Most companies now have AI licenses and a handful of enthusiastic users. Everyone else sat through a course or clicked through an e-learning module, picked up a few prompt tips, and was back to the old way of working two weeks later.
That is rarely the employees' fault. Usually it is the format: the course was about AI in general, and their own work never appeared in it. So the first question is not which provider, it is which format.
AI training for employees: four formats compared
| Format | For whom | Duration | Outcome | Fits when | Doesn't fit when |
|---|---|---|---|---|---|
| Online course, e-learning (e.g. LinkedIn Learning, Coursera, Microsoft Learn, OpenAI Academy, Claude Academy) | Every employee, self-paced | A few hours, at your own pace | Basic concepts, first prompts, awareness of risks | many people need a baseline quickly and cheaply | a team should use the tool on its own data and processes |
| Open public course (e.g. General Assembly, university executive education) | Individuals from different companies | One hour to several weeks | Overview, exchange with peers, often a certificate of completion | individuals want to get started or go deeper | a whole team should change how it works together |
| In-house team training on real systems | One team or department | One day, with preparation beforehand | First workflows running on your own data, a list of the next ones | a tool is approved and the team has recurring work | no tool has been approved yet |
| AI Champions Program | 20 to 30 volunteers from several departments | Several weeks or longer | Internal people who build; use cases in production | AI should stick for good and a sponsor is in place | all you need is a one-off introduction |
Course platforms also sell company licenses. The real difference from in-house training is whether your tools and data are connected before anyone practices.
When an online course or a public course is the better choice
Specialty Tokens wrote this page, and we only offer the bottom two formats. For some goals, the first two are still the better choice:
- Basics for the whole workforce: e-learning scales and costs little. Microsoft Learn has a Copilot learning path with exercises for sales, finance, HR, operations and other roles; Anthropic's Claude Academy and OpenAI Academy offer free courses; Google AI Essentials on Coursera takes under ten hours.
- Individuals who want orientation: a public course costs less than an in-house session and adds perspective from other companies.
- No tool approved yet: make that decision first, for example in a session for executives, then train.
We compare named providers, with strengths and limits, in corporate AI training providers compared.
What corporate AI training should include
- The approved tool: training happens in the tool you pay for, with your accounts and data rules.
- Your own tasks: people practice on the weekly report and the shared inbox, not on made-up examples.
- Split by audience: leadership, business teams and IT have different questions. The European Commission also recommends accounting for people's knowledge, experience and context of use.
- Data rules and checking: what may go into the tool, and how do you check a result before it goes out?
- Durable features: projects, connectors and reusable instructions, not features on their way out. OpenAI retires custom GPTs on December 11, 2026.
- A result after the session: what is still running on Monday?
- A record: content, date and participants. The Commission does not require a certificate.
How in-house training on your systems works
AI training from Specialty Tokens starts before the first session. We agree which tools are approved and connect them through the Model Context Protocol (MCP) to your systems: ERP, CRM, accounting, ticketing, file storage. Then you pick the format:
- Tool training: ChatGPT, Claude, Microsoft Copilot, Gemini, or Langdock, on your reports, inboxes and analyses.
- Technical depth: Power Automate and Copilot Studio for teams that want to automate themselves; Claude Code and Codex for engineers.
- Literacy and records: AI literacy training with data rules and documented content.
- Leadership: AI training for executives on value, cost, risk and governance.
- A standing capability: the AI Champions Program.
Sessions are taught by the people who build production AI systems for clients, led by directors Nik Redl and Thomas Schlossmacher.
What changes afterwards
At TimeTac, the marketing and growth team, with no engineers on it, went live with ten automations after an AI enablement workshop and an internal AI platform: 117 runs in three weeks, 54 of them unattended. At Speedinvest, the pan-European VC, enablement ran team by team with Affinity CRM and the fund's ERP connected; the dealflow inbox automation alone saves an estimated 10 to 20 hours a week. At Nyra Health, a four-hour hands-on workshop gave a skeptical, experienced engineering team a fresh look at Cursor and Claude Code. With the Austrian Federal Economic Chamber (WKO), we run a series of in-person executive AI sessions as a partner.
AI literacy under Article 4, also for US and UK companies
Since February 2, 2025, Article 4 of the EU AI Act has required companies that use AI to take measures for their staff's AI literacy. Per the Commission, that includes companies outside the EU when the AI is used in the EU or affects people there. The Digital Omnibus softened the wording in July 2026 but kept the obligation. This is not legal advice.
Everything built in a training is yours: no per-seat fees, no dependency on us. If you are unsure which format fits, talk to us.
