Economics matter, there is no hiding it, large-scale AI inference is expensive
Thought Leadership by Andrew Stanton Analyst & Consultant – Proptech-PR
For the last two decades, the software industry has been built around a predictable commercial model Software-as-a-Service (SaaS). Companies paid a fixed monthly or annual licence fee per user, gained access to a cloud platform, and scaled usage gradually as their business grew.
The economics were simple, a proptech platform might charge an estate agency £X per branch or £Y per user, or £Z per month for enterprise access. Whether a user logged in ten times a day or once a week, the provider’s infrastructure costs remained stable. Cloud hosting, storage, and support costs were manageable and predictable. With gross margins often exceeding 70–80%.
Artificial intelligence is now disrupting that model entirely
The next generation of proptech platforms will not simply provide software tools. They will deliver outcomes. AI systems will negotiate with landlords, qualify leads, write listings, answer tenant queries, manage maintenance workflows, analyse portfolios, produce valuations, chase arrears, and potentially run entire operational processes autonomously through agentic workflows.
But unlike SaaS, AI does not scale cheaply, and that changes everything.
From access to software to consumption of intelligence
Traditional SaaS charges customers for access but AI services increasingly charge providers for computation. Every AI-generated email, valuation report, maintenance summary, tenant interaction, or sales negotiation consumes tokens — the underlying units of computation used by large language models.
This introduces a fundamentally different cost structure. A CRM platform can serve thousands of users with low incremental cost. But an AI agent managing complex workflows continuously may generate millions, or even billions of tokens per customer every month.
In practical terms, a SaaS platform might cost a provider pennies per user per day. An AI-driven operational platform could cost pounds, tens of pounds, or eventually hundreds of pounds per user depending on how autonomous and sophisticated the workflows become.
The more value AI delivers, the more compute it consumes. That creates a paradox for the Property/proptech industry. The more intelligent the system becomes, the more expensive it becomes to operate.
Agentic workflows will multiply consumption
Today, most businesses use AI intermittently. A user opens ChatGPT, asks a question, receives an answer, and leaves. That is not the future being built. No, the future is agentic.
Agentic systems do not wait for prompts. They continuously execute tasks, communicate with other systems, make decisions, trigger workflows, retrieve data, generate outputs, and monitor environments autonomously.
In a proptech context, an agentic workflow might monitor inbound property enquiries, qualify leads automatically, book viewings, generate follow-up emails, update CRM records, analyse tenant affordability, produce compliance documents, chase outstanding paperwork, recommend price adjustments. And communicate with landlords, coordinate contractors, produce management summaries and forecast portfolio performance.
And all of this happening 24 hours a day, and every single action consumes compute. Every API call, reasoning cycle, retrieval process, memory operation, and multi-agent interaction increases token expenditure.
A traditional SaaS platform might process database transactions cheaply. In contrast an AI-native platform may perform thousands of reasoning operations for what previously required one button click.
The hidden economic problem of AI
Much of the current AI boom is being subsidised. Investors are funding enormous infrastructure expansion while many AI providers operate at thin margins or losses to accelerate adoption.
But over time, economics matter, there is no hiding it, large-scale AI inference (the process of using a trained learning machine model to make predictions, decisions or generate content from new unseen data) is expensive. Running advanced models continuously across enterprise workflows requires GPU infrastructure, high-performance networking, vector databases, memory systems, multi-model orchestration, continuous fine-tuning, monitoring and safety layers, context management, and long-running agent processes.
This is not comparable to hosting a conventional CRM or property management platform, the cost base is dramatically higher. As AI agents become embedded into operational workflows, providers will eventually need to pass those costs onto customers. That may fundamentally reshape software pricing.
Stanton’s Analysis
Why proptech could become one of the most expensive AI verticals
Proptech is particularly exposed because property operations involve enormous amounts of communication, coordination, documentation, and repetitive process management. These are precisely the tasks AI excels at.
But they are also tasks that generate continuous computational load, consider a medium-sized estate agency with forty staff, 10,000 monthly tenant interactions, 2,000 property enquiries, Hundreds of maintenance tickets, ongoing landlord communications, automated compliance monitoring, and you guessed it AI-driven marketing workflows.
If every operational layer becomes AI-assisted or AI-managed, token consumption scales aggressively. We are talking big figures, now extend that to build-to-rent operators, housing associations, commercial real estate firms, global brokerage networks, and institutional portfolio managers, then the compute requirements become enormous.
Unlike SaaS, where additional users barely move infrastructure costs, AI-native workflows scale proportionally with usage intensity. This creates a future where businesses may no longer pay £200 per month for software access but instead pay £2,000–£20,000 per month for autonomous operational intelligence. Especially if multiple AI agents operate simultaneously across departments.
Will we see the return of usage-based pricing?
Ironically, AI may reverse one of SaaS’s greatest commercial advantages: predictable pricing. Many AI companies are already experimenting with token-based pricing, consumption billing, outcome pricing, agent execution fees, workflow metering, and compute-tier subscriptions.
This resembles utility billing more than software licensing. In the future, firms may receive invoices based not on the number of users but on the number of autonomous actions, AI processing hours, reasoning depth, agent concurrency, memory usage, workflow complexity, and the big kicker token consumption
Businesses adopting heavy agentic automation may discover they are effectively employing digital workforces that require continuous computational energy to operate. The software industry may slowly evolve from selling products to selling machine labour.
The economic divide that could emerge
This creates a potentially significant divide in the market. Large enterprises may benefit enormously from AI automation because replacing large operational teams with AI agents still produces net savings despite higher software costs. But smaller agencies face pressure.
A traditional independent estate agency paying £Z per month for SaaS tools may struggle if fully AI-native operational platforms eventually cost several multiples more, up to 8 or 10 times.
Some firms may adopt partial automation; others may operate hybrid models where human staff manage tasks that remain too computationally expensive to automate continuously. The assumption that AI will always make software cheaper is unlikely to hold true at enterprise scale.
In many cases, AI may produce dramatically better outcomes while simultaneously becoming significantly more expensive than legacy SaaS.
The real question is whether the outcomes justify the cost
The critical issue is not whether AI costs more, the real question is whether businesses receive disproportionately greater value in return. The Holy Grail hockey stick return that business owners and operators dream of.
In the cold light of day if an AI platform replaces five staff members, operates 24/7 at a higher level of service, improves lead conversion, reduces void periods, automates compliance, increases portfolio efficiency, and enhances tenant satisfaction, then even a much larger monthly bill may still represent exceptional ROI.
It is too early to judge but is the property/proptech world now transitioning. Moving from software tools that assist humans toward autonomous systems that increasingly perform the work itself. And when businesses stop paying for access to software and start paying for digital labour, the economics of technology fundamentally change.
The SaaS era was built on cheap scalability. Will the AI-agent era be built on expensive intelligence.
PROPTECH-PR is both a consultancy for Proptech Founders and an Advisory Service for the wider Proptech & Real Estate Industries, if you are running low on ideas, strategy or revenue, or just plain confused – please book an exploratory personal call to see if I can help. Using this LINK.
