Top AI Tools for Architects: Eight Compared by Job and Price

Eight tools cover the jobs a small practice actually pays for: ChatGPT and Midjourney for concepts, Veras and MyArchitectAI for renders, Claude for specifications, Otter and Microsoft 365 Copilot for minutes, and PandaDoc for fee letters. Entry plans run about $8 to $30 a seat a month, so the limits matter more than the price.

None of the eight draws or models for you, because your CAD or BIM software still does that. Each one takes over a repeat task around the drawings, and each hides a catch that the price page does not lead with: a minutes cap, a privacy mode kept for the dearer plan, or a base licence you must already hold. The table puts the catch beside the price.

Follow me on Instagram@sagnikteaches

Prices are USD list prices as of November 2026. Confirm each on the vendor's page before paying, because Microsoft is running a first-year offer on Copilot and several tools charge far less on annual billing than month to month.

Connect on LinkedInSagnik Bhattacharya

The eight tools in one table: job, entry price and the catch

Read the last column before the price. A $10 plan with no private mode can cost you more than a $60 plan that has one, and a recorder that stops at minute 1,200 costs more than its sticker once you buy the plan above it. A sum for a whole practice comes next, then each job gets its own section.

Subscribe on YouTube@sagnikteaches

What a five-person practice pays, and what it gets back

Take a practice of two directors, two architectural assistants and a part-time office manager, with three projects at concept stage this month. The stack I would start them on is one ChatGPT Plus seat, one Claude Pro seat on annual billing, one Otter Pro seat on annual billing and Veras Pro on annual billing. Midjourney, MyArchitectAI, Copilot and PandaDoc wait until a job asks for them.

Now the time side. The figures below are my assumptions for that practice, not measurements, so swap in your own volumes. Every "with AI" figure includes the minutes spent reading the output against the source, which is where the saving is won or lost.

At $50 an hour of staff cost (use your own figure), 11 hours is $550 against a stack costing $74.33, and the stack repays itself after about 1.5 hours ($74.33 divided by $50 is 1.49). I left rendering out of the time sum because the saving depends on how you produce views today. The useful finding is that price is rarely the obstacle at these levels. The checking minutes are, and so is any limit that forces you onto the next plan up.

Concepts: ChatGPT for the options, Midjourney for the mood

ChatGPT Plus is the cheapest way to turn a client's brief into written options before anyone opens the modelling software. OpenAI's Projects feature works on the Free plan as well, so the $20 mostly buys higher limits and more files per Project (5 on Free, 25 on Plus). Park your house questions and tone of voice in a Project, then use a prompt like this for the first pass on a small extension.

You are helping a small architecture practice think through concept options. Do not design anything; describe approaches in words.

Client brief (illustrative): single-storey rear extension to a mid-terrace house, budget $140,000 to $170,000, kitchen-diner with garden access, one rooflight, keep as much garden depth as possible.
Site facts I have confirmed: north-facing rear, 9 m garden, neighbour's extension on the left boundary.

Give me three distinct concept approaches. For each: a two-sentence idea, what it does for daylight, the main cost driver, one planning or neighbour question I must check, and one point the client will probably push back on.
Mark anything you assumed with [ASSUMED].

A typical reply gives a full-width glazed "garden room", a split plan with a small courtyard, and a deeper box lit from above. It will also slip in things you did not ask for: a ceiling height marked [ASSUMED], a cost per square metre, and a line saying one option "should not need planning consent". Delete the last two outright, because the tool knows neither your market nor your planning authority. Keep the questions it raises, since a fair few of them are ones you would have asked on day three. The output is a conversation starter for the client meeting, never a design.

Midjourney covers the other half of concepts: material, light and atmosphere references for a moodboard. Its Basic plan is $10 a month, and Midjourney's plan comparison puts Stealth Mode, which keeps your images private, on Pro ($60) and Mega ($120) only. The price for any job built on a client's site photo or sketch is therefore $60, an extra $50 a month, and the $10 plan stays for references that carry no client information. The series tutorials on Midjourney concept images for client meetings and the Stealth Mode privacy settings go through both sides in more detail.

