The returns from AI come from one person owning the decisions and the build. Here's what that role does, what it's worth, and how to get it affordably.
Most AI conversations start with a tool pitch and end with a subscription. While the businesses who are actually getting something out of AI may have started there, they didn’t end up there. They found a person, someone who could look at how work moves through the business, pick the two or three places where a machine would genuinely help, created and led a plan to build those, and left everything else alone.
The gap between those two groups is wide. AI Lab Australia’s January 2026 report, which models 500 Australian enterprises using Deloitte, Tech Council, ABS, and National AI Centre data, sorts adopters into three tiers. Around 60% are basic users saving one to two hours a week on drafting and admin. Roughly 35% reach an intermediate level, which the report associates with a 45% profitability uplift. About 5% are fully enabled, where that figure is 111%. Those are modelled numbers rather than audited results, so hold the exact percentages loosely. The shape is what matters, and the shape is stark. Every tier has access to the same tools. What separates them is whether somebody in the business owned the job of putting those tools to work.
What that person actually does
Owning AI is more specific than it sounds. It means deciding which problems are worth automating and, more usefully, which ones aren’t. It means knowing what the current tools can and can’t do this month, which changes faster than most teams can track. It means building the thing rather than writing a paper about it, checking the output before it reaches a customer, and watching what breaks in the weeks after launch, which is where most AI projects quietly stall. The Goldman Sachs 10,000 Small Businesses survey of 1,256 US owners in early 2026 shows how much sits on that role. 84% named increased efficiency and productivity as their main benefit from AI, and 87% said it augments their people rather than displacing them, so the appetite isn’t the issue. Only 14% had AI built into core operations. The two blockers they named most often were a lack of technical expertise (49%) and difficulty choosing the right tools (48%). Both of those are one person’s job description.
What the role is worth
A great example is a management consulting firm we worked with who have 60 staff, 38 of them billing. Fee earners were losing about 5 hours a week each to file notes, scoping documents, and chasing information from clients. A capable AI lead doesn’t remove all of that. Say they recover 3 hours of it with a system that drafts file notes from call recordings and pre-fills scoping documents from prior matters, with a person approving each one before it goes anywhere. Across 38 people that’s 114 hours a week. If only a third of it converts into billable work at $220 an hour, the firm is $8,300 a week better off, or roughly $380,000 across a year. Set that against the $250,000 a good AI lead costs in the current market and the recovered time covers a full year of the role inside the first couple of months of running. And thats with one task and one implementation, but these sort of results are happening week in week out across similar sized businesses.
Somthing worth noting is that it depends on a couple of things holding. The work has to be aimed at a real bottleneck, because speeding up a step that was never the constraint just moves the queue. And the recovered hours have to go somewhere that earns. Time that comes back and doesn’t get used isn’t a saving, it’s a quieter week.
There’s a second return that’s harder to see. Nearly half of the businesses in the Goldman survey said choosing the right tools was a blocker, and the cost of choosing wrong is real money in abandoned pilots, overlapping subscriptions, and work rebuilt twice. Someone who has done this before spends a fraction of that, mostly by saying no early.
Most businesses can’t carry that hire
Even at those returns, $250,000 a year is a lot for a 60 person firm to commit to, and there usually isn’t a full year of AI work to keep that person occupied once the first few builds land. Promoting from inside is harder than it looks too. AI Lab Australia puts more than half the SMB workforce at basic or novice AI literacy with only 10% advanced, and someone growing into the role learns on your live systems and your money.
What you need is the judgement and the build, not the desk. That’s what our Technology Partner service is for. You get an allocated block of expert hours each month, every AI decision and build sits with that one role, and the ongoing running costs of what we put in are rolled into the same fixed monthly price so the number is known before the month starts.
One warning if you’re comparing providers. Plenty of people will tell you they know AI on the strength of a few proofs of concept and a weekend project. Ask what they’ve put into production in the last six months and who is using it today. The report also notes that growing SMBs are 1.8 times more likely to invest in AI than declining ones, which is worth remembering when this gets treated as a discretionary line.
What this means for you
If nobody in your business owns AI, the results you’re getting are the basic tier results, an hour or two a week and not much else, no matter which tools you’ve bought. The fix is a person with the judgement to pick the right problem and the ability to build the answer, and you can buy that by the month rather than by the year.
Start with our AI Roadmap Interview. We’ll walk through where your time and margin actually go, then show you the first build worth doing and what it should return.
