UA Design  /  internal proposal
August 2026

Keep the geometry.

Moving off ChatGPT to Claude, and putting our renders in a tool that was built to draw buildings instead of one that guesses at them.

$400
per month, all in
+$240
net monthly increase
4 wks
migration, no downtime
scroll to orbit
The ask

Two jobs, one tool.

We pay for one general purpose tool and use it for two unrelated jobs: everyday knowledge work, and architectural rendering. It is competent at the first and improvising at the second.

This proposal splits them. Claude for the work, a purpose built renderer for the drawings. Seat cost does not change, because Claude Team and ChatGPT Business are priced identically at every tier. What changes is what the seat includes, and where our renders come from.

Line itemMonthlyAnnual
Claude Team, 8 Standard seats$160$1,920
Visoid Premium, 1 seat$40$480
Hosting Fees$200$2,400
Total, annual billing$400$4,800

Against roughly $1,920 a year on ChatGPT seats today, the net increase is about $2,880 a year.


Why Claude

Cowork is the ceiling, not the day one win.

Every assistant can answer a question. The difference now is whether a tool can be handed a task and left alone to finish it. Cowork works inside the folders and tools you point it at, runs multi step work start to finish, and returns a finished file rather than text somebody still has to reformat.

Worth being straight though: today our teams draft emails and ask questions. Nobody is going to open Cowork on day one and be transformed. I will be using it immediately for infrastructure and automation work, and the studio grows into it.

Price is no longer an argument in either direction.

Standard seats are $20 annual and $25 monthly on both platforms. On 10 August 2026 OpenAI added a ChatGPT Business Premium seat at $100 annual and $125 monthly for 5x usage, which is identical to Claude Team Premium, and both let you mix seat tiers in one workspace. There is no cheaper tier, no cheaper upgrade path, and no cost penalty for being wrong. The decision rests on capability and output quality, which is where it should rest.

Included, and not a separate line item.

Team seats carry Claude Design alongside Claude Code and Claude Science, and our content is not used for model training by default. That last one is a contractual position rather than a setting somebody has to remember to toggle, which matters given how routinely client material goes into these tools.

Every month we stay, the switch gets more expensive. The cheapest moment to change platforms is before anyone has built anything on top of one.

Right now there is almost nothing to migrate. No custom assistants anyone depends on, no workflows built around one vendor, no habits to unlearn. The only thing worth carrying across is each person's accumulated context, and there is a supported import for exactly that.

The argument, in one interaction

Same model. Two renders.

geometry preservedOur building
geometry reinterpretedSomething like it
drift  0%

Ask a general image model to change one material and it regenerates the whole frame. Window proportions move. The roofline shifts. It looks impressive, and it is not what we drew.

Recommendation

Visoid Premium, one seat, billed annually.

Our teams work in SketchUp. Visoid takes SketchUp models and viewport exports directly and holds the underlying geometry, so a material or a season can change without everything else in the frame changing with it.

Pro
$15per month, annual  /  $30 monthly
Renders1,000 / mo
Resolution2K
Model uploadno
Multi promptno
Batchno
Premium  /  recommended
$40per month, annual  /  $80 monthly
Renders3,000 / mo
Resolution4K
Model uploadyes
Multi promptyes
Batchyes
Resolution decides it

Pro caps at 2K. That is a review asset. It is not a presentation board, it is not print, and it is not something we hand a client at scale.

Model upload is the whole point

Uploading the actual SketchUp model rather than a flat screenshot of it is what makes geometry preservation reliable, and it sits behind Premium.

Not the render count

One shared account will not approach 1,000 renders a month. Do not buy Premium for the credits.

Annual billing, specifically.

Visoid roughly doubles on monthly billing. Premium is $40 annual against $80 monthly, so committing annually saves about $480 a year, which is more than the annual plan costs. For a production workflow rather than a trial, monthly billing buys a second subscription we never receive.

One seat, one owner.

A seat shared informally across the studio conflicts with the terms of an individual plan, leaves nobody able to find their own past work, and collides when two people render at once. The fix is not more seats. Name one owner, route requests through them, log demand, and revisit Visoid's Team plan at 90 days with real numbers instead of a guess.


Also considered

Most published comparisons of AI rendering tools are written by companies selling AI rendering tools, and each concludes its own product wins. Weighted on hands on testing instead.

