What this is about
A few years ago the question in most SMEs was: "Should we use AI at all?" In 2026 the more useful question is: "Where does AI create real value for us — and are our people skilled enough to use it well?"
Availability of the technology is barely the limiting factor any more. Generative AI sits inside office suites, search engines, CRM systems, development environments, design tools and specialised business applications. At the same time, AI agents, retrieval-augmented generation (RAG), vibe coding and automated workflows open up possibilities that go far beyond writing a text with ChatGPT.
That is exactly where the new gap opens: the tools improve fast. Skills, processes and governance do not automatically keep pace.
The OECD names missing skills as one of the central barriers to AI adoption: more than half of the SMEs that have not yet adopted generative AI cite a lack of capabilities as an obstacle. At the same time, only a small share of the workforce needs highly specialised AI development knowledge. For the majority what counts is digital competence, data understanding, critical thinking, problem solving, creativity and the ability to apply AI correctly in their own work context (OECD, 2026a, 2026b).
For management the conclusion is this: not everyone has to build RAG systems or program an agent. But teams do have to understand which tool suits which task, how good results come about, when outputs need to be checked, which data must not go into arbitrary systems, and when a chatbot turns into an agentic process with additional risk. AI competence is therefore not a single skill but a tiered capability model.
This article shows which capabilities are practically relevant for Swiss SMEs, how they can be structured across six levels, and how leaders can derive a realistic learning and implementation agenda from that.
Why AI skills are becoming a competitive factor for SMEs
The OECD describes skills as one of the decisive factors in whether companies actually extract productivity gains from AI. More than half of the SMEs without generative AI name missing capabilities as a barrier. At the same time, companies report that AI can partly compensate for existing skill and staffing gaps. So AI creates new competence requirements and helps reduce existing bottlenecks at once (OECD, 2026a).
For SMEs this matters more than for large corporates. They typically have fewer specialised roles. Marketing, sales, operations, HR or the management team often have to assess and deploy new technology on top of day-to-day business. A separate "AI department" is rarely the realistic answer.
The OECD also shows that most employees do not need highly complex AI specialist knowledge. Fewer than 1 % of workers require advanced AI-specific skills; what is broadly relevant are digital skills, data analysis and interpretation, plus complementary capabilities such as problem solving, creativity, innovation and management (OECD, 2026b).
The SME takeaway: The strategic question is not how many employees become "AI experts". What matters is whether the right people have the right depth of AI competence for their role.
Being able to use AI is more than good prompting
Prompting stays important. But anyone who reduces AI competence to "writing good prompts" is measuring one slice only. In practice it takes at least six competence fields.
1. AI literacy — understanding what AI can and cannot do
AI literacy starts with a realistic basic understanding: What is generative AI? How do answers come about? Why can a model phrase something convincingly and still be wrong? What is the difference between a model, a chat interface, an agent and a tool? These fundamentals matter because wrong mental models lead to wrong decisions.
The European Commission lists as minimum questions for an AI literacy initiative, among others: What is AI, how does it work, which AI is used in the organisation, and what opportunities and risks exist? Role, prior technical knowledge, experience and usage context should be taken into account as well (European Commission, 2026).
2. Tool choice — not every task belongs in the same chat
A common productivity mistake is solving every problem with whatever tool happens to be familiar. Source-grounded learning, current web research, image generation, video, voice cloning, coding and process automation all place different demands. AI competence therefore also shows in understanding the use case first and choosing the tool second.
For an SME this capability saves cost and time immediately: the most powerful or best-known tool is not automatically the best one — the best one fits the data source, output, integrations, risk profile and work process.
3. Prompting and context engineering
Good results depend not only on the question but on the context. Goal, role, relevant background, desired format, quality criteria and boundaries should be supplied so the model can actually follow the task. At an advanced level, context engineering is added: which information belongs in the context at all? Which documents are supplied via retrieval? Which tool descriptions are necessary — and which only generate unnecessary tokens?
Go deeper: Prompt engineering for SMEs — with frameworks and practical examples for outputs you can use straight away.
