Claude or ChatGPT for a five-person company: what we actually use and why
Ben McDermott · September 2, 2026 · 3 min read
We build on Claude. Clients ask why almost every week, usually because they've already got ChatGPT and want to know if they're doing it wrong. They're not. Here's the honest version of how the two compare for a small business, and the part that matters more than the choice.
Where Claude wins for a small business
- Writing in a specific voice. Give Claude two pages describing how you write and a few real examples, and the drafts stop sounding like a chatbot. For proposals, follow-ups, and customer replies, this is most of the value.
- Following long instructions. A 40-line skill that says exactly how to turn a call transcript into a proposal, in what order, with what exceptions, gets followed. This is the difference between an assistant and an automation.
- Working with a lot of context at once. A year of call notes, a pricing sheet, and a contract template in one conversation, without losing the thread.
- Projects, skills, and Cowork. Shared workspaces where your team and Claude work on the same documents with the same rules, and reusable instructions anyone can run.
- Reading and writing files, calendars, and inboxes with a person reviewing before anything goes out.
Where ChatGPT wins
- Images. If you need product mockups, social graphics, or quick illustrations, ChatGPT's image generation is the practical choice.
- Familiarity. More of your staff have already used it, which lowers the training cost on day one.
- Breadth of consumer integrations and plugins, and a voice mode people already like for thinking out loud.
- If you're deep in Microsoft 365, Copilot deserves a look before either, because it's already in your license and inside the apps your team uses.
What they cost
For an individual, both run about $20 a month for the paid tier. Team plans for either are in the $25 to $30 per person range. For a five-person company that's $100 to $150 a month total, which is less than one hour of most consultants. Check current pricing before you buy; both change their plans often.
The part that matters more than the model
Here's what we've learned building this with clients: the results come from the context and structure you give the tool, not from which tool it is. A business with a well-written context file describing its voice, pricing, and process, and five skills for its weekly jobs, will get better output from either model than a business pasting one-line prompts into the best model on earth.
That's also why the choice is lower-stakes than it feels. The context files and the skills are the valuable part, and they transfer. We've moved clients between models in an afternoon.
What we recommend, in practice
- Pick one and commit for ninety days. Two half-used tools are worse than one well-used one.
- Write the context file first. Your voice, your pricing, your terms, how you handle the five most common situations. Two pages. This is the afternoon that pays for everything after it.
- Build one skill for one weekly job. Proposal drafting is usually the best first one.
- Add a review step to anything customer-facing. A person presses send.
- Only then think about automation: scheduled jobs, inbox connections, and the rest.
We default to Claude because steps two and three go faster and hold up better there, and because we've built on it for long enough to know its edges. If your team is already productive in ChatGPT, we'll build there and tell you when a workflow would go better on the other side.