What a Small Practice Really Pays for Its Software Stack.

The licence fee is the smallest line on the bill. Here is the full cost of an architecture software stack for a practice of five or fewer, and where it leaks.
The Subscription Total Is the Number You Can See
Every principal knows the subscription total. It sits in the accounts, it renews on a schedule, and it is the figure that gets quoted when someone asks what the software costs. It is also the smallest and most honest part of the real bill.
Build the stack the way a five-person practice actually buys it. A CAD or BIM package, priced per seat per year. A visualisation engine. A vector and layout tool for boards and the Design and Access Statement. A photo editor for post-production. A video editor for the fly-through that clients now expect. A PDF suite for issuing drawings. That is six categories, and for most offices they are six separate vendors with six separate renewal dates.
Multiply by five seats and the number arrives before anyone has drawn a line. Most principals stop counting there. They should not.
The quiet extras
Beneath the headline licences sit the items nobody tracks as a stack: plugin licences, asset libraries, cloud storage tiers that creep up as project archives grow, font licences, render farm credits for the deadline week. Each one is small. Each one is easy to approve in isolation. Together they can rival a part-time salary line, and because they renew on different dates and charge to different cards, nobody ever sees them added up.
If you have started experimenting with AI tools, add another category. Usage-based generation behaves differently from a seat licence. Published model pricing is quoted per clip, per second or per image rather than per person. Aggregated listings put a five-second video clip from current models somewhere between roughly $0.29 and $0.72 depending on the model, and a ten-second output costs about twice that before a single failed attempt is counted. Image models are cheaper per attempt, at around $0.05 to $0.07 per image on one published table, but a moodboard with eighty exploratory images still produces a bill that nobody budgeted for. These are third-party figures that move often, so check each vendor's checkout page before you plan around them.
The structural point matters more than the exact numbers. Each vendor runs its own credit pool, and those balances do not transfer. A comparison of current video tools notes explicitly that Kling, Veo and Runway credits cannot be moved between services. Unused credit on one tool does nothing for a shortfall on another.
Idle seats are the first leak
Seat counts rarely match usage. A visualisation licence bought for the whole office is opened by one person, twice a week. A video editor was bought for one competition entry and has not launched since. Before you look at any other cost, list every seat and the last date it was actually opened. That audit costs nothing and usually pays for itself in the first hour.
The subscription total is the part of the bill that renews on a schedule. It is not the part that grows.
Seats Are Cheap. Hours Are Not.
Put a chargeable rate on the problem. A practice billing £60 an hour that loses 45 minutes a day per person to tool friction is carrying a larger cost than any subscription in the stack. Across five people and a working year, that is well over 800 hours, or roughly £50,000 of chargeable time spent moving files rather than designing. The figure is illustrative, not a benchmark, but the arithmetic is worth running with your own numbers.
Follow one planning submission
Walk through a single planning application. The model is built in one package. It is exported to another for rendering. The renders round-trip to a photo editor for post-production. The finished images then go into a layout tool for the Design and Access Statement, alongside drawings that were exported from the first package as PDFs.
Count the file conversions. Model to render engine: one. Render to photo editor: two. Photo editor to layout: three. Drawings to PDF to layout: four. Every conversion is a place where a setting can differ, a scale can drift or a layer can drop out. Every one is also a place where somebody has to stop and check.
Version drift is the expensive failure
The costly failure is not slowness. It is version drift. A late change to a window position goes into the model, but one export is not repeated. The elevation on the board no longer matches the model, and nobody notices until it is on the planning officer's desk or in front of the client. The fix takes ten minutes. The credibility cost does not come back so quickly.
Interior design has the same failure in a different form. A mood and material board is assembled across four tabs: a supplier site for samples, an image generator for atmosphere, a layout tool for the board, a spreadsheet for the finish schedule. By the time the client signs off, the schedule says one oak and the visuals show another. Nobody made a mistake. The stack made the mistake possible.
AI generation adds its own verification burden here. Image models can alter window proportions or change materials between iterations, and generated camera moves can misrepresent circulation. Vendor capability statements are vendor claims, not proof of production reliability, and nothing in the current evidence supports treating generated output as dependable construction information. The labour to check and repair it belongs in the cost, not outside it.
Compare the AI bill with saved chargeable hours, not with the headline subscription price.
Learning Curves and the Cost of Switching Between Tools
Every additional package is a training bill, and in a small practice somebody always pays it unbilled. Count the weeks a new graduate needs before they are genuinely productive in each tool. Then count who covers the gap: usually the person who would otherwise be on a fee-earning task.
Specialists become single points of failure
In a five-person office, each specialist tool tends to belong to one person. One person knows the render pipeline. Another is the only one who can produce a clean competition panel. When the render specialist is on leave in the week an entry is due, the entry stalls. A larger firm absorbs this with a dedicated BIM manager and a visualisation team. A practice of five has no such buffer, so every tool added is carried by the same few people, and every absence is a risk.
Teaching software as well as design
Part I and Part II students arriving on placement bring one toolset from university and meet another in the office. The practice ends up teaching software as well as design. That is a real contribution to the profession, but it has a cost, and the more tools the stack contains, the more of the student's placement is spent learning interfaces rather than learning how a scheme gets built.
