Your job software can now tell you why a job lost money. Only if you told it first

Fergus, the job management platform used by more than 26,000 tradespeople across Australia, New Zealand and the UK, has launched what it calls AI-first job management.

There are three parts. Fergus Assistant sits inside the web product where owners and office staff run jobs, schedules, quotes, invoices and reporting. On mobile, through the Fergus Go app, a tradie can type or talk on site to capture what happened, and have it land in the job, a quote or an invoice instead of being retyped back at the office. And a new AI-first experience, Fergus AI, is rolling out across all three markets next week. The assistant and the mobile app are available now. The release doesn’t mention pricing.

The examples Fergus leads with are the ones any trade business owner will recognise. Ask it “Why is this job losing money?” or “Which jobs made us the most money last month?” and it answers. Describe a report and it builds one.

That is genuinely useful. But one line in how Fergus describes it matters more than any of the features.

The answers come from what is already in Fergus

Fergus’s pitch is that there is no separate system to integrate, no data to move and no prompting skill required, because the platform already holds the customer, site, job cost, quote and invoice records. The AI works from that.

This is the right design. It is also the whole catch, and it is worth saying plainly before anyone asks their software why a job went sideways.

An AI that answers from your job data can only be as right as your job data. If the hours and materials on a job are complete, “why is this job losing money?” gets a real answer. If they are not, it gets a confident answer about the wrong thing. It will not say “I don’t know, half your labour is missing”. It will find the most plausible explanation in what is there.

We made a version of this argument last week about AI agents and what they can actually read. This is the trades version, and it is more concrete, because job costing has a very specific list of things that either went in or did not.

What “why is this job losing money?” actually needs

For a job’s profit to be knowable at all, by a person or a machine, four things have to be in the job.

• The labour, logged against the job. Not “general”, not “workshop”, and not reconstructed from memory on Friday afternoon. If two days on a bathroom reno got clocked to the van, the AI will tell you the reno was a winner.
• The materials, costed to the job. Supplier invoices matched to the job they were bought for. Stock pulled from the van counts too, and it is the bit that most often goes missing.
• The variations. The extra point the client asked for on the day, the rotten subfloor nobody could have quoted. If they were done but not recorded, the job looks like you underquoted when really you under-billed.
• The original quote, with line items. “Why did this lose money?” is really “where did actuals drift from the quote?”. A single lump-sum figure gives the AI nothing to compare against.

Most trade businesses have one or two of those nailed and the others patchy. That is normal. It just means the first honest thing an AI job assistant will tell many owners is not where their margin went, but a guess.

Voice capture is the fix hiding in the launch

Here is the part of the release that deserves more attention than the chat questions.

Job costing fails for one reason: entering things at the time is a pain. You are under a house, your hands are filthy, and the timesheet app wants six taps. So it waits until the end of the day, then the end of the week, and by then it is fiction.

Talking into your phone at the van before you drive to the next job, “two hours on the Smith job, used the last of the 20 mil PEX, client added a tap in the laundry”, is the first genuinely realistic way to get that information in while it is still true. If the mobile capture works as described, it does more for the quality of your data than the question-answering does for your insight. The questions are the reward. The voice notes are the work that earns it.

Keep a person between the AI and the customer

The release is about getting quotes and invoices drafted faster. It doesn’t say anything about checking them before they go out, and that is worth setting up yourself.

For the first couple of months, treat anything AI-drafted that a customer will see as a draft. A quote is a promise. If a spoken note becomes a quote with the wrong quantity or a missed line, the customer holds you to the number, not to the software. Faster drafting plus a thirty-second human read is still a big time saving. Faster drafting that sends itself is a new way to underquote at scale.

The twenty-minute test before it lands

Before you ask the AI anything, try this. Pick your last five finished jobs and work out, by hand, from what is recorded in your job system, whether each one made money and roughly how much.

• If you can do it and the numbers look right, your data is ready. Ask the AI the same question and see whether it agrees with you. That is also the fastest way to learn how far to trust it.
• If you can do it but the numbers look wrong (“that job definitely didn’t make 60%”), something is not being recorded. Usually labour. Fix the habit before you trust the answers.
• If you can’t do it at all, the AI can’t either. It will just sound more certain than you.

None of this is Fergus-specific. Whatever you run jobs in, the principle is identical, and AI features inside trade software will not stop at one platform.

The func.digital take

The most useful thing about this launch is that it is not another tool. For a tradie already paying for Fergus, the AI turns up inside the software they already use, which is exactly where AI should live. No integration project, no new login.

That puts the effort where it always belonged: getting the job in cleanly at the start and the costs in honestly along the way. It is the same point we made about the 16 hours a week tradies lose to admin. Fix the flow first, and the smart features have something worth being smart about.

The bit we help with sits at the front of that flow. How an enquiry arrives from your website, whether it lands in your job system as a job with the right details rather than as an email someone retypes, and whether it gets followed up before the customer rings the next tradie. If you want a plain read on where your jobs start leaking information, before any AI starts answering questions about them, that is what our free digital systems audit covers. No jargon, no hard sell.