We Hold AI to Standards We'd Never Hold a Human To

And it's costing us...

We build AI bots that answer phones for businesses across New Zealand. And almost every time, we hit the same wall. Not a technical wall. A people wall.

The moment we demo an AI solution - usually one enhancing an existing process - the questions start flying. Is the language perfect? Does it pronounce all words correctly? Is every piece of information 100% accurate? Does it warn callers before collecting their personal details? Is it New Zealand Privacy Act compliant? Does it use zero-retention mode so nothing ever touches a disk?

Fair questions, all of them.

But here's the thing. We've seen the thing it's replacing. We've seen medical receptionists asking patients for their date of birth, home address, and reason for visit - out loud - in a waiting room full of strangers. We've seen personal information taken over the phone - a channel I'm not fully convinced is as secure as we'd like to think, written on Post-it notes, and stuck to monitors visible through street-facing windows. We've heard wrong information given out with complete confidence. We've seen junior staff fielding complex calls with no script, no escalation path, and no idea what to do.

Humans are brilliant. Humans are also imperfect, distracted, undertrained, and sometimes just having a bad day.

Nobody's demanding we get rid of humans. This technology can simply field a large portion of calls, freeing the humans to specialise and focus on the work that actually needs them. Better use of their time means better quality from them too.

And it's not just good for the staff - it's good for the customer. AI means people can get information at a time that suits them. Ten o'clock at night. While they're walking the dog, using their voice instead of filling out a form. On whatever channel feels natural to them. The alternative is forcing customers to call during business hours - when they're likely busy, possibly stressed, squeezed into a five-minute gap in their day. You're not getting them at their best, and that frustration often colours the entire interaction before it's even started. A customer who can get help on their own terms is a calmer, happier customer. That has to be a good thing, right?

Yesterday I was at an AI conference aimed at small and medium businesses. There were some genuinely good speakers showing real, practical benefits of AI - and that was encouraging to see. But a number of the sessions leaned heavily on cautionary tales: chatbots saying the wrong thing, runaway token costs, legal liability. Important topics, for sure. The problem is that those stories rarely came with the other half of the picture - the guardrails that now exist, the configuration options, the maturity of the technology today compared to when those incidents happened.

When the risk gets all the airtime and the solutions get none, businesses walk away spooked.

Here's the thing about most of those horror stories: they come from early-generation AI - misconfigured, unguarded, deployed without thought. That's not the technology we're working with today. Modern AI has guardrails. It can be configured to escalate, disclaim, and stay in its lane. A balanced conversation would include that.

We're not saying AI is perfect. It's not. We're saying the bar shouldn't be "is this perfect?" - because nothing is. The bar should be "is this better than what we're doing now?" And in most cases, with the right setup, it is.

The double standard is real, and it's holding good businesses back. AI gets scrutinised to within an inch of its life. The status quo gets a free pass. Meanwhile, the patient's personal details are on a Post-it note on the monitor.

Incendo builds practical AI solutions for New Zealand businesses. If you want to talk through what this means for you - get in touch.