentals and safe use | How models work, limits and hallucinations, data classification, usage policy and the AI Act | 2 hours |
| 2. Working method with Claude and ChatGPT | Instructions with context, templates, projects with reference documents and iteration | 3 hours |
| 3. AI for sales and prospecting | Ideal customer, account research, email and LinkedIn sequences, meeting preparation and CRM | 3 hours |
| 4. AI for export | Market selection, multilingual product sheets and catalogues, trade fair preparation and international quotes | 3 hours |
| 5. AI in management and automation | Documents, spreadsheets and reports; how to connect AI with Holded, Clientify and Make | 2 hours |
| 6. Workshop with your company's cases | Each participant solves real tasks and leaves with their own templates and a usage plan | 3 hours |
Training can be live online, in person or in-company, in Spanish, Catalan or English. It includes materials, instruction templates for each role and recordings of the sessions for review. At the end, the company has an AI usage policy and a record of the training delivered.
AI training subsidised through FUNDAE
Spanish companies that pay vocational training contributions have an annual training credit, applied as a rebate on their Social Security contributions. Vycte's courses, including AI training with Claude and ChatGPT, can be subsidised with that credit, provided FUNDAE's requirements and deadlines are met.
The key points: the training action must last at least two hours, the start must be notified in the FUNDAE application no later than two days before it begins, workers' legal representatives must be informed beforehand where they exist and, in companies with more than five employees, part of the cost is private co-financing. We explain it step by step in the guide to FUNDAE-subsidised training.
How to measure the impact of AI training
If you do not measure, training stays a good feeling. Before the course, pick three or four specific tasks per area and note how long they take today and what result you get. Then compare at 30, 60 and 90 days.
- Time per task: preparing a proposal, researching an account, translating a product sheet or closing a monthly report.
- Adoption: how many people use the company accounts each week and with which templates.
- Quality: errors caught in review, client responses and consistency of copy across languages.
- Business indicators: meetings booked, qualified leads, product sheets published per market and response time to clients.
The natural next step is usually to automate what the training has put in order: connecting the CRM, invoicing and AI with Make automations or tackling a digitalisation project with Holded, Clientify and Make. If the goal is to sell abroad, our digitalisation and export services work on the same foundation.
Frequently asked questions
Which is better for a company, a ChatGPT course or a Claude course?
It depends on the tool you already use and the tasks involved. What matters is learning a working method, classifying information and reviewing results, and that applies to both. In our training we work with Claude and ChatGPT, compare their business plans and help decide which one fits each department, without depending on a single tool.
Is it mandatory to train employees in artificial intelligence?
Article 4 of the EU AI Act, applicable since 2 February 2025, requires companies that use AI systems to take measures to support the AI literacy of their staff. It does not require a certificate or a specific level per person, but the team should understand uses and risks. Training staff and documenting it is the simplest way to comply.
Can an AI course be subsidised through FUNDAE?
Yes, if the requirements for company-scheduled training are met: a minimum length of two hours, notification of the start in the FUNDAE application no later than two days before, prior information to workers' legal representatives and application of the rebate before the financial year closes. The amount depends on each company's available training credit.
How long does AI training for companies take?
Our sample programme is around 16 hours split into six modules, but it is adapted: management may need a two-hour session on safe use and internal policy, while a sales team may need a longer workshop focused on prospecting. Splitting it into two- or three-hour sessions works best, so people can apply what they learn in between.
Is it safe to enter company data into ChatGPT or Claude?
It depends on the plan and the data. Anthropic and OpenAI state that, by default, they do not train their models on content from their business plans, while personal accounts have different terms. Even so, classify your information, anonymise client data, avoid entering confidential information unnecessarily and review the processing of personal data with your GDPR adviser.
