Before You Build Voice Agents - the Quick Win Is Dictation

The last two articles were about the big vision. Voice as an interface, queried reliably instead of hallucinated.

Between today and that goal lies work. Connect data sources. Settle permissions. Define context. None of it is sexy.

Meanwhile a gain is lying around that almost everyone walks past. It sounds mundane: dictation.

The number that's hard to argue with

A practiced person types 40 to 60 words a minute. Anyone speaks at 130 to 170.

Three times as fast. Not an estimate from a glossy brochure, but a stable constant of human motor skill. Typing is a learned, slow translation of thought into finger movement. Speaking is the direct channel.

Run it for one role. Two hours a day producing text - emails, notes, proposal drafts, minutes. Dictate half of it instead of typing: 30 to 40 minutes back. Per day. No major AI project required. The feature ships with every modern operating system.

Scale it to a team. Ten people, 30 minutes, 220 working days. Over 1,100 hours a year.

Now subtract half of it, generously, for onboarding, corrections and the days it doesn't fit. What's left beats any other AI initiative at maturity level zero - at a fraction of the effort and risk.

The write-heavy roles gain most. Sales with call notes. Support with tickets. Consulting with minutes. Management with the flood of small decisions that otherwise drown in the inbox.

The real gain isn't the time. It's the friction that disappears.

Less friction beats saved minutes

Typing creates a threshold. If you have to formulate a thought and then type it cleanly, you'll skip it when in doubt. The note after the client call you'll write later and then forget. The objection in the workshop nobody captures.

Dictation lowers that threshold. Ideas get secured before they evaporate.

The behavior changes measurably. More gets documented. Earlier. More completely.

For what comes next, that's gold. This early-captured documentation is exactly the knowledge a good system will need later. Today's quick win is tomorrow's data foundation.

The honest limits

A promise without limits is marketing, not advice. Dictation fails in four places.

Noise. Open-plan office, workshop, train. Recognition suffers. Without a decent microphone you correct more than you save.

Sensitive content. Personal data, health data, contract details - spoken out loud, possibly through a cloud service. Under Swiss and EU data protection law that's not a detail, it's a compliance question. Settle in advance what may be dictated and where it goes.

Mumbling and pace. Rush your speech and you get poor results. Dictation is a skill, not magic. A few days of getting used to it are part of the deal.

Jargon. Product names, abbreviations, industry terms come out badly at first. Trainable. In the beginning it costs patience.

None of these limits is a knockout. But hiding them produces exactly the disappointment that strangles every further voice initiative: "we tried it, it didn't work."

Why the small step sets up the big one

Voice AI at scale rarely fails on the technology. Almost always on adoption. People who have never spoken to a machine won't trust it with decisions.

Dictation clears that hurdle at the lowest possible threshold. Low risk. The benefit is immediate. And it gets people used to speaking to software instead of operating it.

When real voice intelligence arrives later, you meet a team that already knows the gesture. You're no longer selling a revolution. Just the next logical step.

A second effect gets overlooked. Dictation forces clarity. Say a thought out loud and you notice faster whether you've finished thinking it. Many report that dictated notes come out more structured than typed ones. Speaking forces a thought to take shape immediately. The speed tool quietly becomes a thinking tool.

Where we stand

We advise most mid-market companies to take the unspectacular order. First the quick win that saves money today and builds adoption. Then the expansion into an interface, once the data foundation holds.

Start with the ERP voice agent before your team enjoys dictating, and you're building on sand.

Your next step

Before you roll out dictation: do the math. What do you actually get back?

Our ROI Calculator does it in two minutes. Roles, time shares, team size in. Time saved out - as a number that's hard to argue with in a budget conversation.

The number first. Then the decision. In that order you'll convince the skeptics too.