A 10-person commercial real estate firm getting serious about AI for the first time should plan to spend somewhere between $6,000 and $60,000 in its first full year, and the size of that number is almost entirely a choice about scope, not a fixed price. The floor — around $6,000 to $12,000 — buys per-seat AI subscriptions for the people who will actually use them plus one hands-on training session to make them fluent. The ceiling climbs when you add a custom automation project on top, at which point a single build can run $25,000 to $90,000 and pull your annual figure well past $60,000. The mistake most principals make is treating “AI budget” as one line item to size against a benchmark. It is three separate decisions, each with its own price and its own payback, and getting the order right matters more than getting the total right.
The one-number answer, and why it is a range
If you need a single planning figure to put in front of a partner, use $15,000 to $40,000 for a firm of your size that wants meaningful progress this year without betting the budget on a custom build. That range covers AI subscriptions for the whole team, a proper training investment to get real use out of them, and enough set-aside to run one small, scoped automation pilot if a clear bottleneck justifies it.
The reason it is a range and not a price is that the three things you are buying scale independently. Subscriptions are a few thousand dollars a year and barely move. Training is a one-time cost measured in low thousands. A custom automation project is the variable that swings your total by an order of magnitude — skip it and you are near the floor; commit to a production build and you are near the ceiling. So before you size the budget, decide which of the three you are actually funding this year. The full decision logic for whether an off-the-shelf tool or a custom build fits your situation is laid out in our buy-versus-build playbook for small CRE firms; this piece prices the outcome of that decision.
The three buckets your AI budget splits into
Every AI dollar a small CRE firm spends lands in one of three buckets. Naming them separately is the whole trick, because it stops you from either underspending on the cheap thing that creates the most value (fluency) or overspending on the expensive thing before you are ready for it (a custom build).
| Bucket | What it buys | Annual cost for a 10-person firm | Recurring or one-time |
|---|---|---|---|
| 1. Per-seat subscriptions | ChatGPT, Claude, or Microsoft Copilot licenses; a proptech AI add-on | $1,500–$5,000 | Recurring |
| 2. Fluency and training | A hands-on workshop that teaches staff to use the tools on real CRE work | $2,000–$15,000 | Mostly one-time |
| 3. Custom automation | A built workflow — lease abstraction, deal screening, back-office reporting | $25,000–$90,000 per project | One-time build + ~15–20% yearly upkeep |
The buckets are ordered deliberately. Bucket 1 is table stakes and cheap. Bucket 2 is the highest-return spend a small firm can make, because tools nobody knows how to use return nothing. Bucket 3 is powerful but optional, and it should never be the first check you write.
Bucket 1: Per-seat AI subscriptions
This is the baseline layer, and for a firm your size it is genuinely inexpensive. A business-tier seat of ChatGPT, Claude, or Microsoft Copilot runs roughly $25 to $30 per user per month. You do not need to buy a seat for all ten people on day one — the pattern that works is licensing your four to six power users first: the brokers writing offers, the analyst doing market write-ups, the ops person handling lease summaries and tenant email.
At six seats and $30 a month, that is about $2,160 a year. Add a proptech tool with AI features baked in — a CoStar or Crexi subscription you may already carry, or a listing-marketing add-on through Buildout — and the AI-attributable slice of that bucket lands between $1,500 and $5,000 annually. For context on what “normal” looks like, firms of 10 to 50 people already spend $400 to $600 per employee per month across all their software (Cledara). A few AI seats are a rounding error against that, which is precisely why this bucket is never where a small firm should hesitate.
One caution specific to your world: if you handle confidential deal data, buy business or enterprise tiers, not the free consumer versions. The paid tiers keep your prompts out of model training by default, which matters when the “document” you paste is a client’s rent roll or a purchase agreement.
Bucket 2: Making the team fluent
This is the bucket firms underfund, and it is the one that decides whether the other two return anything. A subscription that sits unused is pure waste, and the default outcome for a busy brokerage is exactly that — a seat gets bought, used twice, and forgotten.
A hands-on fluency workshop closes that gap. Market rates for a focused team training run roughly $2,000 to $15,000 depending on length and depth, and for a 10-person firm a single well-run session sits comfortably in the lower half of that band. The content that matters for CRE is narrow and practical: how to prompt an assistant to draft a letter of intent from your terms, summarize a 60-page lease into the fields you care about, turn comps into a market write-up, and triage a full inbox — the daily writing and reading work that eats a broker’s afternoon. Getting a whole team to competence on those tasks in a single sitting is the highest-return line in the entire budget. We make the case for training-before-tooling in full in our look at sequencing AI spend at a small firm.
Treat this as mostly a one-time cost with a small annual refresh. You train the team once, then budget a few hundred dollars and a couple of hours a year to bring new hires up to speed and cover what has changed.
