You heard the conversation. This is the thing we were talking about.

The short version of the hour: AI made building cheap, and it did not make trusting cheap. Everything below is what we do about that — the operating system we run, and the product it produced. FieldTaskora is proof-of-work software for businesses with crews in the field, and it is built the way the conversation describes.

Your Next Employee Might Be AI

Building, Trusting and Governing the Future of Work · 1 hr 4 min

AI, building with it, and the harder half — how you come to trust it. Governance, and what work looks like when software carries more of the responsibility.

Listen to the episode

What we got into

The frame

Treat it like a new employee

Nobody hands a new starter the keys on their first morning. They get a process to follow, a written standard, boundaries, and someone checking the work until it earns trust. A model deserves exactly the same and usually gets none of it.

Counter-intuitive

The best instruction is often “slow down”

Speed is the part that arrives free. The instruction that improves the output most is the one that stops it charging ahead — check this first, prove that, do not move until the gate passes.

The bottleneck

It builds faster than anyone can read it

The constraint has moved. It is no longer how fast software can be written, it is how fast a human can understand what was written well enough to stand behind it. That gap is where the risk now lives.

Accountability

Nobody gets to blame the model

When AI-built work goes wrong, responsibility does not transfer to the thing that typed it. Somebody signed it off, shipped it, or sold it. Deciding who that is before it happens is most of what governance means.

What changed

Nobody wants the pitch any more

The startup conversation has moved from describing an idea to showing a thing that works. If you can build it in a fortnight, a deck explaining what you would build is a strange thing to bring to a meeting.

Where to start

Practically, today, without it going wrong

You do not need a strategy to begin. You need one job worth doing, a written definition of done, and a check that has to pass before the work counts. That is a week, not a programme.

And then the part with the crews in it

The reason any of the above matters to us: it is what the product is made of.

On the record

The gap is rarely a safety gap. It’s an evidence gap.

Your crews almost certainly did the job properly. The question is whether you can put the proof — dated, attributed, tied to that asset — in front of someone standing in front of you, three weeks later, on a site you have since left.

On the buyer

It isn’t built for the person doing the work.

It’s built for the person who has to prove it. The compliance manager, the contracts manager, the owner whose payment is held because the pack isn’t right. That distinction is the whole product.

On the ask

The answer is usually yes. We just need the problem.

The hard part now isn’t building the solution — it’s getting a clear description of what’s actually broken. Most people have been told no so often they’ve stopped asking. Tell us the real one.

It is easy to say “put guardrails on it”. This is what they cost.

I build a product where AI does the work rather than sits beside it, so the governance is not a policy document — it is code somebody has to maintain. What that actually looks like in ours:

  • Every AI action is metered and logged per organisation, behind a spend cap and a circuit breaker — so a runaway costs money once, not indefinitely.
  • A document a stranger sent us is data, never instructions. If you feed a model a customer’s specification, you have handed an outsider a keyboard unless you have decided otherwise on purpose.
  • Every AI path fails soft and says so on screen. No key, no answer, and the product tells the user that rather than quietly producing something worse.
  • A person closes every gate that matters. The agent drafts, matches, prices and packs. It signs off on nothing.

None of that is the interesting engineering. All of it is the reason the output can be relied on. If you are working out the same problem — in a product, or in how your own business runs — I am happy to compare notes properly rather than over a webinar.

And if you run crews in the field, that is the product — closest fit is utility and civil contractors, councils, inspection and certification businesses, and trades big enough to have a compliance headache. See the industries we build for.

If you came for the AI part

A fair few people will have listened to that conversation without a field crew in sight. The short version of what we run here, in case it’s useful:

  • Generation is the cheap layer.Code, copy, documents, research, video. It’s the least interesting part.
  • Gates are what earn the speed back. A ticket for every piece of work, a fixed definition of done that nothing skips, and checks that have to pass before anything ships.
  • Adversaries are the part nobody builds.A second model whose only job is to disagree with the first, and audits that go back and re-check what we already signed off. That’s where we find the things that were quietly wrong.

The one-line version: AI made building cheap, and it did not make trusting cheap. Everything above exists in the gap between those two.

If you’re working through the same problem in your own business, I’m genuinely happy to compare notes — mark.kuch@fieldtaskora.com. No pitch.

Beyond Growth with Reyan Fernando

Reyan spent two decades in IT before any of this — he founded and scaled his own IT company and led it through a merger with a managed services provider. That is a large part of why the conversation went where it did: he has sat in the seat, and he asks the questions of someone who has had to make payroll.

These days he is a Professional EOS Implementer®, working with leadership teams across Geelong, Melbourne, Sydney and Colombo on the unglamorous machinery of a business that runs without chaos — a clear vision, the right people in the right seats, and the discipline to actually execute. If your business has outgrown the way it is run, he is worth a call.

Beyond Growth is his long-form conversation with founders, leaders and operators — leadership lessons from business, life, and the journey within. Past guests include a detective of two decades who left to start an environmental business, an AFLW pioneer who debuted at 27, family-business operators and fellow technology founders. Worth subscribing to even if you never listen to ours.

Your crew runs the job. FieldTaskora writes the report.

Leave your email and we’ll let you know as early access opens.

No spam. We'll only email you about early access.

Worth reading next