Site icon Kevin Waites Consulting Services

How MCP lets AI work across your nonprofit’s tools

How many tabs do you have open right now to get one donor email out the door? For a lot of small teams it looks something like this:

I call this the app shuffle, and it’s where a lot of a small team’s week disappears.

Most leaders I talk to use AI like a search engine. They ask it a question, get an answer, and then do all the work themselves, shuffling between the same four tabs. MCP is the piece that lets AI help with the shuffle itself.

What MCP is

MCP stands for Model Context Protocol. It’s an open standard that lets an AI agent connect to your other tools so it can read from them and, where you allow it, make changes in them.

Think of it as a universal plug. Instead of every software company building a custom connection to every AI tool, each one builds an MCP connection once, and any agent that speaks MCP can use it.

With those connections in place, an agent can work through a donor report from start to finish:

A good way to picture it is hiring a chief of staff. If you didn’t give them access to the CRM, the donor list or the calendar, you’d be setting them up to fail. AI is the same. Once it has access, you stop copying data into a chat window and the AI goes and gets the data itself.

Long-running work in the background

This is where it gets useful for a small team. Most of your day goes to digital paperwork, the kind of work that’s simple but slow.

Picture asking an agent: “Scan our last six months of donor emails, find the three most common questions people ask, and draft an FAQ page that answers them.” In a chatbot, that’s 20 prompts of copying and pasting back and forth. An agent with connections to your email and files can work through it in the background while you’re in a board meeting, and you come back to a finished draft.

Software is being built for agents

For years, business software has been built around screens and dashboards for people to click through. More and more, it’s being built so an agent can operate it directly. People call this headless software. The tool still does the work, but a person no longer has to drive every click. Big platforms, Salesforce included, are opening themselves up to agents this way.

Underneath, a well-built agent system has three parts:

That last part matters more as agents get more capable. If an agent can work in Salesforce, Canva and Mailchimp at the same time, someone has to make sure it doesn’t send something strange to your biggest donor, and that job stays with you. The agent gathers the data and drafts the work, and you give the yes or no and the final polish.

How to get your nonprofit ready for MCP

  1. Map one app shuffle. Pick a task that bounces between tools, like a donor email or a monthly report. Write down every tool it touches and every copy-paste step.
  2. Ask your vendors about MCP. A growing number of tools, including CRMs and office suites, offer MCP connections. Find out which of yours do.
  3. Start read-only. Let an agent read from your tools before you let it change anything in them.
  4. Hand over the whole task. Give the agent the goal instead of prompting it one step at a time, then review the finished draft.
  5. Keep the final approval. Decide which outputs need a person’s sign-off before they go out, and make that a step the agent can’t skip.

For leaders: give your AI the access you’d give a new hire

A teammate is only as good as the tools they can use. When you’re looking at AI tools, ask what they can connect to, and what happens once they’re connected.

Get those answers right and a team of three starts to work like a team of ten, with your staff spending that time on the mission.

I’ve spent the last few weeks building MCP connections into Cosly, the AI staff team I’m building for small nonprofits.

Exit mobile version