Bucket 3: One automation project (optional in year one)
This is the bucket that turns a five-figure budget into a six-figure one, and it is optional — many firms should not fund it in their first year. A custom automation project has SFAI or another builder construct a workflow that runs without a person babysitting it: pulling structured data out of your lease stack, screening and ranking inbound deals from broker blasts, or assembling CAM and investor reports from your ledger.
A scoped project for a small firm runs $25,000 to $90,000 for the build, with the pilot phase at the low end and a full production workflow at the high end. On top of the build, budget roughly 15 to 20 percent of the project cost per year for maintenance, because document formats drift, integrations break when a vendor ships an update, and the underlying models change. A $45,000 build therefore carries something like $7,000 to $9,000 a year in ongoing upkeep — a real number that belongs in your annual plan, not a surprise in month twelve.
The discipline that keeps this bucket from becoming a mistake is to fund it only against a bottleneck you can measure. If your analyst spends two days a month abstracting leases, or a broker loses deals because inbound listings pile up unread, the math is easy to run and a build can pay for itself. If you cannot name the specific hours it saves, do not build yet. The question of when to stop renting a proptech tool and build your own is worked through in our guide on when to fire a proptech vendor and build in-house.
Three budget tiers for a 10-person firm
Here is the whole model as three tiers you can adopt directly. Pick the one that matches your appetite and your evidence, not the one that matches a competitor’s press release.
| Tier | What it includes | Annual total | Right for |
|---|---|---|---|
| Starter | 4–6 AI seats + one fluency workshop; no custom build | $6,000–$12,000 | A firm proving the value of AI on daily work before committing capital |
| Standard | Firm-wide seats + workshop + a scoped automation pilot | $15,000–$40,000 | A firm with one clear, measurable bottleneck worth testing a build against |
| Ambitious | Seats + training + a full production automation in one domain, with maintenance | $40,000–$90,000+ | A firm that has already validated a workflow and wants it running in production |
Most 10-person firms should start at Starter or Standard. The Starter tier is not a lesser choice — it is the correct choice for a firm that has not yet made its team fluent, because spending on a custom build before your staff can use a chatbot is building the second floor before the first. The Ambitious tier earns its number only after a pilot has produced a real result on your own files. A lean firm’s structural advantage is how fast it can adopt a new tool once it decides to, an argument we develop in the small CRE firm AI manifesto.
What the benchmarks actually say
Outside numbers are useful as a sanity check, less so as a target. The most-cited 2026 benchmark puts average AI spend at about $2,068 per employee — but that average hides a wide split: the median company spends under $200 per employee, while the top 10 percent spend $2,800 or more (Rize). At ten employees, the average implies roughly $20,000 a year, which lands squarely in the Standard tier above. Small businesses overall average around $18,000 annually on AI tools and subscriptions, with subscriptions making up roughly half of that spend (Preferred Data).
Real estate specifically underspends relative to other sectors, and the reason is instructive. Deloitte’s 2026 Commercial Real Estate Outlook, drawn from more than 850 executives, found 81 percent naming data and technology as their top area of spending focus, yet only 7 percent reported a transformative impact from AI so far (Deloitte). JLL’s 2025 technology survey is starker: 88 percent of investors are running AI pilots, 87 percent are raising technology budgets for AI, but only about 5 percent report achieving most of their goals, and more than 60 percent describe themselves as unprepared to execute (JLL, via Propmodo). The gap between money spent and value captured is not about spending more. It is about sequencing — which is the whole point of the three buckets.
How to sequence the spend so you do not waste it
The order in which you release the budget matters more than the total. A firm that spends $40,000 in the wrong order gets less than a firm that spends $12,000 in the right one. The sequence that works:
- Buy the seats first, this month. A few thousand dollars, immediate. Give your power users paid, business-tier access to a capable assistant.
- Train within the same quarter. Do not let two months pass between buying tools and teaching people to use them, or the tools go cold. The workshop is what converts Bucket 1 from a cost into a capability.
- Run for a full quarter before considering a build. Let the team use AI on real work long enough to surface where a chatbot stops being enough and a real automation would pay. That lived experience is what tells you whether Bucket 3 is justified — and where.
- Fund one build, not three. If a bottleneck earns a custom project, scope a single workflow, run a pilot before the full build, and measure the result on your own data before writing the production check.
This is why the JLL and Deloitte numbers look the way they do: firms that spend on Bucket 3 before Bucket 2 end up in the 60 percent that feel unprepared, because they bought a system nobody was ready to run.
When to spend nothing new this year
An honest budget guide has to include the case for zero incremental spend, because for some firms it is the right call. If your team is not yet using the AI tools you already pay for — a Microsoft 365 subscription that includes Copilot, or a proptech platform that quietly shipped AI features last quarter — the correct move this year is not to buy more. It is to make the team fluent on what you have and measure what that alone changes.
