Is AI expensive or are you budgeting wrong?

This article originally appeared in the 6th edition of Unlocked by GSL.

An operator I spoke with recently flinched at a €200 monthly AI subscription. The same week, they signed off on a €70,000 hire without a second thought. That reflex, comfortable with payroll and suspicious of software, is the most expensive mistake in real estate technology right now.

The problem isn't the price. It's the budget line.

For the past two decades, SaaS has taught us to think of software as a predictable IT expense: a monthly subscription that scales with headcount. AI doesn't behave like that, and it shouldn't be measured like that. Benchmarked against Microsoft 365, a €300 subscription looks indulgent. Benchmarked against the junior analyst it replaces, it's the cheapest hire you'll ever make.

In 2026, the operators who get this right will be the ones with budgets that reflect what AI actually replaces. They will book AI under headcount, where it belongs.

The category error

Software used to be a tool you gave a person so they could do their job. The person was the unit of work, the software supported them, and pricing was flat and per seat.

AI inverts that. The person shifts from using the tool to directing it. Output that used to take a junior analyst three days now takes a senior operator ninety minutes of prompting, reviewing, and refining. The software is doing a meaningful share of the job itself.

That breaks two assumptions every finance team still works from. Consumption replaces seats: two employees with the same job title can consume wildly different amounts of AI, and that's a signal of leverage rather than a budgeting problem. Output replaces access: AI sells throughput, not the licence to log in. The right comparison is no longer "what does this tool cost compared to other tools," but "what would this output cost if a person produced it."

The €200 teammate

Make it concrete. A power user at our company runs €200 to €400 a month, all in: a Claude Max plan, an IDE like Cursor, a few specialised tools, and the tokens consumed by an MCP server connecting it all.

In eighteen months, that person's output on the right tasks has gone from doubled to quadrupled. Not across the board, not magic, but real and measurable.

Now the alternative. A junior hire in most European markets runs €5,000 to €7,000 a month in loaded cost: salary, social contributions, holiday, equipment, onboarding. Add a three-month ramp and the cost of attrition, and an early-career hire sits north of €70,000 in year one. The AI teammate costs roughly three percent of that. Productive on day one, doesn't churn, doesn't need re-onboarding. It won't replace the junior entirely, but it absorbs enough work that one experienced person now operates at the capacity that previously required two.

This is why the right reference class is the organisational chart and not the IT stack.

Why this belongs in HR

If AI replaces tools, IT owns the budget. If AI replaces output, HR owns it. The conversation has moved from licences to productivity, capacity planning, and compensation.

Three shifts follow. First, give every employee a personal AI budget of €200 to €500 a month, depending on role. Poor budgets kill experimentation. Individual stipends enable employees to take ownership and figure out how to use it.The cost of one person underusing it is trivial. The cost of one person discovering a workflow that compounds across the team is enormous.

Second, hire for AI fluency, the new Excel. Ask candidates how they use AI today, what they've stopped doing because of it, and what they've started doing that wasn't possible before. The answers separate candidates from those who are open to exploration and those who are not.

Third, pay for leverage. If two property managers do the same job and one ships three times the output because they're AI-native, that should show up in pay, not as a bonus for using a tool but as recognition they're operating at a different level. The operators that work this out first will win the talent fight in 2027.

The honest costs

None of this is free. Token costs are variable, so budget for a range and treat spikes as a leading indicator rather than a problem. AI output also needs an editor. Bad output flowing unchecked, whether it’s resident communication or financial reporting, will end up backfiring.

Governance is the first genuinely new line item. The EU AI Act's first wave of obligations lands in August, and any operator using AI in tenant-facing workflows, hiring decisions, or automated communications will need classification, documentation, and training records on file. None of this is expensive in isolation, a few thousand euros a year for most mid-sized operators, but it's a cost that didn't exist eighteen months ago, and discovering it in an audit costs ten times more than budgeting for it now. Treat it like insurance: small, recurring, non-negotiable.

The line item

A subscription that costs less than a team lunch can do the work of a hire that costs more than a company car. As a real estate operator, internalise this in your company strategy and your competitive advantage will compound.

The right question isn't whether AI is worth the spend, but whether your budget structure has caught up to what you're actually buying. For most, it hasn't. The tools are right. The line is wrong.

Move it.