Rendering: Veras or MyArchitectAI, and what your BIM licence may hold

Veras, now part of Chaos, turns a view of your model into a photoreal image. Its page lists Revit, SketchUp, Rhino, Archicad, Vectorworks, Forma and Allplan as host apps, with Enscape, V-Ray, Corona and a web app as other places it runs. Starter is $202.80 a year ($16.90 a month) for 500 credits and Pro is $348 a year ($29 a month) for 2,000, both on annual billing. Chaos equates Starter to roughly 50 images at 1K resolution a month and Pro to roughly 200, and the page advertises a 14-day free trial. Month to month, Pro showed about $59 in a late-October check, so the billing toggle matters more than the plan name; read the cart total on the Veras page before you pay.

MyArchitectAI starts from sketches, photographs and uploaded SketchUp or Archicad designs instead. Its page offers ten free renders to test and says plans start from $29 a month, but it does not state what each plan allows or what the licence says about commercial use. Settle both questions in writing before its images go anywhere near a client. A longer comparison sits in the tutorial on AI rendering tools for small architecture practices.

If you use Archicad, look at your subscription first. Graphisoft lists an AI Visualizer inside Archicad 30 on its Studio, Collaborate and Max plans, so a practice already on one of those can test it for nothing extra before buying a separate renderer. Confirm that your own plan carries it.

A usable-image test settles the choice faster than any feature list. Export three model views, run them through two tools, and count the images you would be happy to show a client without redrawing. Suppose one in four is usable. Pro's roughly 200 images a month then yields about 50 usable ones, which is $0.58 each ($29 divided by 50). A concept stage of three options with six views each needs 18 usable images, so about 72 generated at that hit rate: Starter's 50 images cannot cover one project, and Pro covers just under three. If your hit rate is one in three, a project needs 54 images and Pro covers three with 38 to spare. The plan you need follows the hit rate, not the price.

One failure to plan for: a render of a flat-roofed extension comes back with a pitched roof and an extra window on the side elevation, and it goes to the client as "the look we are going for". Two weeks later the client asks where the side window went. Compare every generated image with the elevations before it is sent, caption it "AI-assisted illustration, not the design", and put the elevation beside it.

Specifications: Claude Pro with your own clause library

Claude Pro costs $20 on monthly billing or $17 a month when you pay $200 up front for the year. Projects come with it (Free accounts get up to five), and a Project can hold your master clause library, with files capped at 30 MB each. That library is what separates a useful spec draft from a generic one. The rule that matters most sits in the instruction: use clauses from the library, mark anything new, and name no product or standard that is not in the library.

Use only clauses from the attached master library. Where the library has no suitable clause, write a short draft and mark it [NEW - CHECK].
Never name a manufacturer, product or standard number unless it appears in the library.
After the clause, list every fact I still have to confirm and who could confirm it.
Source notes follow.

The before and after below shows the difference. The "before" is a site note of the kind that sits in a notebook for a week:

ext wall insul 100mm PIR between studs?? check w eng, fire stopping at cavity TBC, same as last job

With the rules above, the draft comes back as a short clause in house format: insulation of the thickness shown on the drawing, fitted between the studs and fixed as the manufacturer instructs, followed by three flagged items. The thickness is taken from a note that carries a question mark. No product is named. The cavity fire-stopping detail waits for the engineer. "Same as last job" gets listed as a fact that cannot be resolved from the notes. Without the rules, the same input often returns a named board and a performance figure, and that is the version that causes trouble because it reads as finished.

Whatever Claude drafts, check these five before issue: standard numbers, product names, performance values, dimensions, and any statement about regulations in your jurisdiction. Keep a "verified by" initial on each clause. The longer walk-through is in the tutorial on using Claude for specifications and planning reports.

Minutes: Otter Pro's 1,200 minutes against Teams recap

Otter's free Basic plan gives 300 transcription minutes a month, 30 minutes per meeting and three file imports in total, which is enough to try it on a few calls. Pro is $16.99 on monthly billing or $8.33 a month billed annually, with 1,200 minutes and up to 90 minutes per meeting. Business is $30, or $19.99 on annual billing, with meetings of up to four hours. The Otter pricing page carries the current numbers, and Otter's home page says it joins Zoom, Teams and Google Meet calls and can capture on mobile, so try it on one site meeting before relying on it there.

