Last verified: October 2026.
In advising animal protection organizations on their software, we have found that the old question, “Is this easy for my team to use?”, is no longer enough. More and more of the work inside a CRM (your contacts and donor database), a project management tool, or a database is done by an AI agent: your Claude or ChatGPT reading and updating records, and pulling reports together. So before you buy, ask five more questions:
- Can your AI connect to it?
- How does the vendor charge when an AI does the work?
- Could something simpler do the job for now?
- Will it fit the size you hope to reach?
- Do you need to build your own?
1. Can your AI connect to it?
This is the first thing we check. A connector is a built-in link that lets your AI read and update a piece of software for you (the technical term is an MCP server). Look the tool up in the Claude connectors directory or in ChatGPT’s apps before you sign anything.
Think twice about buying software your AI cannot reach. However well designed the screens are, a person clicking through them will not keep up with an AI agent that does the same job while you sleep. We expect organizations to move away from tools that are pleasant to edit by hand but offer no way in for AI.
If there is no connector, check whether the software has an API (a way for other programs to talk to it) or a command-line tool (a program you run by typing commands). Either can work well. Basecamp, for example, publishes an official command-line tool instead of a connector.
To use one, someone on your operations or IT team usually builds an organization-wide skill: a saved set of instructions, with a little code, that calls the API or command-line tool so that everyone’s AI can read and update the software. On a Claude Team plan, an admin can then share that skill with the whole team. Skills that run a command-line tool work in Claude Code (Claude’s app for working with files and code on your computer), but not in the regular chat app.
This route needs an owner. The skill can break when the vendor changes their API or tool, and someone on your team has to fix it. Vendors maintain their own connectors, so when a good connector exists, it is less work to look after.
One more check: some connectors only hold one account at a time. Claude’s Slack connector, for example, connects to a single workspace. If your team works across several, plan for that.
2. How does the vendor charge?
Look at how a tool charges as well as what it costs. Per-seat pricing assumes people log in to do the work, and an AI agent working on your behalf does not need a seat, so some vendors are already changing how they charge.
Be most careful with usage-based pricing, where you pay per query or per record read, often through credits. AI agents look things up often, and not always efficiently, so a bill that looks small for a human team can grow quickly once AI is doing the reading. Ask the vendor how AI and API use is billed before you commit. Our FAQ on what AI actually costs covers the AI side of the bill.
Be careful with free products too. Some free products make their money from your data, and free plans often have fewer connectors, tighter limits, and no admin controls. Before you put supporter or donor records into anything free, read what the vendor does with them, and check what happens to your data if the free plan ends. Our data security and privacy basics guide covers what to look for.
3. Could something simpler do the job for now?
Often, yes. A spreadsheet in Google Sheets or a base in Airtable is enough for many teams to start with, and both connect to AI easily. Moving your data to a bigger system used to be a painful project that needed a consultant. AI now makes most of that move much easier, so starting simple no longer locks you in.
4. Will it fit the size you hope to reach?
The classic advice still applies: do not just buy software for where you are at the moment. Buy software for the scale you hope to achieve. If you process thousands of applications a year, check that the tool can handle that comfortably. But if your organizational strategy is to reach 10,000 supporters in 2 years, you should buy software with that in mind. An enterprise system such as Salesforce is very unlikely to be outgrown, but for a team that will stay small it is usually far more cost and complexity than you need.
5. Do you need to build your own?
AI has made it easy to build custom software, and we believe there is huge potential for non-technical advocates to build custom tools to solve their needs (our guide to vibe coding for animal advocates shows how). However, we have also seen some organizations build things they did not need, or build when they should have bought. Most of the cost of custom software comes after it is built, in keeping it running. Teams with technical staff can manage that, but teams without them often end up patching the same tool for hours every week, which is time taken away from their work for animals.
Two questions help you decide between building and buying:
- Is the problem unique to you? Managing contacts, projects, and email are problems almost every organization has, and good software already exists for them. Build only where your work is unusual enough that nothing on the market fits.
- Are you the expert? Building works best when the person directing the AI knows the subject well enough to tell when the tool gets something wrong. The vibe coding guide’s example is a researcher who built a calculator around studies he had collected himself, so he could check every number it produced.
Additionally, with AI models such as Claude and ChatGPT becoming so powerful, some teams are finding that a task no longer requires software at all. For example, a recurring job that once needed a dedicated automation platform such as Make can now be a Claude scheduled task.
Our rule of thumb: do not buy or build new software until it is clear your AI cannot do the job now, or within the next six months, and the need cannot wait. A custom build can take several months, and in that time a new AI model may make it unnecessary. We apply the same thinking to our own work: we avoid taking on any single automation that would take more than two months to deliver. If you do need an automation, our FAQ on n8n, Make, Apps Script, or custom code helps you choose.
Joining your tools together with AI
You do not need one platform that does everything. Once your tools connect to AI, the AI can work across them for you. A weekly summary can pull from your project tool, your calendar, and your meeting transcripts in one go, without you buying an all-in-one system. AI still needs to know how your team uses each tool, such as which projects matter and what your labels mean. Writing that down as a skill is how you teach it. Our guide to organizing information for AI shows where to start.
