Most people overcomplicate AI adoption so badly that they end up doing nothing. They evaluate tools for weeks, sit through demos, read comparison articles, and then quietly go back to working exactly as before. Here’s the reality: you don’t need an IT team, a rollout plan, or a consultant. You need one working workflow, the right tool, and about 15 minutes.
Two hours a week sounds modest. Multiply it by 52 and that’s more than a full working week recovered every year – from a single workflow.
Why most AI setups stay unused
The overcomplicated trap is real. Most SMEs either buy too many tools at once or spend so long evaluating options that they never deploy anything. A few months later, someone cancels the subscriptions because “we never really used it.”
The core mistake is starting with the tool instead of the problem. It sounds obvious, but it almost never gets followed.
Enrico Sottile, Technical Engineer at what., puts it well:
Non-tech companies should use polished, user-friendly interfaces – less power, but less fear. The right tool for the right person matters more than the most powerful option.
This is a critical insight that most AI vendors conveniently ignore. A developer might be comfortable in a terminal, running Claude Code alongside command-line tools. A project manager at the same company might need the clean Claude.ai interface. A marketing manager at an SME with no technical background? They need something even more guided than that.
There’s no single “right” AI tool. There’s the right tool for the right person. And for most non-technical business owners and their teams, the polished consumer interface – even if it offers less customization – is the one that actually gets used.
The gut-check question to ask before adopting any AI tool: does this save at least 30 to 60 minutes every single week? If the answer is no, it’s a toy, not a tool.
Also relevant: Before you automate anything, it’s worth making sure the underlying process is actually worth automating. Read why fixing your workflows before touching AI is often the higher-ROI first move.
The 3 AI workflows that deliver results in 15 minutes
Here are three specific workflows that deliver real, measurable time savings – drawn from what our team actually uses.
Stop letting meetings drain your time: automate notes and follow‑ups
Meeting-related admin quietly eats enormous chunks of the working week. Notes, follow-up emails, task lists from a single client call – it adds up faster than most people realise.
What the workflow actually looks like at what.: Optiverse records and transcribes the call. A project manager then triggers a custom-built skill that instructs Claude to fetch the transcript and do two things: draft follow-up emails for external participants, and generate Slack summaries for internal ones. The result is that nobody takes manual notes during the meeting, and instead of writing a follow-up from scratch, you’re reviewing and approving a draft that’s already in the right format and tone.
That last part matters more than it sounds. A well-written follow-up isn’t just a time task – it’s a judgment task. Getting the content, format, and tone right can easily eat 30 minutes on its own. This workflow cuts that to a quick review.
The time saving is roughly 30 to 60 minutes per meeting-heavy day. For anyone running three or four client calls a week, that compounds fast: easily 90 minutes or more recovered weekly from this single workflow alone.
The Swiss-specific angle matters here too. We switched internally from Google Gemini to Optiverse – a Swiss company that handles German and Swiss German significantly better, and connects directly to Slack and Linear. If your meetings happen in Swiss German or a mix of languages, using a tool that actually understands the dialect isn’t optional. It’s the difference between a useful summary and a confusing one.
Turn a full planning afternoon into a 10-minute brief
Lucas Stebler, Shopify Lead at what., shared something that illustrates this better than any generic example: for a new project, he feeds Claude one example milestones file and a short task list. Claude then creates all the tickets in the correct projects, links dependent ones, builds the milestones Excel from a template, assigns them, and sets everything to “refine.” It pulls context from Linear, Slack, and email through MCP connections – the integrations that act as the glue between Claude and all these platforms. In one case, it caught a go-live date that existed only in a Slack message, not in any brief.
A project board that used to take an afternoon now takes 5 to 10 minutes.
MCPs (Model Context Protocols) are worth understanding here, even if you never touch the technical side yourself. They’re what allow Claude to reach into Slack, pull from Linear, and cross-reference email threads without manual copy-pasting. Without them, these workflows are half as powerful. With them, AI stops being a writing assistant and starts being something closer to an operational layer across your tools.
