Claude or ChatGPT for a UK business? Decide by the job, not the brand.

Claude Team Standard is $20 a seat a month on annual billing or $25 monthly, with a two-seat minimum and a Premium seat at $100 or $125 for five times the usage. ChatGPT Business is $20 or $25 on the same terms, with a two-seat minimum and, since August 2026, a Premium seat at $125 with five times the included usage. At 24 September rates that is about £15 or £19 a seat for either, before VAT, and both bill in dollars. Enterprise on both sides is a sales conversation: Anthropic lists $20 a seat plus usage billed annually; OpenAI publishes no price and procurement reports put it around $60 a seat with a large seat minimum.

The individual plans are also close: Claude Pro at $17 or $20, ChatGPT Plus at $20, with heavier tiers above both. None of this should decide the purchase. A firm of eight people is choosing between two bills of about £1,450 a year each. What differs is what the eight people will get done.

List prices on 24 September 2026. £ equivalents at USD 1 = GBP 0.7545, ex VAT.

OpenAI does not use content from ChatGPT Business, Enterprise or Edu workspaces to train its models by default, and workspace owners control retention, with deleted conversations gone within 30 days and Enterprise able to set custom retention from 90 days. Anthropic’s position on Team and Enterprise is the same: no training by default, retention controlled by the workspace.

The consumer plans are where firms get caught. ChatGPT Free and Plus conversations are used for training unless the user turns it off in data controls. Claude Free, Pro and Max users are asked to choose, with the option presented switched on and five-year retention if it stays on. In both cases the person deciding is an employee with a personal login, not the firm. If your staff are on personal accounts of either, that is the first thing to fix, and it costs a few pounds a seat.

On residency, both process in the United States on their first-party plans and both offer the standard contractual clauses with the UK Addendum. Anthropic contracts with UK firms through its Irish entity; OpenAI through its Irish entity for European customers. If you need in-region processing, Claude is reachable through AWS Bedrock in London and ChatGPT’s models through Azure OpenAI in UK South; both become development projects rather than subscriptions.

Long documents. Contracts, tender packs, board papers, a year of board minutes: Claude handles very long inputs and holds the thread across them, and it is the tool we reach for when the question is “what does this 300-page thing actually say”. Careful writing. Client-facing prose in a firm’s own voice, where the cost of a confident invention is a lost client. In our own use across two products and the client work, Claude is the model we trust to say “the document does not say” rather than fill the gap, and that habit is worth more than any benchmark to a professional-services firm. Software. Claude Code is the tool our own engineering runs on, and it ships with the Pro and Team plans; a firm with one technical person gets a coding agent in the same subscription.

Claude also comes with Projects, which let a team keep templates, tone and reference material in one shared context, and the Microsoft 365 and Chrome extensions, which put it inside Word, Outlook and the browser. For a firm that writes for a living, that is most of the working day covered.

Images. Marketing teams, agencies and anyone producing visuals get native image generation and editing in ChatGPT; Claude does not do pictures. Voice. ChatGPT’s voice mode is a working product for people who think out loud or use it on the move. Breadth of integrations. The catalogue of third-party apps and GPTs is wider, and the ecosystem of tutorials, templates and hires who already know it is larger, which matters when you are training twenty people rather than two. Habit. If the team already lives in ChatGPT, switching costs are real and mostly invisible; a firm that has built a hundred custom GPTs should not throw them away for a marginal writing improvement.

ChatGPT is also, fairly, the safer answer for a general-purpose rollout across a mixed workforce where nobody has a strong preference. It does most things well and most people have used it.

Firms are not one job. The table below is how we allocate seats when we roll out for a client, and it usually ends with both products in the building. The cost of that is nothing: two Business plans with the seats split are the same money as one plan with everyone on it.

Both vendors now sell more than a chat window, and this is where the choice starts to matter for a firm that wants AI to do work unattended. Anthropic’s route is Claude Code, which ships with Pro and Team seats and will build and run software against a codebase, plus the API for anything that has to run inside your own systems. OpenAI’s route is a comparable coding agent, custom GPTs for packaged assistants, and its own API. The seat plans on both sides do not include API usage; the moment a process runs on a schedule, reads an inbox or answers a phone, it is on the API at per-token prices and it is a build, not a subscription.

Two things from running that kind of system in production. The model is rarely the expensive part; the channels around it, telephony, messaging, the integrations with the systems of record, are. And the small, cheap models do most of the work once a workflow is designed properly, with the frontier model reserved for the steps that need it. That is true of both vendors, and it means the agent decision is about the engineering partner and the design, not about which logo is on the API key.

Both vendors give a UK firm a data processing addendum, standard contractual clauses with the UK Addendum for the transfer to the United States, and SOC 2 reports through a trust portal. Anthropic additionally publishes ISO 27001:2022 and ISO/IEC 42001:2023, the AI-specific management standard, which is a useful line on a client questionnaire. OpenAI publishes its enterprise privacy commitments and offers custom retention and data residency options on Enterprise. For an SME the practical difference is small; for a firm answering procurement questionnaires from large clients, having both vendors’ documents to hand is the point, and both make them easy to get.

Buying on a benchmark. The models change every few months and the gap between them on any published test is smaller than the gap between a team that has been trained and a team that has not. Buying personal accounts because they are “the same thing”. They are not: the business plans do not train on your data and the consumer plans, on both sides, default the other way. And buying seats without a policy and without Projects or custom instructions set up, then concluding after a month that AI does not work for the firm. Adoption is the expensive part; the subscription is the cheap part.

If the firm is not sure which way it leans, a two-week paid trial of both, with the same five tasks given to both by the same people, settles it faster than any comparison article, including this one.