The unit economics of an AI employee, worked through

Cheap is not the same as value

A $99 per month subscription is easy to approve. It sits below most business credit card thresholds, it does not require a capital expenditure conversation, and the vendor demo makes it look like a straightforward win. So it gets signed up, it gets opened a few times, it gets mentioned at a team meeting, and three months later it is just another line on the software statement that nobody has cancelled because nobody has evaluated it.

The economic question that most Australian SMBs skip is simple: what does this tool need to displace before the spend is justified? The answer is not a feeling or a productivity score. It is a dollar figure, calculated from real loaded labour costs and real hours saved. Until you do that calculation, you are not managing an AI investment, you are running a subscription.

This piece works through the full model. Every number used is either an Australian benchmark or a scenario you can replace with your own inputs. By the end, you will have a formula you can apply to any AI tool your business is considering, and a clear view of which implementation patterns produce value and which ones quietly consume it.

Three ways AI creates value

Before running the numbers, it is worth being precise about what AI tools actually do to a business economically. There are three distinct mechanisms, and they compound differently.

Time displacement

The most direct mechanism. A task that took a staff member 2 hours now takes 20 minutes. The recovered hour and 40 minutes is either redirected to higher-value work or absorbed into capacity headroom that lets the business grow without adding a headcount. Time displacement is the mechanism most vendors lead with in their demos, and it is real when the tool is genuinely embedded in a workflow rather than used occasionally.

Time displacement is also the easiest to measure. You track how long a defined task takes before and after the tool is introduced, multiply the delta by your loaded hourly rate, and you have a dollar figure. Repeat that across every person and every task the tool touches and you have your annual gross saving.

Error reduction

Less visible but often more valuable. Manual processes accumulate errors: wrong figures in proposals, missed follow-up emails, inconsistent client reports, data entry mistakes that surface as invoice disputes three months later. Each error costs rework time, and some errors cost client relationships or compliance penalties.

For an Australian professional services firm, a single onboarding error that delays a client project by a week can represent $2,000 to $8,000 in delayed revenue or goodwill cost. An AI that reduces the error rate on a high-stakes process by 60 per cent is creating economic value that does not show up in hours-saved spreadsheets but absolutely shows up in the P&L over a 12-month period.

Capability expansion

The mechanism that has no direct human equivalent. A 12-person Sydney practice cannot employ a staff member to respond to website inquiries at 11pm on a Saturday. They also cannot manually personalise follow-up sequences for 400 leads per month at an individual level. An AI tool does both. That is not a cheaper version of something you were already doing; it is a capability your business did not have before.

Capability expansion is the hardest to quantify before implementation but often the highest-value outcome in practice. A health practice that starts capturing 30 per cent more weekend inquiries within a 2-hour response window is growing revenue that would not have existed without the tool. The ROI calculation for this mechanism is comparative: what would it cost to replicate this capability with human staff?

The break-even calculation, worked through

Start with the most common AI subscription tier: $99 per month, or $1,188 per year. The break-even question is: how many hours of staff time need to be saved per month for the tool to pay for itself?

Loaded staff cost in Australia varies by role, location, and business overhead, but a realistic working range is $35 to $80 per hour for SMB employees. The ATO standard benchmark for employer on-costs sits around 20 to 25 per cent above base wage for super and payroll tax alone. Add a proportional share of office, management, and software overhead and most roles land 35 to 55 per cent above the base wage.

At a conservative loaded rate of $50 per hour, the break-even calculation is direct:

$99 ÷ $50 = 1.98 hours per month.

That is the break-even threshold. Less than 2 hours of saved staff time per month covers the subscription. At $70 per hour (a mid-senior loaded rate for a coordinator or office manager in Sydney), the threshold drops further:

$99 ÷ $70 = 1.41 hours per month.

The bar is genuinely low. Most tools marketed to SMBs can clear it in the first week of use on a single defined task. The reason so many subscriptions fail to create value is not that the break-even is hard to reach. It is that the tool is never properly embedded in a workflow where it can save consistent hours.

Now scale up. A well-configured AI workflow assistant that saves 1 hour per day for one staff member is saving 20 hours per month. At $50 per hour loaded, that is $1,000 in recovered capacity per month, or $12,000 per year, from a $1,188 annual spend. Net annual value: $10,812. The ROI is approximately 910 per cent. The tool that looked like a nice-to-have at $99 per month is returning nearly 10 times its cost in recovered labour.

The three-variable model

The full model has three inputs. Once you have these, you can calculate net annual value for any AI tool against any role in your business.

Net annual value = (Hours saved per week × Loaded cost per hour × 52) − Annual tool cost

For a $99 per month tool, annual tool cost is $1,188. At a loaded rate of $50 per hour, the three scenarios break down as follows.

Even the Low scenario produces a positive return. One hour per week of genuine, consistent time saving across a process that used to be done manually returns $1,412 net per year. That is not transformative, but it is positive, and it compounds as usage matures and the team gets faster with the tool.

The Mid scenario at three hours per week is where most well-implemented tools land inside the first six months. Three hours per week is realistic for any AI tool embedded in a high-frequency process: daily reporting, client follow-up, content drafting, or document summarisation. At $6,612 net per year, this is a strong business case that clears most SMB investment thresholds.