The same applies across adjacent disciplines. A landscape practice has to check planting-season and species claims that an image model may invent. A product designer may need watertight geometry that an AI-generated mesh does not provide until topology has been repaired. A planning consultant must preserve the evidential chain behind visual impact material. In each case a new tool arrives with its own checking discipline, and someone has to own it.
A fifty-person firm can afford a tool for every job. A five-person practice can only afford the tools it can actually keep fluent in.
Consolidating Without Giving Up Capability
Before cutting anything, test which jobs genuinely need a separate tool. Concept massing, planning drawings, DWG and DXF round-trips, IFC4 export, renders, retouching, boards and a practice website can all be handled inside one platform. That is the argument for consolidation, and it is worth being specific about what it covers.
Drawing and modelling
Linea is the BIM modelling lane: browser-based drawing and modelling where the plan and the model are one building rather than one generated from the other. It imports DWG, DXF and IFC, including native files from Revit and ArchiCAD, and exports DWG, DXF or IFC4. The IFC4 export carries rooms as spaces with floor area and volume, base quantities on walls and slabs, and each type's layers. Schedules of rooms, doors, windows and stairs can be placed on sheets as tables.
Vector handles the drawing-production side. Drop in a vector-ready PDF from AutoCAD, Revit or SketchUp LayOut, or a DWG or DXF, and it lands as editable linework at true scale. Where a sheet holds twenty-odd drawings on the same layers, the import finds the separate drawings and a click crops to the one you want. Export goes out as PDF, SVG, DWG, DXF or PNG.
Boards, post-production and the shared files
The shared-file problem shrinks when the surrounding work lives in one account. Layout builds crit boards, competition panels and Design and Access Statements at true sheet size, with a Sheet mode for the board and a Document mode for the statement or portfolio, and exports a vector PDF at true dimensions. Retouch and Image cover post-production and visualisation. The hand-offs between five applications become one project.
| Job | Typical separate tool | Where it sits on ArchAdemia Tools |
|---|---|---|
| BIM modelling, plans, IFC4 export | Dedicated BIM package | Linea |
| PDF, DWG, DXF import and true-scale drawing | CAD viewer plus vector editor | Vector |
| Crit boards, competition panels, DAS | Desktop publishing package | Layout |
| Visualisation and post-production | Render engine plus photo editor | Image and Retouch |
| Practice website | Separate site builder | Web |
The honest limits
Be realistic about what consolidation cannot do. A practice with a heavy Revit or ArchiCAD dependency from consultants will keep those tools. The saving comes from what sits around them, and the open IFC import keeps the exchange workable rather than forcing a choice. Generation is also a separate cost: the paid AI actions use plan credits, each shown before it runs, so the credit allowance still needs watching in a busy competition fortnight.
One plan against several seats
The economics are straightforward to state. ArchAdemia Tools is £19 a month billed yearly, or £29 monthly, with a 7-day free trial and no card needed. That is one plan covering the studios rather than a seat-by-seat build-up. The sensible test is to run a real live job through the trial before committing, rather than judge it on a demo project.
Consolidation is not about owning fewer tools for its own sake. It is about removing the hand-offs where hours and accuracy leak away.
A Cost Audit a Principal Can Run in One Afternoon
None of this needs a consultant. A principal can run the audit in an afternoon with the accounts to hand and a pen.
- List every paid tool and every seat. Include the plugins, the storage tier and the font licences. Next to each seat, write the last date it was actually opened. Cancel anything that has sat idle for a quarter.
- Time one real submission end to end. Pick a live planning application or competition entry. Record the time from first massing to issued PDF. Note each export, conversion and manual re-check. That total is your friction cost.
- Price the exit as well as the entry. For each tool, check the file formats it exports, whether you can reach an archive, and what happens to old projects if the subscription lapses. A tool that holds your project history in a proprietary format is a liability, whatever it costs per month.
- If you use AI generation, record the credit mechanics. Note each tool, plan, monthly credits, rollover rules, resolution limits and commercial-use terms. Then measure one real workflow including rejected images, video reruns, upscaling, mesh repair and manual redraws. Allow for failed generations explicitly: forty five-second attempts a month at about $0.30 each is roughly $144 a year in direct generation, and ten hours of review and rework will exceed that at ordinary charge-out rates. The sum is illustrative, so run it with your own usage.
- Set a rule for new tools. Nothing joins the stack unless it replaces something or removes a hand-off. Review the whole stack every January against the hours it saved, not against the features it lists.
Where a price in this article came from a third-party comparison, treat it as a planning figure and confirm it at the vendor's own checkout before you rely on it. Several widely discussed tools, including some current consumer and API plans from the major assistant vendors, did not have reliable primary pricing available at the time of writing, and they are best marked as to verify rather than assigned an annual total.
The takeaway is a change of denominator. Stop asking what the software costs per seat and start asking what it costs per issued submission, in subscription, in credits and in the hours spent checking and repairing what comes out. A small practice that answers that question honestly usually finds that the stack it can afford is smaller than the one it owns.
Keep reading.
Try ArchAdemia Tools for yourself
Draw it, model it, render it, publish it. One place, built for architects and small practices. From £19 a month, all in.


