OpenAI made this whole market. ChatGPT’s launch in late 2022 turned language models from a research curiosity into a consumer product, and it has stayed the most used AI tool since: Pew Research Center’s February 2026 survey found 44% of US adults using it, and OpenAI has cited over 900 million weekly users. For your firm, that popularity has a practical meaning before any procurement decision is made: your people already know it, and some of them already use it for work.
What you are buying
OpenAI’s offering splits into three things worth distinguishing:
- ChatGPT, the app: the chat interface on web and mobile, in free and paid personal tiers, and in Team and Enterprise tiers for organisations. The business tiers are the ones that matter for a firm: they exclude your conversations from model training by default, add workspace administration, and carry terms you can actually show an auditor.
- The API: the same underlying models, sold per use for building into your own systems. This is what powers custom automations, document pipelines and assistants. It is also where costs are measured per job rather than per seat, which is the right shape for automation work.
- The model range: OpenAI maintains a spread from fast, cheap models to its most capable reasoning models. The names change often enough that chasing them is a mistake; what matters is that the range lets you match cost to task, which any well-built system should do automatically.
The tier decision in one view:
| Tier | Training on your data | Admin controls | Right for |
|---|---|---|---|
| Free / Plus (personal) | Possible; opt-outs are personal settings | None | Private experimentation only |
| Team | Excluded by default | Workspace, seats, sharing | The standard firm starting point |
| Enterprise | Excluded, contractually | Full: SSO, retention, compliance | Larger firms, higher stakes |
| API | Excluded on business terms | Per-key, per-system | Automations and custom builds |
What it does well
The reliable wins are the unglamorous ones:
- Drafting. First versions of emails, letters, summaries, meeting notes and internal documents: an hour of writing becomes minutes of review.
- Summarising. Long documents, threads and transcripts reduced to what matters, with the original one question away.
- Explaining. Unfamiliar jargon, a regulation, a spreadsheet formula: plain English on demand, at whatever depth you ask for.
- Structuring. Messy notes turned into clean tables, lists and templates; voice and image input make the capture easier still.
None of this replaces judgement; all of it removes typing.
The equally reliable failure mode is treating fluency as accuracy. ChatGPT will occasionally state something false with complete confidence. The realistic limits guide covers this properly; the short version is that anything factual going in front of a client gets verified by a person who can tell.
The data questions
Before client or sensitive data goes anywhere near it:
- A business tier in place, with training exclusion confirmed in writing
- OpenAI’s data processing agreement signed, sub-processors noted
- Processing location and retention recorded in your DPIA
- The one rule circulated and enforced: client personal data only ever in the approved, contracted tool, never in anyone’s personal account
The most common ChatGPT incident in small firms is not a vendor failure: it is a staff member pasting client information into a free personal account, because no approved alternative was named. The approved-tools policy template is the one-page fix, and the UK governance guide shows how it sits inside Consumer Duty, SM&CR and UK GDPR.
Where it fits
Pick ChatGPT as the general-purpose staff tool when no single office suite dominates your workflow, or when you want the largest ecosystem and the fastest-moving feature set. If your firm lives in Microsoft 365 or Google Workspace, compare it honestly against Copilot and Gemini, which win on proximity to the work. For judgement-heavy long-form writing, run the same real tasks through Claude before deciding. The cluster overview has the three questions that settle the choice.
Where I fit in
The gap between “we bought ChatGPT Team” and “our processes are faster” is where most firms stall. I close that gap: an Automation Audit finds the processes worth pointing it at, staff training builds the habits, and a Kick-starter wires the API into the workflows where a chat window is not enough. If you want the shortcut, book a call and bring the task that eats the most of your team’s week.