The cap is where practices get caught. Five meetings a week at 50 minutes is 250 minutes, or about 1,083 a month, so Pro fits with a little to spare. Replace one of those with a two-hour design review and the week becomes 320 minutes, about 1,387 a month: Pro runs out before the month does, and the long meeting is cut at 90 minutes anyway. For a practice with regular long reviews, Business at $19.99 is the real entry price, or you record only the meetings where decisions are made.

Microsoft 365 Copilot Business is the other route, and it only makes sense if your team already meets in Teams. It costs $21 a user a month on an annual commitment ($25.20 billed monthly), as an add-on to an eligible Microsoft 365 plan. Microsoft's page currently shows $18 for the first year, valid to 31 December 2026 with an annual commitment. The app is now called Microsoft Copilot, though product pages still say Microsoft 365 Copilot Business in places. Teams' intelligent recap needs either Teams Premium or a Copilot licence, so check what you already hold before adding either. The same licence helps with Word drafting, which is why it comes up again below.

This extract shows the kind of exchange a recorder gets wrong. The table after it is how a first-pass summary tends to read:

Contractor: We can pour the slab Thursday week if the engineer signs off the rebar.
Engineer: I'll look at it, but I might not get there before Friday.
Contractor: Then it's the Monday after. Client, are you happy with that?
Client: Fine, as long as the kitchen supplier still delivers on the 14th.

Three fixes in three lines is a plausible rate, and it is why the earlier sum credits a 15-minute saving per meeting rather than the full 25. Tell attendees you are recording and note it in the minutes, since the rules on recording differ from place to place. The tutorial on architect meeting minutes with AI covers the same-day issue routine.

Fee letters: a chat assistant and Word, or PandaDoc

Drafting a fee letter is a job for the chat assistant from the concepts section, and Copilot in Word does the same if you hold the licence. A separate paid tool does something else: it sends, tracks and gets the letter signed. A filled-in request to the assistant looks like this.

Draft a fee letter for a small residential extension. Use these stages and fees (illustrative): concept $2,400, developed design $3,600, technical design $4,800, site stage as hourly at $95.
Include: scope per stage, what is excluded (structural design, surveys, planning fees, boundary and neighbour matters), assumptions about client decisions and response times, how variations are priced, payment terms of 14 days, and that the offer is valid for 30 days.
Write it in plain English and mark every figure or term I gave you with [CONFIRMED] so I can see what you added.

The draft will have the right bones: stages, exclusions, assumptions and a validity period. Read the exclusions first. One slip to look for is a sentence such as "fees include all consultants", which the prompt never asked for, and which would commit the practice to paying engineers out of its own fee. Delete anything not marked [CONFIRMED] unless you add it on purpose. Keep the wording in a ChatGPT Project or a Claude Project rather than a custom GPT, because OpenAI stopped custom GPT creation in late September 2026 and existing ones stop running on 11 December 2026. The tutorial on AI for architects' fee proposals and briefs has the stage-by-stage structure.

The sum for a signing tool is short. PandaDoc's Starter plan is $19 a seat a month on annual billing, with unlimited uploads, e-signatures and its AI Assistant; the free plan allows five documents a month; CRM integrations, approval workflows and AI data extraction belong to Business at $49. Four fee letters a month with 20 minutes of signature chasing each is 80 minutes, which at $50 an hour is about $67 of time for a $19 tool. If you send one letter a quarter, the sum reverses and Word plus email wins.

Which tools to buy first at three practice sizes

The larger figure is $80 for the assistant seats, $39.98 for Otter, $58 for Veras and $38 for PandaDoc. Whichever row you are in, run each tool for 30 days with a simple log. An illustrative month for the minutes recorder might read: 14 meetings recorded, 5.5 hours saved, 9 errors caught (names, dates, owners), no file you hesitated to upload. Keep a tool only if its log shows hours saved well above its price and nothing you would regret sharing.

Three checks before client files go into any of these

First, find the training setting for the plan you will really use. Claude's pricing page lists model training as opt-out on Free, Pro and Max and as none by default on Team and Enterprise. ChatGPT's business plans and Microsoft 365 Copilot also leave business content out of training by default. Team plans from Claude and ChatGPT both need two seats, so a solo principal either pays about $40 to $50 for that default or removes names, addresses and drawing numbers before upload.

