M.A.I. Consulting
GuidePractitioner9 October 20264 min read

Three repeatable AI workflows for small teams: grants, donor updates, meeting recaps

Brief the tool on mission, audience and constraints, feed it real inputs, and have a person who knows the facts check every claim before a grant, donor update or meeting recap goes out.

Small teams write grant applications, donor updates and meeting recaps with nobody whose job title covers them. AI can turn that from a dreaded solo task into an afternoon, but only if the tool is given real material and a person checks what comes back. A tool alone makes each step of a broken process faster, which is not the same as making the process better.

The three workflows below share one shape: brief, feed real inputs, draft in small pieces, check by someone who knows the facts.

Step zero: brief the tool before you type the task

A staff member at a small advocacy organisation asked an AI tool to "make our newsletter better" and got something polished and generic that could have belonged to any organisation. The fix was not a better prompt about newsletters. It was answering three questions first.

Without these, the tool falls back on the most ordinary version of what you asked for. That is an accurate reflection of an underspecified request, not a failure of the tool. Two minutes on these questions usually does more than any later prompt engineering. Use the same three answers at the top of each workflow below.

Workflow 1: a grant draft

A four-person community organisation had to submit six applications in a quarter, with the programme director writing them alone at nights and weekends. A four-step workflow turned a multi-day ordeal into an afternoon.

  1. Start from the funder's language. Give the tool the guidelines and priorities, and ask for an outline that mirrors the funder's structure and terms, not your internal vocabulary.
  2. Draft one section at a time against your own programme notes, so each answer rests on specifics instead of generic nonprofit language.
  3. Have a second person read for mission accuracy, not just grammar. A fluent paragraph that overstates impact or misdescribes the population served is a real risk.
  4. Keep a running file of what worked. Save sections that were funded and reuse them as reference examples next time.

The point is a process that survives whoever happens to be free that week.

Workflow 2: a donor update

A small environmental nonprofit owed major donors a quarterly update, and the executive director, not a writer, had a spreadsheet of outcomes and a folder of field photos. The tool did not write the update. It turned her raw material into a structured first draft she could fix, cut and make sound like herself.

The human pass matters most here. The draft will read smoothly and say almost nothing wrong, which is the risk: a number out of context, a programme described more broadly than it operates, an outcome that sounds bigger than it is. Read for whether every claim is exactly true. An update that overstates impact costs more trust than a plainer one.

Workflow 3: a meeting recap

Most meeting notes die in a shared document nobody reopens. The gap is the ten minutes of working out who owns what, which AI is fast at. A meeting is done when someone knows what to do next.

Before sending, check for any action assigned to someone who was not in the meeting, and any deadline that sounds invented. Both happen more than you would expect and undermine trust in the whole recap.

What to do next

Series · Working with AI day to day · part 9 of 9
Keep reading
03 ยท Team AI Training

How should AI be used in my role?

If this is the question on your desk, a thirty-minute call tells you whether the service fits, or that you do not need us yet.