Last winter we put a rotary phone in a theatre foyer in Eindhoven. Twenty-six days. 613 conversations. Visitors picked up the handset after a show and told an AI what stayed with them. We called it P.I.M., the Parktheater Impact Meter. The full story is in the case study. This piece is about what we'd tell another venue thinking about something similar.
Why voice beats forms in a foyer
Cultural venues spend serious money trying to understand their audiences. Star ratings, exit surveys, post-show emails. Most of it produces what you'd expect: a number you can put in a quarterly report and very little of what actually moved the people who came.
The problem isn't bad surveys. The problem is that survey-shaped thinking changes the answer. When a visitor sees a form, they answer the form. When a visitor talks to a phone, they tell stories. We measured 142% more words per response on average compared to written formats, and the qualitative quality jumped further than the count suggested. People said the kind of thing they'd say to a friend, not to a comments box.
Voice is the most natural interface there is. From the moment we can speak, we use it to make sense of what just happened. The trick of an installation like this is to hand visitors something familiar enough that they don't think about the technology, then to step out of the way.
What "142% richer" actually means
The number is not magic. It's a measured comparison: average words per voice response in our pipeline, against average words per response in conventional written formats used by the same venue. 142% more, on average. The variance is high. Some visitors said two sentences. Others spoke for over two minutes.
The substantive shift is what you do with that volume. Written feedback compresses to ratings. Voice feedback opens up. We saw:
- Reflections on what visitors planned to do differently after the show
- Specific, actionable feedback about elements of programming and venue experience
- Emotional impact dimensions you cannot capture in a five-point scale
- Stories about who visitors planned to share the experience with, and why
- Honest criticism about things that did not land, framed in language a programmer can act on
None of that fits in a star rating. All of it fits in a structured AI summary that runs nightly into a dashboard.
A survey gives you a number. Voice gives you the words behind it.
What to think about before placing a voice AI installation
If you run a cultural venue and you're thinking about something similar, here are the questions we wish more venues asked first.
1. What's the question you actually need answered?
Voice AI is most useful when there's a question your existing tools can't answer. "Was the show good?" is not that question, your booking and exit data already answer it. "What stayed with the visitor after the show?" is the kind of question voice is built for. Be specific about what you'd do differently if you knew.
2. Where does the installation belong?
Foyers, lobbies, queue lines, and the spaces between exhibits work best. The visitor should already be in transition, with the work fresh in their mind, and a moment of attention to spare. Avoid placing the installation where visitors are tired or distracted. Avoid placing it where it competes with bar service or coat check.
3. Whose voice is it?
The voice persona is the most important design decision. It should match the tone of your venue. P.I.M. spoke a warm, considered Dutch, no scripted catchphrases, no hype. A natural-history museum would want a different voice. A nightlife brand would want a third. The installation will sit inside your house style for a long time. Treat the persona as a brand asset.
4. How does GDPR work?
From day one, not bolted on. We anonymise in real time, with personally identifying details stripped before storage. Retention windows are short and configurable per venue. The processing region is documented. We can run the pipeline inside your tenancy if your legal team requires it. None of this is exotic, but it has to be in the design from the start, not after the pilot.
5. What's the dashboard story?
If the data does not land somewhere your team already looks, it does not exist. P.I.M. integrates with the theatre's existing audience-research tools. The dashboard is a complement, not a replacement. Plan the integration before building the installation.
6. What's the sustainability of the run?
An installation that runs for one weekend is a marketing stunt. An installation that runs for a season is research infrastructure. Plan for the second from day one: power, network, content moderation, on-call protocol, end-of-life. We did 26 days unattended with zero content incidents because the safety architecture was designed for the long run.
Where this works beyond theatres
The architecture transfers. Museums (with characters from the era of an exhibition), attractions (with queue-line companions or in-set characters), exhibitions (with vitrines that explain themselves), retail (with brand-voiced assistants for products that need a story), brand activations (where a phone or a kiosk holds a real conversation instead of an iPad).
The form factor doesn't have to be a phone. We've sketched versions inside vitrines, as historical-character intercoms, as queue-line companions, and as room-scale installations. The hardware adapts. The voice engine, conversation design, content safety, GDPR pipeline, and dashboard all carry over.
What it costs to find out
A 30-minute call is the cheapest way. Email lets@reckon.works with what your venue is, who walks through it, and what you'd like to learn. If voice AI fits, we'll scope a pilot. If it doesn't, we'll say so.
More on what we build: voice AI installations. Or read the Parktheater case study in full.