The High scenario at five hours per week is achievable for tools that touch team-wide workflows rather than individual tasks. A practice management AI that automates appointment confirmations, intake document collection, and post-session notes for five practitioners simultaneously is realistically in this range. At $11,812 net per year from a $99 per month spend, the economic case is unambiguous.

Apply these numbers to your own business by replacing the loaded rate with your actual figure and the hours with a conservative estimate based on the specific process the tool will handle. The conservative estimate is important: do not use optimistic vendor projections for the hours input. Use what you actually observe in the first 30 days.

The scenarios that do not work

The break-even model is also useful for identifying implementations that will not pay off, before you commit three months to building them. There are three patterns that consistently produce negative or near-zero ROI in the SMB context.

Low-rate role, marginal time saving

Consider a $99 per month tool deployed to automate a task performed by an admin coordinator on a base salary of $55,000. Loaded rate is approximately $38 to $42 per hour. The tool saves 15 minutes per week on a data entry task. Annual gross saving: 0.25 hours × $40 × 52 = $520. Annual tool cost: $1,188. Net annual value: negative $668.

This is not a failure of the tool. It is a failure of scoping. The tool may be excellent at what it does; the problem is that the process it is handling has insufficient volume and insufficient labour cost to justify the subscription. The fix is either to expand the tool to handle a higher-volume process for the same role, or to redirect the subscription to a process where the loaded rate and hours saved are materially higher.

Single-user dependency without succession planning

A tool that lives in the workflow of one person and is not documented, not embedded in team process, and not transferable is a high-risk spend. When that person leaves, the tool either gets cancelled with zero institutional knowledge transfer, or it gets kept open by a replacement who does not know how to use it and reverts to the manual process anyway.

This is especially common in Australian SMBs where AI adoption is driven by one digitally-fluent team member. The fix is to treat AI tool onboarding like any other business process: documented, tested, and not dependent on a single person to function. If you cannot describe how the tool fits into the workflow in writing, the tool is not yet an asset, it is a personal habit.

Configuration overhead that consumes the savings

Some AI tools require significant ongoing configuration: prompt tuning, integration maintenance, output review cycles, and regular retraining on new data. If that configuration work takes 4 hours per month and your loaded rate is $60 per hour, you are spending $240 per month in labour to manage a $99 per month tool. Total monthly cost: $339. That changes your break-even from 1.98 hours to 6.8 hours of saved time per month.

Configuration overhead is the most common silent cost in AI tool implementations and the one most frequently omitted from initial business cases. Before signing up, ask the vendor for a realistic estimate of weekly maintenance time for a business your size. If they cannot give you a specific answer, assume 2 to 3 hours per month and price that into your model before you decide.

What Bizkook clients actually see

Across our Sydney consulting engagements with Australian SMBs between 5 and 50 staff, the range of recovered capacity we see in the first 6 to 12 months after a properly scoped AI implementation falls between 3 and 12 hours per week. At typical loaded rates, that translates to $500 to $2,500 per month in capacity value per implementation.

Professional services firms, which include accounting practices, legal firms, marketing agencies, and management consultancies, tend to see the higher end of that range. Their workflows are document-heavy, client-facing, and highly repeatable. A well-configured client communication and reporting workflow in a 15-person accounting practice routinely saves 8 to 10 hours per week across two or three senior staff, producing $1,600 to $2,800 per month in recovered capacity from tools that cost $150 to $400 per month in subscriptions.

Allied health practices see strong results in the 4 to 8 hour per week range. The highest-value implementations are appointment management and clinical documentation support, where AI drafts structured session notes from audio or dictation and the practitioner reviews and signs off. A psychologist or occupational therapist who spends 90 minutes per day on notes and reduces that to 30 minutes is recovering 2.5 hours per day, or 12.5 hours per week. At a loaded rate of $65 to $80 per hour for a registered practitioner, that is $800 to $1,000 per week in recovered capacity, from a tool that typically costs $200 to $350 per month.

The pattern across both sectors is consistent: the best returns come from high-frequency, high-stakes, document-heavy processes where the staff doing the work have a loaded rate above $50 per hour. The worst returns come from low-frequency, low-stakes, low-rate processes where the tool is solving a problem that was not expensive in the first place.

What to do next

Run the three-variable model on every AI tool you are currently paying for, not just the ones you are considering. Pull your loaded hourly rates, estimate hours saved honestly using actual observed time rather than vendor projections, and calculate net annual value for each subscription. You may find that one or two tools are producing excellent returns and two or three others are marginal or negative.

For tools you have not yet committed to, use the break-even calculation before you sign up. It takes five minutes and prevents the most common failure mode in SMB AI adoption: subscribing to tools that solve small problems at a rate that does not justify the spend or the implementation time.

If you are deciding which function to automate first rather than which specific tool to buy, read which AI employee to hire first for the cost-of-delay sequencing method that ranks your six business functions before you spend anything. For business owners managing variable revenue, cash flow forecasting for lumpy revenue covers the financial discipline that underpins any AI investment decision.

Our AI consulting engagements begin with the ROI model for every tool under consideration. It is the fastest way to separate genuine business cases from subscriptions dressed up as strategy.