The director's objection is fair. You're already short-handed. Adding "maintain the audio guide" to someone's plate isn't a free decision, and nobody's pricing page mentions it.
So let's just answer the question. For a modern AI or app-based guide, the realistic ongoing cost is 5 to 15 staff hours per month at a mid-sized museum with an active program. For an old hardware guide running the same kind of program, it's 20 to 60 hours, sometimes more, and most of it has nothing to do with content. That gap is the whole story.
Worth it? At 10 hours a month for a system that reaches a meaningful share of your visitors in their own language, yes, almost always. There are exceptions, and we'll get to them. But the math is much friendlier than the objection assumes, and most of the friction you've been told to expect comes from a model of audio guide that the field is leaving behind.
What staff time actually goes into a guide today
Strip away the legacy assumptions for a second. On a modern guide, here's the work that recurs:
Content edits. A new acquisition arrives. A wall text gets rewritten. The curator changes their mind about how to frame a difficult work. With a CMS-driven system you load the new material, write a sentence or two of guidance about how it should be discussed, and move on. Twenty minutes per change is typical. A busy month might bring eight to twelve of these.
New tours. A school holiday is coming up. The education team wants a 30-minute kids' route. A donor wants a private tour highlighting a specific theme. Setting up a new tour from existing knowledge takes a few hours, sometimes an afternoon. You're sequencing stops, writing a persona prompt, choosing a tone. Real work, but not production in the old sense.
Reviewing AI output. This is where serious institutions spend honest time. Even when the system is well-tuned, a curator should walk a new tour, listen as a visitor would, and flag anything that lands wrong. Sensitive topics, contested attributions, recent restitution claims, anything where institutional voice matters. Two hours a month is a reasonable baseline. More around big openings.
Promotion and signage. The guide doesn't promote itself. New QR posters when a tour launches. A line in the welcome script. A social post when a fresh route goes live. We've covered why this matters in training museum staff on audio guides, and the short version is: this is the highest-leverage hour you'll spend.
Analytics and tweaks. Once a quarter, look at which tours visitors finish, which stops get skipped, where questions cluster. Make a few targeted edits. An hour, maybe two.
Add it up and you land in the 5 to 15 hour range for an active mid-sized museum. That's it. That's the maintenance cost.
How the old model burned hours
Now compare the legacy hardware setup, because this is the experience most directors are actually thinking of when they hesitate.
A hardware audio guide is a small operations business inside the museum. Every morning someone unplugs 60 to 100 devices from a charging rack and counts them. Every evening someone collects them, checks each for damage, plugs them back in. A device walks out the door — budget says 5 to 15 percent annual attrition — and someone files a replacement order. The headphones get sticky. The rubber covers tear. A unit freezes mid-tour and a visitor wants their five euros back.
That's before any content work. We've watched museums absorb 0.25 to 0.5 FTE on hardware audio guides at serious institutions, and the bulk of it is just fleet. Distribution, recovery, charging, repairs, returns, vendor RMA cycles. None of that touches the visitor experience in any positive way. It's pure overhead.
Then there's the content side, which on the old model is its own production cycle. A new permanent gallery opens. The vendor schedules a recording session four weeks out. Scripts are drafted, reviewed, revised, sent to voice talent. A studio day is booked. Translations are commissioned across however many languages you support, with a separate session per language. Files come back, get QA'd, get loaded onto the device fleet. Months elapse. Real money moves.
When something inevitably needs to change — a new attribution, a re-hung gallery, a temporary loan — the same cycle starts again. Most museums simply stop updating. The guide becomes a fossil. We covered the financial side of this in the five-year total cost of ownership piece, but the staff cost is its own problem. People stop believing the guide is something they can shape, so they stop trying.
Realistic time estimates per museum size
Numbers help, so here are the ranges we've seen hold up across institutions actually running modern systems. These assume the guide is live, has decent adoption, and isn't being neglected.
Small heritage site or single-collection museum (under 30,000 annual visitors). Two to five hours a month. Usually one person, often the director or a senior curator, doing it themselves. Light cadence: a content sweep every quarter, a new tour once or twice a year, occasional edits when something changes on the floor.
Mid-sized museum (50,000 to 250,000 visitors, active exhibition program). Five to fifteen hours a month. Often split between a curator (content), a marketing or visitor experience person (promotion), and reception staff (distribution support). Add an extra five to ten hours in the month before a major opening.
Large institution (250,000+ visitors, multiple permanent collections, regular temporary exhibitions). Fifteen to forty hours a month, occasionally pushing higher around big launches. Usually distributed across two to four people rather than concentrated. The work scales with the program, not with the system: more exhibitions means more new tours, more sensitive content review, more languages worth checking.
For comparison, the same three institutions on a legacy hardware guide running the same program: roughly 10 to 25 hours, 30 to 80 hours, and 80 to 200+ hours respectively. The shape of the curve is similar; the absolute numbers are several times higher, and most of the increase is invisible — it lives in receptions, IT, facilities — rather than in a budget line called "audio guide."
Where staff time is genuinely needed (and worth it)
Some of the hours above aren't just maintenance. They're the work that makes the guide actually good. Worth being clear about which is which.