4. Critical review, data protection and governance
A linguistically convincing output is no proof of quality. Employees have to recognise when source checks, four-eyes review or human approval are required. Equally important is the question of which data may be entered into which system. As soon as AI touches customer data, contracts, internal knowledge bases or operational systems, an individual productivity question becomes a governance question.
AHEAD OF TIME supports companies here with AI transformation, education & engineering — from tool-stack audit through policies and AI Act classification to training and technical implementation.
5. Understanding agents, RAG and automations
Advanced AI usage shifts from the single prompt to the system. A RAG setup connects a model to selected knowledge sources. Tool calling grants access to defined functions. Agentic systems can plan multiple steps, call tools and keep working based on intermediate results.
Not everyone has to be able to build these architectures. But leaders and process owners should understand what changes: a system that only drafts a text carries a different risk profile than an agent that sends emails, modifies CRM data or triggers orders.
You will find practical examples in the AI & automation topic cluster — including AI agents, multi-agent systems and automation setups for SMEs.
6. Token, cost and energy efficiency
Competent AI use also means not burning compute unnecessarily. Long redundant contexts, needlessly verbose outputs or a large reasoning model for trivial tasks drive up cost without improving quality. In API-based applications billed per token, this competence becomes directly commercially relevant.
The goal is not "as few tokens as possible" but: as little unnecessary compute as possible at the quality the use case requires. That includes lean contexts, appropriate output lengths, caching of recurring content and choosing the smallest model that reliably meets the quality bar.
Not everyone in an SME needs the same AI competence
A sensible capability model is role-based. Management makes different decisions than marketing, HR or a development team. Uniform mandatory training for everyone does create fundamentals, but it does not replace role-specific depth.
| Role | Particularly relevant AI skills | Typical question |
|---|---|---|
| Management | Opportunities/risks, governance, use-case prioritisation, investment decisions | Where does AI create measurable business impact — and which risk do we accept? |
| Marketing & sales | Research, content, tool choice, prompting, automations, quality control | Which tasks can be done faster without losing brand and quality? |
| HR & people | AI literacy, data protection, policies, training, human oversight | Which roles need which skills, and how do we avoid shadow AI? |
| Operations | Process analysis, automations, agents, data quality, exception handling | Where does automation pay off, and where must a human decide? |
| IT / development | APIs, RAG, agents, MCP, vibe coding, security, evaluation | How do we integrate AI into existing systems and permissions in a controlled way? |
Six levels: a workable AI competence model
For the AI skills check by AHEAD OF TIME we use a tiered model with six independent levels. It is not meant to sort people into "good" and "bad" but to make visible which capabilities are already there and where the next sensible learning step lies. Each level has its own question catalogue; whoever clears the threshold unlocks the next one.
Level 1 — AI Beginner You know individual terms and tools, but the overall picture is still forming. The biggest lever right now is the fundamentals: what a language model actually does, where it is reliable and where it is not.
Level 2 — AI Practitioner You already work with AI tools and make workable day-to-day decisions. What is still missing is the systematic side: which tool when, and how to verify results reliably.
Level 3 — AI Advanced You have an overview of the tool landscape, know the typical failure modes and can judge results properly. The next step is towards building: your own workflows, context and guardrails instead of one-off requests.
Level 4 — AI Integrator You no longer use AI tool by tool but connect tools into workflows: data moves from one system to the next and steps build on each other. The next lever is making those chains reliable and traceable.
Level 5 — AI Architect You build things yourself — agents, automations, data paths. You know not just the tools but where they break: permissions, retries, failure modes. The next step leads away from technology towards the organisation.
Level 6 — AI Strategist You think about AI from the organisation outwards: who is allowed to do what, what a decision rests on, which dependencies a company takes on. This is the level at which AI either carries its weight or becomes expensive.
Why six levels instead of four? Because otherwise the jumps get too big. Between "uses tools safely day to day" and "assesses architectures" sit two very different capabilities in practice: chaining tools into workflows (integration), and building and operating those workflows yourself (architecture). And above that lies another distinct layer that is not technical but organisational: permissions, decision bases, dependencies.
In addition, specific competence dimensions can be assessed separately — AI terminology, tool choice, or efficient use of tokens, cost and energy. That is useful because someone can prompt very well and still have clear gaps in data protection or tool selection.
