Voice AI installations

Spaces that listen back. Objects that have something to say.

From the moment we can speak we use our voice to communicate, learn, and connect. Yet the moment people walk into a venue or up to a brand, they're handed a screen. We think that's a missed opportunity. So we build the alternative, end-to-end.

Voice is the most natural interface there is. Use it.

A voice AI installation is an experience where people simply talk and the space responds. A retro phone that asks how the show landed. A room where light shifts with the conversation.

We design, fabricate, and integrate in-house. Voice AI sits next to AV, lighting, sound, custom hardware, and software. Eindhoven means we have access to the prototyping infrastructure to make any of this real. The point is not novelty. The point is that real spoken responses give you something a screen never can: rich, qualitative, human.

A space that holds a real conversation tells you things no survey can.
What we build

Voice in. Story, light, sound, and movement out.

  • Talking objects. Phones, radios, intercoms, vitrines. Familiar form, unfamiliar intelligence. The object hides the technology so the visitor can be a visitor.
  • Interactive characters. A historical figure, a character from a show, an in-universe persona. Real-time conversation in your tone of voice. Safe-by-default content boundaries.
  • Responsive environments. Rooms that shift with what's said: lighting cues, soundscape, projection, even motion. Voice as a real-time director.
  • Live audience research. Pick up, talk, hang up. The structured insight (themes, sentiment, per-show patterns) lands in a dashboard, not someone's inbox.
  • Brand and event activations. Trade shows, openings, product launches. Voice AI that holds attention without an app or a queue.
Why we're built for this

Voice AI plus the hardware to hold it.

Most voice-AI work today happens on a screen. We're one of the few studios in Europe that builds the entire stack: the audio pipeline, the conversation design, the prompt architecture, the safety layer, the cloud workflow, the dashboard, and the physical object that holds it. Hardware design and fabrication, in-house, in Eindhoven.

The voice engine inside every installation we ship is Switch Board AI, a named Reckon. Studio product. Switch Board AI powers Parktheater's P.I.M. today, and the same engine drives a portable Voice AI audio tour we install on existing exhibits, RFID-aware environments, and lighting environments where rooms respond to who's in them. Voice today, fuller environments tomorrow, the same Claude-native stack.

  • Ultra-low-latency audio pipelines tuned for natural conversation
  • Content-safety prompt architecture that runs unattended for weeks at a time
  • GDPR-first data handling: minimisation, anonymisation, retention windows
  • Custom hardware: Pi-based or industrial, with networking, NeoPixel and AV control
  • Live dashboards that integrate with the research tools your organisation already uses
  • Multilingual: Dutch, English, with options for German, French, and others
  • Pathway into ambient sensing, RFID, and lighting environments for museum-scale builds
Proof

P.I.M. at Parktheater. 26 days. 613 conversations.

A retro rotary phone in a theatre foyer. After every show, visitors picked up the handset and told an AI what stayed with them. No app, no form, no account. Twenty-six days of unattended operation. 613 conversations. 142% richer than what written forms typically capture, with feedback the theatre could act on the next morning. Privacy-aware from day one.

613
Conversations
26 days
Live, unattended
142%
Richer than written forms
Read the full case
Frequently asked

Practical, before you ask.

What does a voice AI installation cost?

It depends on whether the project starts from your existing object or includes hardware design, fabrication, and AV integration. Pilots typically start at five-figure euros. We send a one-page proposal after a discovery call. The pilot scope is the conversation we'd have first.

How do you handle GDPR and recordings?

From day one. Conversations are processed in anonymised form by default, with personally identifying details stripped before storage. Retention is short and configurable per venue. We can run pipelines inside your tenancy if your legal team requires it.

How long does a pilot take?

From green light to a working installation: six to ten weeks for a phone-style installation, longer if hardware is bespoke. Then a live pilot of two to four weeks before going into a permanent run. Each phase has a clear go or no-go.

Can it work in our language and tone of voice?

Yes. The conversation persona is designed with you. We've shipped Dutch and English installations and can do the major European languages. Tone, pacing, vocabulary, and content boundaries are all part of design.

Do you also work with experience designers and animatronic builders?

Yes, where it's a natural fit. We can drop the voice and dialogue layer into installations other studios are building. Strictly under non-compete terms with the production studio.

Want a space that listens back?

Voice AI installations, end to end.

Objects, characters, rooms. Voice in, story out, in your venue. lets@reckon.works.

Email lets@reckon.works See P.I.M. at Parktheater