Launching new tours. A guide with one tour gets used by maybe a quarter of the audience it could reach. Add a kids' tour, a quick-visit highlights tour, a deep-dive thematic tour, and adoption climbs. This is where staff time creates leverage, and it should be deliberate, not reactive. Plan two to four new tours a year and you'll see the difference in numbers.
Reviewing AI output for sensitive content. Restitution discussions. Colonial collections. Contested attributions. Religious objects. Anything where the institutional voice matters more than the average. We strongly recommend a curator listen to any tour that touches these topics before it goes live, and re-listen quarterly. This is one place where staff time is non-negotiable, and where AI without human review is genuinely the wrong answer.
Marketing and adoption. The single highest-return hour you can spend on an audio guide is the one that gets more visitors to use it. Reception script tweaks. Social posts. Email mentions to members. Better signage at the entrance. None of this is glamorous. All of it moves the number that matters.
Curating themes around the calendar. A Pride month tour. A Black History Month route. A late-opening "after dark" tour. A specific exhibition spinoff. These don't need to be built from scratch — they're recombinations of existing knowledge with a new persona and sequence. A few hours each. They keep the guide feeling alive in a way that no permanent content can.
These are the hours that are actively worth defending against the "we're too short-handed" objection. They're not maintenance. They're program work that happens to live inside the audio guide tool.
Where it's basically nothing
A different category: things that used to consume staff time and now don't.
Routine maintenance. No daily charging routine. No device counts. No lost-and-found inquiries. No vendor support tickets about a frozen handset.
Language additions. On a modern AI guide, adding Japanese or Basque or Brazilian Portuguese is a configuration change, not a recording project. Zero staff hours per language after the first one. The old model required a full studio day per language per content update.
Hardware lifecycle. No fleet replacement at year five. No procurement cycle for new devices. No write-offs.
Translation QA on existing content. When the AI handles translation in real time from your source material, a content edit propagates across languages automatically. Old model: every edit is a translation project times the number of languages.
Day-to-day distribution support. Visitors scan a code on their own phone. No queue, no desk staffing, no battery anxiety. Reception still mentions the guide — that's still the highest-leverage line in the script — but they don't manage it.
If you take the 0.25 to 0.5 FTE that a hardware program used to consume and ask where it went, this is where. Not a small saving. The kind of saving that lets a curator actually work on content instead of inventory.
When it isn't worth it
Honest answer: there are museums where even a modest cadence is too much.
A two-room volunteer-run heritage house with no paid staff. A site that opens three days a week with one person at the desk who's also the educator and the bookkeeper. An institution where the closest thing to "digital capacity" is one trustee who knows how to use Canva. If nobody owns the guide — even at two hours a month — it will rot, and a rotting guide is worse than no guide. Visitors notice. Reviews mention it. Staff stop recommending it because they've stopped believing in it.
The threshold isn't size. It's whether someone at the institution can plausibly take ownership. We've seen 15-person museums run excellent guides because the director cared. We've seen larger museums with formal "digital strategy" titles where the guide languished because no individual person felt accountable.
If you can't name the person before you sign, that's a yellow flag. Not a no, but a strong signal that the project needs to start with org design rather than software selection. Sometimes the answer is to wait until you can hire — or share — the role. Sometimes it's to scope the guide deliberately small (one tour, two languages, an annual review) so the work fits the available capacity. Either is honest. Buying a guide and hoping someone steps up rarely works.
The opportunity-cost frame
The director's question deserves the comparison it implies. Ten hours a month spent on the guide versus ten hours a month spent on what?
Updating the website? The website reaches visitors before they arrive. The guide reaches them in front of the work. Different funnel.
Producing a printed exhibition guide? A printed guide costs paper, design fees, reprints when content changes, and reaches one language. Same staff time, narrower reach, higher unit cost.
Hosting more in-person tours? In-person tours are wonderful and have a hard ceiling — they reach the visitors who can attend at the scheduled time in the language the guide speaks. Useful, but not scalable.
A podcast or video series? Higher production cost per minute. Lower in-museum impact. Different audience.
Direct interpretation through wall labels? Always worth doing, and not in tension with a guide. Labels and audio guides do different work.
There isn't another medium that delivers the same in-gallery, multilingual, on-demand, updatable interpretation for the same monthly hour count. That's the case for the guide as a use of staff time. Not "it's free." Not "it runs itself." But: of the things you can do with ten hours a month to improve what visitors actually experience inside the building, this is one of the highest-return options on the menu.
A closing position
Staff time is the cost nobody puts on the proposal. We've put it on this one. For a modern guide, the real number is small enough that the objection doesn't survive contact with the math, provided someone owns the work.
If your institution has that person, even part-time, an AI museum guide is one of the better uses of their hours. If it doesn't, fix that first. The guide will be better for it, and so will everything else they end up touching.
One point worth naming on the commercial side. Platforms like Musa price on per-interaction or revenue share with no minimums, so the staff-time calculation above isn't competing with a fixed capex bill — the hours only earn against usage that's actually happening. For a small institution that was priced out of audio guides entirely under the old hardware model, that matters: the role you're defining can be sized to the visitor engagement you actually have, not to the infrastructure bill you'd have to grow into.
If you want to think through what the staffing model would look like at your institution specifically, we're happy to walk through it. We'll be straight about where the hours land and where they don't.