Two things shipped on agor.me this week. The mic icon in the chat widget is live. Click it, and you are talking to Leo, our voice agent. Tell Leo what you want, and Leo can book a consultation directly into the calendar without you ever touching a form. The same widget still does text. Same agent, two ways in.
The second thing is quieter, and I think more useful in the long run. Every post on this blog now has a Leo-voiced audio version. There is a small play button under the title. Press it, and the post reads itself in the same voice you would meet on a call. You can read with your eyes, or you can listen on a walk, in a car, while you cook. The words and the audio are the same words, and the player highlights them as Leo reads.
Both of these are small. Neither is a moonshot. They are the things I want to write about anyway.
Why the chatbot in the corner has been wrong for years
Most consulting sites have had a chatbot widget for years now. You know the shape. A colored circle in the corner that opens a small box. You type a question. It answers with a marketing line and a link to a contact form. Sometimes a human takes over, sometimes not, mostly not.
That widget has always been a costume. The pitch was "we use AI in our work", and the proof was a chatbot that could not actually do any of the work. The asymmetry is comical when you stop to look. The whole point of an AI consulting firm is that AI can act inside a real business and produce real outcomes. The website should be the first place you see that. Instead it has been the last place.
The mic icon on agor.me is a small fix to that. You talk. Leo listens. Leo has the same set of tools the text version has. Leo can check the calendar, hold a slot, send the booking email, and confirm. The flow takes about ninety seconds when you know what you want. There is no form. There is no "we will get back to you within one business day". You hang up with a meeting on the calendar.
It is one feature. It will book a meeting when a meeting is the right next step, and it will say it does not know when it does not know. The point is that the front door of the site finally matches the pitch. If we tell you AI can do real work in your business, the front door has to be a place where AI does some real work in ours.
The other half: read it or hear it
The blog audio is the lazier version of the same idea. I write these posts long. They run two thousand words. Most people who would get value out of them will never sit at a desk and read two thousand words. They will be driving to a meeting, walking the dog, on a flight, between calls.
So the posts read themselves now, in Leo's voice, generated server side by xAI's Grok TTS. The first time you load a post, the audio renders and caches. After that the player streams the cached MP3 in a few hundred milliseconds. The player highlights words as Leo speaks them. You can click any word in the post and the audio jumps to that point. It is a small thing, but the alignment is the kind of detail you only notice when it is missing.
The audio is not a robotic narration. It is the same voice you would talk to if you opened the chat widget and clicked the mic. The site is starting to have a voice in the literal sense. One agent, one voice, one set of words across reading, listening, and speaking. The shapes of the surface are different. The thing behind the surface is the same.
What this is a small piece of
Voice as an interface to real work has been promised for so long that we have all stopped paying attention. Most voice products you have used in the last decade have been timer-and-weather toys, with a little smart-home glue. The thing those products never crossed was the line into doing real work that has consequences. Booking a meeting on a real calendar with a real client is on the other side of that line. Not by much. But it is on the other side.
The blog audio is the same line, smaller. Decent narration on demand for any piece of writing was a research demo two years ago. Now it is a build step in a daily blog pipeline that nobody thinks twice about. The cost is low enough to give every post the treatment whether anyone listens or not.
This is the texture of how AI lands in real businesses. Not as a single big launch. As a hundred small surfaces that quietly start to do things they could not do six months ago.
The way to test if you are early on this curve is simple. Click the mic on the chat widget. Tell it what you want. See what happens. Then come back and press play on a post. Listen to it while you do something else.
If those two minutes feel obvious, you are ahead of where most of the market is. If they feel new, that is the gap. The gap is where the next two years of work in your business is going to come from.