What makes this work isn’t magic – it’s the combination of a clear example input, deliberate permission settings (Claude has to ask before most actions, which doubles as a QA checkpoint), and a human reviewing the output before anything goes live. The underlying principle: let AI handle the low-leverage, repeatable work and keep your own time for judgment and client work.
Getting AI to process the documents and emails you’re drowning in
This one applies broadly. Enrico summarises it well from a technical perspective: “I gather tasks and requirements and ask AI to analyse and summarise. It actually saves me a lot of time and mental energy.”
The same principle applies far beyond development work. A wave of client feedback that would take hours to manually cross-reference against agreed scope? Feed it to Claude alongside the relevant Figma designs, Slack threads, and QA tickets. In 30 minutes rather than three hours, you get a clear view of what contradicts the agreed design, what’s a genuine gap, and what’s already been flagged.
For email specifically: professionals spend somewhere between four and six hours a week on drafting and triage. Even a lightweight setup – a prompt template that handles your most common inquiry types, with AI drafting replies for human review – can recover one to two hours weekly. Setup time is under 15 minutes once you have a template written.
The key habit across all three workflows is the same: AI drafts, human approves. You’re not removing yourself from the loop. You’re moving yourself to a review role instead of a creation role.
Reading tip: 2026 Will Be the Year of Autonomous Workflows – Here’s Why – and starting with these small workflows now is exactly how you build toward that.
What a lean AI stack looks like and what it costs
“AI stack” sounds intimidating. It doesn’t need to be. For most SMEs, the minimal effective setup is two to three tools, used properly, that talk to each other.
Here’s a realistic starting point:
| Tool | Purpose | Approx. monthly cost |
|---|---|---|
| Claude Pro or ChatGPT Plus | Writing, summarising, drafting, analysis | ~CHF 20 |
| Make or Zapier – optional | Connecting tools and triggering workflows (becoming less essential as MCP adoption grows) | ~CHF 10–20 |
| Optiverse (or similar) | Meeting transcription and summaries | ~CHF 15–25 |
| Buffer or Publer | Social content scheduling | ~CHF 10–15 |
Total: roughly CHF 55–80 per month. A single ChatGPT Plus or Claude Pro subscription alone can return three to five hours per week in time savings if used consistently. The ROI case is not hard to make.
The golden rule: go deep with two tools, not shallow with ten. A well-used Claude subscription beats five underused apps every time.
One note on Make and Zapier specifically: these automation connectors have been the standard glue between tools for years, but their role is shifting. As MCPs become more widely adopted – allowing AI models to connect directly to platforms like Slack, Linear, or your CRM – the need for separate automation middleware starts to shrink. For now, they’re still useful for many setups. But don’t treat them as a permanent fixture.
One thing worth flagging specifically for Swiss SMEs in regulated industries (finance, health-adjacent, legal) is data residency. Many mainstream AI tools process data on servers outside Switzerland. If that matters for your compliance obligations, it’s worth checking. At what., when it’s required, we work with Swiss-hosted infrastructure to keep data where it needs to stay. You can read more about what that means in practice in our guide to data privacy and security for AI systems.
Once you have a working stack, the next lever is connecting it properly to your existing systems – your CRM, accounting software, inbox. That’s what turns a useful workflow into a reliable one. Proper tools integration is what separates “nice to have” from “we’d never go back.”
Start with one AI workflow, then scale
Two hours a week is not a stretch claim. It’s what one well-chosen workflow reliably delivers – and we’ve seen it happen for clients and internally at what. more times than we can count.
You don’t need to automate everything at once. Pick the workflow that costs you the most time this week. Set aside 15 minutes. Run it with real emails, a real project brief, or a real meeting recording. See what comes back.
The businesses that get the most from AI aren’t the ones with the most sophisticated setups. They’re the ones who started with one thing, got it working, and built from there.
If you’re not sure where to start, or you want to connect your AI tools properly to your existing systems, that’s exactly what we help with. Talk to our AI automation team – no sales pitch, just a practical look at where AI actually makes sense for your business.