Second, read what the plan says about generated images. Private modes, commercial-use terms and indemnities differ by plan, and the cheapest plan is often the one without them. The Midjourney sum earlier shows the pattern.

Third, tell the people who carry your risk. Your professional indemnity insurer and your standard appointment terms may both have views on AI-assisted work. Settle them before a client asks. A written practice AI policy settles all three, and the policy and copyright tutorials listed under Further reads go deeper.

Tools left out on purpose

Snaptrude (Pro at $60 monthly or $50 on yearly billing, aimed at early-stage concept design) is design software, so it falls outside a price-and-job comparison of practice tools. Adobe Firefly starts at $9.99 a month for 2,000 credits and is a fair alternative to Midjourney if you already work in Creative Cloud. Adobe says its outputs are safe for commercial use only from its own Firefly models and that indemnification is for eligible enterprise customers, so partner models inside the app are your risk. Fireflies Pro was listed at $10 a seat on annual billing and $18 monthly in late October, so confirm it on the vendor's page; it is a straight alternative to Otter.

Questions practice owners ask before paying

Can client drawings go into a personal ChatGPT or Claude plan?

Check the data setting and your client contract first. Claude's pricing page lists model training as opt-out on Free, Pro and Max, and none by default on Team and Enterprise; ChatGPT's business plans also do not train on workspace content by default. Both team plans need two seats, so a one-person practice pays roughly $40 to $50 a month for that default, or strips names and addresses before uploading.

If I can only pay for one tool, which should it be?

A $20 chat assistant, either ChatGPT Plus or Claude Pro, because one seat covers concept write-ups, fee-letter drafts and clause drafts, the three jobs most practices repeat every week. Add a recorder next, starting with Otter's free Basic plan, and a renderer only once you have a real use for generated views. Try one live brief in both free plans before choosing.

Are AI renders safe to show clients?

Show them as mood or massing studies, labelled as illustrations and placed beside your elevations. Generated images can add windows, change a roof form or shift proportions, so compare each one with the drawings before it leaves the office. Read each tool's terms on commercial use of outputs too: Adobe, for example, says its commercial-safety claim covers its own Firefly models, not partner models.

How do I spot a price that has changed since this was written?

Open the vendor's page and check three things: the billing period (annual prices are shown per month but paid up front), the seat minimum, and any cap on minutes or credits. Copilot Business was showing a first-year offer that ends on 31 December 2026, and Veras charges far less on annual billing than month to month, so read the cart total before you pay.

Further reads

• How Much Does AI Cost a Small Architecture Practice per Month? — A fuller monthly budget for a small practice.
• Is ChatGPT Plus Worth It for an Architect? A Practice Cost Check — Test whether one ChatGPT Plus seat repays itself.
• AI Tools for Architects 2026: A Toolkit From Brief to Handover — The same ground arranged by project stage.
• AI Policy for Architects: A Ready-to-Use Checklist and Clauses — Policy clauses to adopt before client files go in.
• Copyright and Client Data Risks of AI in Architecture Practices — Who owns generated images and where client data goes.
• Free AI Tools for Architecture: Concepts, Admin and Where Free Stops — What the free plans cover before you pay.
• Best AI Platform for Architects: Revit, Archicad or a Standalone Tool? — Whether your BIM subscription already includes AI.
• Claude vs ChatGPT vs Gemini for Architects: Which to Pay For — Which of the three chat assistants to pay for.
• Best AI for Architecture and Interior Design: Concept Image Tools by Cost — What concept-image tools cost a small architecture or interior design studio a month, and what each licence lets you show a client.
• Best AI Tool for Construction Design: What It Drafts and What Needs a Stamp — Which AI design outputs a small builder can issue (sketches, option comparisons, specification wording) and which need an engineer's or architect's sign-off.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: vendor pricing pages for Claude, Otter, PandaDoc, Chaos (Veras), MyArchitectAI, Microsoft 365 Copilot Business and Adobe Firefly, read in November 2026; ChatGPT and Midjourney plan listings.