→ Take the AI skills check now — six levels, a dedicated question catalogue per level, and a classification with a concrete next learning step at the end. An overview of both checks — company maturity and personal AI knowledge — shows which one fits your starting point.
AI competence, governance and AI literacy
For companies with an EU market connection, a regulatory dimension is added. Article 4 of the EU AI Act obliges providers and deployers of AI systems to take measures to ensure a sufficient level of AI literacy among the people who operate such systems on their behalf. The European Commission stresses a context- and risk-based approach: technical knowledge, experience, education, the specific deployment context and the risks of the systems used should all be taken into account (European Commission, 2026).
Important for Swiss SMEs: the EU AI Act does not apply blanket-wise to every company in Switzerland. Whether and which obligations apply depends on role, system and EU connection. Regardless of the legal applicability, the underlying idea makes business sense: whoever deploys AI should enable the affected employees to understand the opportunities, limits and risks of the systems actually in use.
The Commission also does not prescribe that everyone must reach the same competence level. Different learning stages and approaches can explicitly be appropriate. Which is precisely why a role- and level-based model serves SMEs better than a one-off standard course for everyone (European Commission, 2026).
What management should do now
- Make visible where AI is already in use. Do not only record officially licensed tools. Informal use inside teams is part of the starting position too.
- Do not confuse competence with tool access. A ChatGPT, Copilot or Gemini licence is not an enablement concept. Check whether employees understand risks, data rules and quality control.
- Prioritise roles and use cases. Start where AI is used frequently, or where impact and risk are high.
- Measure the skill gap and define learning paths. Fundamentals for everyone, depth by role and need. Assessment and training belong together.
- Connect learning to real application. The strongest learning progress happens on real tasks: your own research, your own processes, your own documents, your own quality requirements.
The OECD reaches a similar conclusion: training is a central lever for turning AI use into better results. Employees who receive training more often report positive effects on job performance and working conditions (OECD, 2026b).
If you want to turn that into a concrete company agenda: AI transformation, education & engineering at AHEAD OF TIME combines assessment, governance, team training, roadmap and technical implementation.
Conclusion: AI competence is not a course, it is an organisational capability
The most important shift in perspective for SMEs is simple: AI competence is not the same as "being able to operate ChatGPT". It covers understanding, application, critical review, tool choice, governance and — depending on the role — advanced technical concepts.
That does not mean everyone has to become an AI specialist. Quite the opposite: a good capability model prevents unnecessary learning effort because it accounts for different roles and levels. Management needs orientation and decision-making ability. Business functions need confident application skills. Technical roles need deeper architecture and security knowledge.
Building these capabilities systematically is what turns AI experiments into dependable ways of working. Which is exactly the subject of part 2 of this series: how can AI competence be tested sensibly, without merely quizzing tool knowledge or memorised terminology?
Q&A — AI skills in Swiss SMEs
What does AI competence mean in a company? AI competence covers understanding the opportunities and limits of artificial intelligence, plus the ability to apply suitable AI tools safely, critically and economically to concrete tasks. Depending on the role, data protection, governance, RAG, agents or technical integration are part of it too.
Does everyone have to become an AI expert? No. Most roles do not require deep development knowledge. What works is a tiered model: shared fundamentals for everyone and deeper competence where tasks, responsibility or risk demand it.
Is a ChatGPT or prompting course enough for an SME? As an entry point a prompting course can be very valuable. For durable AI competence it is not sufficient on its own. Tool choice, critical review, data protection, governance and role-specific use cases belong in the picture as well.
How can an SME measure its employees' AI competence? An assessment should cover several dimensions and not just test brand or tool knowledge. Sensible areas are fundamentals, application, tool choice, security, terminology and — for advanced roles — RAG, agents and architectural understanding. The AI skills check maps this across six levels.
Which AI skills does a management team need? Above all the ability to assess opportunities and risks, prioritise use cases, define governance, judge investments and build the right capabilities in the company. Management does not have to operate every tool itself.
Is AI literacy relevant for Swiss SMEs because of the EU AI Act? For Swiss companies, legal applicability depends on the specific EU connection and the role in the given AI system. In substance, AI literacy is relevant regardless: employees should understand the systems in use, their risks and how to handle them.