What it is
ChatGPT is a general-purpose AI assistant made by OpenAI, available on the web, desktop, and mobile. Give it a request, text, or files, and it can draft, summarize, analyze, or talk through options with you.
Four features matter most for organizational work:
- Projects hold the chats, reference files, and standing instructions for one ongoing effort, so every conversation starts with your context loaded.
- Deep research runs a multi-step search across many sources and returns a cited report. Our deep research FAQ covers what to expect.
- Agent mode takes on multi-step tasks: browsing sites, running analysis, and returning editable documents and spreadsheets. Its output is a draft to check like any other.
- Custom GPTs are purpose-built assistants with their own instructions and reference files, shareable across a team.
Newer additions: plugins now carry capabilities across ChatGPT and Codex, having absorbed the apps directory in July 2026; scheduled tasks run a prompt on a repeat; Sites builds and hosts a small website or lightweight app from a description, in beta on paid plans above Go and unavailable in the EEA, Switzerland, and the UK; Codex, the coding agent, is folding into the desktop app; and browser-driven work is moving into ChatGPT as the standalone Atlas browser retires in August 2026.
Surfaces move fast here, so trust current documentation over a six-month-old article. Canvas, the side-by-side editing surface, has gone from the newest models; writing and code now appear as blocks in the chat.
What it is good at for advocacy orgs
Work that takes roughly 20 minutes to two hours and gets reviewed by a person afterwards:
- Drafting and revising. Supporter emails, volunteer guides, grant report sections, social copy. Give it your existing material as the starting point and the output sounds far more like you.
- Summarizing. A 40-page policy consultation into a one-page brief, or a long transcript into decisions and owners.
- Working with files. Upload a fundraising export and ask which segments lapsed fastest, then have it walk you through the answer.
- Repeatable team tasks. A custom GPT holding your style guide gives everyone the same starting point for press releases or donor updates.
- Landscape scans. Deep research suits mapping who works on an issue, or gathering background before a campaign or funding application.
Where it fits
ChatGPT and Claude overlap heavily. Both are everyday generalists, and for most staff the practical difference matters less than whether you adopt one at all. Pick a default, get people genuinely comfortable with it, and revisit later.
Two things pull towards ChatGPT: familiarity, since many staff have already used it personally, and the breadth of its ecosystem. The specialized tools in this library still apply — Gamma for presentations, coding agents for technical staff.
Watch-outs
- It can be confidently wrong. ChatGPT invents details, including statistics, quotations, and citations, and can misread a source. Give it the documents, ask it to stay within them, and check anything factual against the original.
- Keep organizational work off the free tier. Consumer plans carry different data terms, leave administrators without oversight, and sit outside the vendor agreements your organization may need to point to. On ChatGPT Business, workspace content is excluded from model training by default; consumer defaults differ. Put staff on an organizational plan before anyone touches real work, and read Data security and privacy basics and Is my data used for training?.
- EU and UK organizations have an extra step. European data residency is limited to the Enterprise and Edu plans, so a Business workspace stores data in the US under standard contractual clauses.
- Connectors widen the exposure. Linking Google Drive or a mailbox gives ChatGPT reach across everything that account can see. Fix over-broad permissions before connecting.
- Keep a human in the loop. ChatGPT drafts; someone who understands the audience and the stakes approves. That matters most for anything donor-facing, media-facing, or legal.
Getting started
A realistic first 30 minutes:
- Sign in at chatgpt.com with your work email on an organizational plan. Skip the free tier — it is the wrong place for anything real.
- Take a draft you were going to send anyway, ask for a tighter version and three subject lines, and compare against your own.
- Upload a report your team wrote and ask for a one-paragraph summary and three takeaways. A document you know well shows you quickly where the tool is strong and where it drifts.
- Create a Project with a short description of your organization and audience, then run the same drafting task inside it.
- Read your organization’s AI policy. If there isn’t one, create your AI policy belongs on your leadership agenda.
Pricing snapshot
A free tier exists, and we would steer organizations away from it: different data terms, no administrative visibility, and usage limits that interrupt real work. Evaluate on a paid plan. Individual plans (Go, Plus, Pro) raise usage limits, while Business and Enterprise add team management and administrative controls, from two seats up.
Nonprofits get a substantial discount through OpenAI for Nonprofits, which applies to the Business and Enterprise plans. What does AI actually cost? carries the current figures alongside the other vendors. Plan names and inclusions change often, so confirm before you budget.