Veras

case for
Runs as a plugin inside SketchUp, so there is no export step at all. Given where our teams live, the most credible alternative here.
why not
Near $59 a month, so it costs more and saves nothing. In testing it does not match Visoid on changing one element while leaving the rest untouched.

D5 Render

case for
Full real time engine with AI on top, and holds accuracy across animation rather than single stills.
why not
Heavier tool, capable GPUs, real training investment. A decision for if and when we bring visualization fully in house.

Vizcom

case for
Excellent sketch to render, strong for very early concept work.
why not
Built for product and industrial design rather than architectural output. Solves a different stage of our process.

Keep ChatGPT for images

case for
No migration risk on the visual side.
why not
Another full seat to retain the weakest part of the current setup, and it keeps the geometry problem entirely.
Cost detail
Against current spendMonthlyAnnual
Current: ChatGPT Business, 8 seats$160$1,920
Proposed total$400$4,800
Net increase$240$2,880

Assumes 8 ChatGPT Business seats at the current $20 annual rate. If we are on Plus seats or a different count, this line needs correcting before the numbers are presented.

Monthly billing raises Claude Team to $25 a seat and doubles Visoid Premium to $80, adding roughly $960 a year for nothing.

On the Hosting Fees line.

I run a personal Claude Max 20x subscription at $200 a month. It is what I use for the firm's infrastructure, automation, and internal application work, and it carries usage limits well above a Team seat, which that work requires.

A Team seat does not cover this work, and neither does a Team Premium seat. The usage ceiling on Max 20x is what makes long running agentic sessions possible at all, and those sessions are how the internal tooling gets built. The account also stays personal and separate from the company workspace, which keeps firm data and client context inside the managed organization.

Migration

Four weeks, nothing cancelled early.

Week 1

Stand up, do not switch

  1. Create the Claude Team organization and provision 8 seats. Keep ChatGPT active.
  2. Configure connectors so Claude reaches our real files rather than working from pasted text.
  3. Buy the Visoid Premium seat on annual billing and assign a named owner.
Week 2

Carry the context over

The only thing we actually have to migrate, and the step that usually gets skipped. Claude has a supported memory import. About ten minutes per person.

  1. Each seat holder exports their memory from ChatGPT and imports it into Claude.
  2. Each person reviews what carried over and trims anything stale or wrong.
  3. Shared Projects set up so team context lives in the workspace, not in individual accounts.
Week 3

Run both, in parallel

  1. Teams run real work through Claude with ChatGPT still there as a fallback. No pressure, no ceremony.
  2. Cowork pilot on two recurring tasks that currently cost somebody an hour or more a week.
  3. Visoid renders for at least one live project, compared honestly against what we would have produced before.
Week 4

Decide

  1. Feedback from all 8 seat holders.
  2. Usage analytics reviewed to see whether any seat is hitting limits and needs a Premium tier.
  3. If the team is satisfied, cancel ChatGPT. If not, we lost one month of overlap and learned something worth knowing.
What could go wrong

Cowork burns usage limits faster than chat

assessment
Low in practice. Email drafting and questions sit well inside a Standard seat. This only becomes real if agentic adoption goes well, which is a good problem.
mitigation
Start all 8 on Standard, watch usage analytics, upgrade only the seats that actually hit limits. ChatGPT now charges the same for its equivalent upgrade, so staying does not avoid this.

The team resists another tool change

assessment
Lower than usual. Nobody has built custom assistants or workflows that would be lost. It is a different chat box plus imported context.
mitigation
The parallel run and the memory import exist for this. Nobody starts from zero and nobody loses their fallback until they choose to.

Shared Visoid seat becomes a bottleneck

assessment
Moderate. Depends on render demand we have not measured yet.
mitigation
Named owner, request queue, demand log. Revisit the Visoid Team plan at 90 days with real numbers.

We genuinely miss ChatGPT image generation

assessment
Possible for non architectural work such as marketing and social assets.
mitigation
Reassess at 90 days. One ChatGPT Plus seat at $20 covers it if the need is real, and that is far smaller than keeping 8.

The migration turns out to be wrong

assessment
Low, but the plan should survive being wrong.
mitigation
Nothing is cancelled until week 4 and annual billing is not committed until after the parallel run. Rollback costs one month of overlap.
Decision requested

Approve, and week one starts Monday.

Total commitment during the overlap month is $560. Ongoing commitment after cancellation is $400 a month.

Download the full written proposal →