Fluency training against your existing stack can run a few thousand dollars and unlock capability you are already paying for. Only once your staff is genuinely using those tools, and can point to a specific task where a general-purpose assistant runs out of road, does new spending — a dedicated proptech subscription or a custom build — become a defensible line. Buying software to solve an adoption problem is the most common way a small CRE firm wastes an AI budget. The current landscape of what small CRE firms are actually adopting, and where the value is landing, is surveyed in our state of AI in proptech for small firms.
FAQ
How much should a 10-person CRE firm budget for AI per year?
Plan for $15,000 to $40,000 for meaningful progress in a first serious year: AI subscriptions for the team, a fluency workshop, and a set-aside for one scoped automation pilot. A leaner path — seats plus training with no custom build — runs $6,000 to $12,000. If you commit to a full production automation, a single build of $25,000 to $90,000 pushes your annual total past $60,000.
What is a realistic starter AI budget for a small brokerage?
About $6,000 to $12,000. That funds business-tier AI subscriptions for your four to six power users (roughly $2,000 a year) and one hands-on training session (low thousands) to get real use out of them. It deliberately excludes a custom build, which most firms should not fund until their team is fluent and a specific bottleneck justifies it.
How much of the budget should go to tools versus training versus custom builds?
If you are not funding a build this year, weight it toward training: subscriptions are cheap and fixed, so the training bucket is where the return lives. A typical split without a build is roughly one-third subscriptions, two-thirds training. Once you add a custom automation project, the build dwarfs both — $25,000 to $90,000 against a few thousand each — so it becomes its own decision sized against a measured bottleneck.
How much do AI subscriptions cost per employee for a CRE firm?
Business-tier seats of ChatGPT, Claude, or Microsoft Copilot run about $25 to $30 per user per month, or $300 to $360 a year per person. You do not need seats for everyone at once — license the four to six people doing the most writing and document work first, which keeps the whole subscription bucket to $1,500 to $5,000 a year for a 10-person firm.
Should a small firm budget for a custom automation project in year one?
Usually not. A custom build should be funded only against a bottleneck you can measure in hours or lost deals — repetitive lease abstraction, unread inbound listings, manual CAM reconciliation. If you cannot name the specific time it saves, spend year one on tools and fluency, then decide with real usage data. Building before your team is fluent is the most common way the budget gets wasted.
What does an AI fluency workshop cost?
Market rates for a focused team training run roughly $2,000 to $15,000 depending on length and depth. For a 10-person firm, a single session that teaches staff to draft letters of intent, summarize leases, produce market write-ups, and handle email with an AI assistant sits in the lower half of that range and is the highest-return line in the budget.
How much should we set aside for maintenance of an AI automation?
Budget roughly 15 to 20 percent of the build cost per year. A $45,000 automation therefore carries about $7,000 to $9,000 in annual upkeep, covering the reality that document formats change, integrations break when a vendor updates, and the underlying models get revised. Plan for it up front so it is a line item, not a month-twelve surprise.
Is AI spending worth it for a firm our size, given most CRE pilots fail?
The failure rate is real — only about 5 percent of CRE firms report achieving most of their AI goals — but the cause is sequencing, not firm size. Firms that buy expensive systems before their staff is fluent land in the 60 percent that feel unprepared. A small firm that spends first on fluency, then on a single measured build, is well positioned to avoid that trap precisely because it can adopt fast and decide quickly.
How do we avoid wasting our AI budget?
Release the money in order: buy seats, train the team within the same quarter, run for a quarter on real work, then fund at most one build against a proven bottleneck. Never buy new software to solve a problem that is actually an adoption problem, and never write the production build check before a pilot has shown a real result on your own files.
What percentage of revenue do real estate firms spend on AI?
Real estate spends less than most sectors — under about 1 percent of revenue on AI, against a corporate average closer to 1.7 percent. For a small firm, a per-employee lens is more useful than a revenue lens: the 2026 benchmark averages about $2,068 per employee, which at ten people implies roughly $20,000 a year, matching the Standard tier in this guide.
Key takeaways
- A 10-person CRE firm should budget $15,000 to $40,000 for a meaningful first year, or $6,000 to $12,000 for a lean tools-plus-training start; a custom build adds $25,000 to $90,000 on top.
- The budget splits into three buckets — per-seat subscriptions, fluency training, and an optional automation project — and naming them separately is what prevents both underspending on training and overspending on a premature build.
- Per-seat subscriptions are cheap ($1,500–$5,000 a year) and should never be where a small firm hesitates; business tiers matter for keeping confidential deal data out of model training.
- Fluency training is the highest-return line in the budget, because unused tools return nothing; fund it before any custom build.
- Sequence the spend — seats, then training, then a quarter of real use, then at most one measured build — because the gap between AI money spent and value captured is about order, not amount.
Want a budget number built around your actual headcount, tools, and the one workflow that is really costing you time? A short conversation will size this far better than any market average. Book your free AI-readiness assessment → and we will map what a first year of AI spending should look like — and be worth — for your firm.
Arthur Wandzel
