For the last fifteen years, the 'Per-User, Per-Month' model has been the undisputed king of business software. You hired a new employee, you bought them a Slack seat, a Microsoft 365 license, and access to your CRM. It was predictable, scalable, and—honestly—a bit of a trap. As someone who runs an entirely AI-first business, I’ve seen the cracks in this model widening. AI implementation for small business isn't just about replacing a human task with a bot; it’s about fundamentally dismantling the way we pay for productivity.
We are moving from an era of 'Software as a Service' (SaaS) to an era of 'Software as a Result.' In the old world, you paid for access. In the new world, you pay for outcomes. This shift from fixed-cost subscriptions to pay-as-you-go tokenized models is the single biggest change to SME cash flow since the move to the cloud. If you're still budgeting for seats in 2024, you're likely paying what I call the Shelfware Tax—and it’s costing you more than you think.
The Death of the Digital Landlord
💡 Want Penny to analyse your business? She maps which roles AI can replace and builds a phased plan. Start your free trial →
Traditional SaaS companies are digital landlords. They rent you space (seats) regardless of whether you’re actually 'living' in them. If you have 20 employees, you pay for 20 seats. It doesn't matter if five of those employees only log in once a month, or if ten of them are only using 5% of the software’s capability. You pay for the potential of use, not the value of the result.
AI-driven tools are different. Most modern AI models operate on a 'token' or 'usage' basis. You pay for the words generated, the images created, the data analysed, or the tickets resolved. This is a fundamental shift toward Utility-Based Operations. Much like your electricity bill, if you don't turn the lights on, you don't pay.
When I look at IT support costs for most SMEs, the inefficiency is staggering. You might be paying £50 per user per month for a support desk that only handles three queries for that user all year. An AI-first approach reverses this: you pay pence for the queries handled and zero for the quiet periods. This is where the lean business wins.
The Shelfware Tax: Why Subscriptions Stall Growth
I’ve spent a lot of time looking at the balance sheets of businesses trying to scale. A recurring pattern I see is what I’ve named The Shelfware Tax. It’s the delta between the total cost of your software subscriptions and the actual utility your team derives from them.
In a traditional SaaS model, the software provider is incentivised to make their product 'sticky' but not necessarily 'efficient.' The more people you have using the tool, the more they make. AI flips this incentive. If an AI tool can solve a problem in 30 seconds of compute time, it costs you less than if it takes 30 minutes. The incentive shifts toward radical efficiency.
Consider your HR software costs. Often, you’re paying a 'per-employee' fee for a suite of tools you might only use during the hiring window or annual reviews. In a pay-as-you-go AI economy, you would only incur significant costs when the AI is actually performing talent acquisition, screening, or onboarding tasks. During 'business as usual' months, your software overhead drops to near zero.
The Framework: The Variable Utility Matrix
To understand where to apply AI implementation for small business, you need to look at your operations through what I call the Variable Utility Matrix. This helps you decide which functions should remain on a subscription and which should migrate to a pay-as-you-go AI model.
- High Frequency / Low Complexity (The 'Token' Zone): These are tasks like customer service FAQs, basic data entry, and first-level IT triage. These should be moved to pay-as-you-go AI immediately. Why pay for a 24/7 human seat (or a per-seat SaaS tool) when you can pay per resolved ticket?
- Low Frequency / High Complexity (The 'Advisory' Zone): Strategic planning, high-level creative, and complex legal work. These still benefit from human oversight or specialized SaaS tools. Even here, however, compare Penny vs Quickbooks to see how an AI-first advisor can provide deeper value than a static ledger for a fraction of the cost.
- High Frequency / High Complexity (The 'Co-Pilot' Zone): Software development or deep research. This is where 'Per-Seat' still makes sense for now—giving your best people the best tools to increase their output.
The 90/10 Rule and the Reshaping of Roles
One of the second-order effects of moving to a pay-as-you-go AI model is the exposure of the 90/10 Rule. When AI can handle 90% of a function (like bookkeeping or basic copy drafting), the remaining 10% rarely justifies a standalone role or a premium per-seat software license.
When you stop paying for 'seats' and start paying for 'tokens,' you begin to see your business as a series of workflows rather than a collection of departments. You don't have a 'Marketing Department' cost; you have a 'Content Generation' utility cost. You don't have a 'Customer Service' overhead; you have a 'Query Resolution' variable cost.
This makes your business incredibly resilient. In a downturn, your software costs automatically scale down with your volume. In a boom, you aren't hit with massive 'pro' tier subscription jumps just because you added three more users.
Moving from Budget Approval to Threshold Monitoring
The biggest challenge in this transition isn't the technology—it's the finance department. Traditional budgeting loves the 'Fixed Monthly Cost.' It’s easy to put in a spreadsheet. Pay-as-you-go AI is variable, which can feel like a lack of control.
However, the control actually increases. Instead of 'approving a £2,000/month subscription,' you set 'threshold alerts' for your API usage. You can see, in real-time, which workflows are delivering value and which are burning tokens unnecessarily. This is Financial Granularity that SaaS never provided.
How to Start Your Migration
If you're feeling overwhelmed by the transition, don't try to rip out every subscription at once. Start with the 'Token Zone.'
- Audit your 'Ghost Seats': Look for SaaS subscriptions where 20% of the users are doing 80% of the work. Can those 80% be served by a pay-as-you-go AI interface instead?
- Look for 'Consumption-Based' Alternatives: When choosing new tools, prioritize those that offer a usage-based tier. It’s a sign the company is confident in the efficiency of their AI.
- Embrace the Lean Team: As you move to pay-as-you-go, you'll find you can run a much larger business with a much smaller core team. I am proof of this. My 'team' is a collection of optimized AI workflows, and my 'payroll' is essentially an API bill.
The window for this transformation is closing. The businesses that move from 'paying for seats' to 'paying for outcomes' will have the cash flow to out-invest their competitors. They will be leaner, faster, and more profitable.
Are you still paying for potential, or are you ready to start paying for results?