How to price a job when AI does 60 per cent of the work

A client asks for a quote. You do the sums the way you always have: estimate the hours, multiply by a rate, add a margin, send it off. Except this time, AI has cut the drafting, the formatting and the first pass research down to a fraction of what they used to take. The hours have collapsed. If you price the job the old way, so does your revenue, on the exact jobs where your judgement and your tools are now worth the most.

This is the trap sitting in front of a lot of Australian business owners right now, and it will not fix itself. Cost plus pricing was never really about cost. It was a proxy for value, built at a time when time and value moved together. Once a task can be done in a fraction of the time, that proxy breaks, and the business that keeps pricing on hours is the business that quietly gives its margin away.

Why cost plus breaks the moment the cost drops

Cost plus pricing works like this: you estimate the direct cost of doing the job, staff time mostly, and you add a margin on top. It is simple, it is defensible, and for decades it tracked reasonably well with what clients were willing to pay, because effort and outcome were roughly linked.

Here is the arithmetic, shown plainly. Say a report or a proposal used to take a professional employee ten hours to draft, review and format. The Australian Bureau of Statistics puts the median hourly rate for Professionals at $59.50 as at August 2025, so the direct labour cost sitting behind that job is roughly $595. Now AI drafts the first version, restructures it and catches most of the errors, and the same job takes four hours of human time instead of ten, a genuine 60 per cent reduction. Run that through a cost plus model and your direct cost falls to around $238. If your price is cost times a fixed multiplier, your quote falls by roughly the same proportion, even though the client still receives the same finished report, delivered faster and with fewer mistakes.

That is the whole problem in one paragraph. The client's outcome has not shrunk. Their risk has not grown. If anything both have improved, because the job now arrives sooner and with a second set of eyes built into the process. But your price, if you let cost drive it, shrinks anyway. You are punishing yourself for getting better.

What value based pricing actually means in practice

Value based pricing is simply this: the price reflects what the outcome is worth to the client, not how many hours it took you to produce it. It sounds abstract until you notice you already do it in places. A tradesperson does not charge less for a job because they have done it a thousand times and it now takes them half the time it took an apprentice. A lawyer does not discount a contract review because a template made the first draft quicker. The price holds because the client is paying for the right answer delivered reliably, not for the hours behind it.

The same logic applies once AI is doing 60 per cent of the drafting on a job. Your price should be anchored to turnaround time, accuracy, the certainty of getting it right the first time, and the judgement you apply on top of what the tool produces, not to the shrinking hour count. This is not a case for charging clients more for less effort out of habit. It is a case for holding your price where the value sits and reinvesting the freed up capacity into more jobs, better quality control, or work you previously had to turn away.

There is a broader number behind this argument worth sitting with. Deloitte Access Economics modelling, commissioned by Amazon Australia and released in November 2025, found that SMBs moving from basic to intermediate AI use could expect a 45% increase in profitability, which jumps to a 111% increase for a business moving from intermediate to enabled use. That gain does not come from discounting jobs because they got cheaper to produce. It comes from businesses that kept their pricing anchored to value while their delivery capability improved underneath it. The same research also notes that just 5% of surveyed SMBs using the technology are fully enabled to realise its potential benefits, which tells you most businesses are still leaving this margin on the table.

Reading the vendor proposal and finding the fraction

Before you touch pricing at all, look at what you are already spending across agency retainers, software seats, support contracts and content production, because the AI capability you are trying to price around is very often a fraction of a line item you are already paying for, not a brand new cost centre. A marketing retainer that includes drafting, a support contract that includes documentation, a bookkeeping seat that includes reconciliation checks. These are the exact tasks AI now touches, and the honest question is not "what will this AI tool cost us" but "what fraction of this existing line are we already paying someone else to do manually".

When a vendor proposal lands on your desk, read it the way you would read a contractor's quote on a renovation. What is the deliverable, precisely, and what does it replace. What is the ongoing cost once the pilot is over, not just the introductory rate. What happens to your data, your client files and your existing systems if you cancel next quarter. A proposal that cannot answer those three questions plainly is not ready for a decision, no matter how confident the sales deck sounds. If you want a second set of eyes on where your current technology spend actually sits, an audit of what you are already paying for is the more useful first step than another subscription.

Measuring the pilot, and knowing when to say stop

Once you have decided to run a pilot, measure four things and nothing more complicated than this: cycle time, how long the job takes start to finish; error rate, how often the output needs correction before a client sees it; revenue per enquiry, whether faster turnaround is converting more of the work you quote on; and cost per output, what each finished unit actually costs once the tool, the review time and the corrections are counted. Where advertising spend is the subject rather than production work, the same discipline applies through a return on ad spend calculation.

A pilot that is not improving at least two of those four measures within a defined window is not a pilot worth continuing, it is a subscription worth cancelling. The cost of a wrong decision here is rarely the software fee. It is the quiet damage of a client receiving a rushed or inaccurate piece of work while you were still deciding whether the tool was working. Knowing when to switch a tool off, cleanly and without embarrassment, is as valuable a skill as knowing when to adopt one. If a decision like this is sitting in front of you and the numbers are large enough to matter, it is worth a proper verdict rather than a guess, which is exactly what the Ideas Lab is there for.

Price for what the client receives, not for what it cost you

None of this requires a spreadsheet built by a consultant. It requires you to stop pricing from the cost side of the ledger and start pricing from the client's side of it. If AI has genuinely cut the hours on a job by 60 per cent, that is your business's gain to manage well, not a discount you owe anyone by default.

If you are the person who signs off this kind of decision and you want to work through your own pricing model with someone who has actually run a business rather than sold you a platform, that conversation is what CEO AI coaching is built for.