The most expensive interpretation setup in the room is often the worst

Not long ago I watched a major technology company host a flagship multilingual summit. Big stage, global audience, a main plenary followed by a wall of parallel sessions. On interpretation, they spared no expense — premium on-site booths, a brand-name equipment vendor, multiple interpreter teams. By every budget line, it was the deluxe option.

The result was mediocre. And it cost more than it should have.

I've spent over two decades at an unusual intersection: as an AIIC-level conference interpreter, and as an engineer who builds the platforms that deliver interpretation. From that seat, this outcome isn't surprising — it's a pattern. The organizations that spend the most on conference interpretation are very often the ones who get the least. And the reason is almost never the interpreters or the technology in isolation. It's the mix.

Interpretation isn't one product anymore. It's a portfolio.

A modern multilingual event has at least three delivery modes on the table:

• On-site human interpreters in booths — the gold standard for the high-stakes plenary: the keynote, the moment where tone, nuance and a live Q&A actually decide whether the message lands.
• Remote simultaneous interpretation (RSI) — the same human quality, delivered over a platform, with a fraction of the on-site logistics and none of the receiver hardware.
• AI captioning and machine interpretation — genuinely useful, and improving fast, for high-volume, lower-stakes, cost-sensitive sessions where "good enough, instantly, everywhere" beats "perfect."

Each has a sweet spot. Each has a failure mode. AI on your board-level plenary is a risk you'll regret; two premiumon-site booths on a routine breakout is money set on fire. The skill, the entire skill, is matching the right mode to the right room, and then paying for top quality only where quality actually changes the outcome.

Most buyers never get to that step, because they never get an apples-to-apples comparison in the first place.

Why procurement keeps overpaying

Three forces push large organizations toward the expensive-but-wrong setup:

1. The incumbent reflex. The equipment vendor who did last year's event gets this year's, no questions asked. Safe for the buyer. Rarely optimal.
2. Misaligned incentives. When your interpretation is bundled through a company that also sells the hardware, guess which solution you get steered toward. Not the lean one.
3. No benchmark. Most buyers never put a specialist interpretation architect next to their default vendor and compare. Without that comparison, "we spent a lot, so it must be good" quietly becomes the standard. It usually isn't.

The irony is that the companies most capable of thinking clearly about AI, the ones building it, sometimes applythe sharpest logic to their product and the dullest logic to their own event. They'll dogfood their AI beautifully in the breakout rooms, a smart, defensible call, and then default to the most conventional, most expensive human setup everywhere else, without ever asking whether the overall mix was right.

The rare skill is orchestration

Getting this right takes someone who cansit in two chairs at once: who knows, from inside the booth, exactly where human interpretation is irreplaceable — and knows, from the engineering side, exactly what the platform, the network and the AI can and can't do yet. That combination is rare. There are perhaps a few dozen AIIC members in this region; far fewer who also write code.

I don't say that to boast. I say it because it explains the pattern. The people advising most large events on interpretation are excellent at one half of the equation and guessing at the other. So they over-buy the half they know, and under-think the half they don't.

If you're planning a serious multilingual event, the most valuable question isn't "who's the biggest vendor" or "can AI do it all yet." It's: who can architect the mix so I pay for quality only where it changes the outcome. Ask that early, and the deluxe-but-disappointing result becomes very easy to avoid.

I'm always glad to compare notes with people wrestling with it.

‍