5 min read
The Sunday she realized she was turning down good work
It is 9 on a Sunday night, and Maya is staring at a funder’s guidelines she has read four times without absorbing a word. She writes grant proposals for small nonprofits, one person, no team, and she is good at it. That is the problem. She is booked solid, a new client is waiting, and she is about to email them the sentence every freelancer dreads: “I don’t have capacity right now.” She has turned away three projects this quarter. Not because she could not do the work, but because the part before the work, the research and the blank first draft, was eating the hours she did not have.
Maya is an illustrative composite, not a single real person, but her bottleneck is one almost every solo service provider will recognize. Here is how a small, deliberate AI setup moved her from turning work away to comfortably taking it on.
Where the hours were actually going
When Maya timed herself honestly, the writing was never the slow part. A proposal took her maybe four hours to write once she knew what she wanted to say. Getting to that point took twelve. Every new grant meant reading dense funder guidelines, digging through the nonprofit’s old reports and past applications, and then sitting with a blank page trying to turn a pile of notes into a first paragraph. The research and the blank page, not the writing, were the ceiling on how many clients she could serve.
She had tried the obvious fix: templates. They failed in a predictable way. A recycled proposal reads like a recycled proposal, funders notice, and her win rate depended on each application feeling written for that specific program. Templates saved time in exactly the place where saving time cost her money. She needed something that sped up the grind without flattening the voice.
The setup that changed the math
Maya built a two part system, and neither part required any technical skill beyond copying, pasting, and writing clear instructions.
Part one was research synthesis. She started using Google’s NotebookLM, a tool that lets you upload your own documents and then ask questions answered only from those sources rather than from the open internet. Into it went the funder’s guidelines, the client’s annual report, and two of the client’s previously successful proposals. Instead of reading all of it four times, she asked direct questions: what does this funder say they care about, where has this nonprofit shown measurable impact, what eligibility rules could disqualify us. Because the answers came only from her uploaded files, they were grounded in the real documents, and each one linked back to the exact passage so she could verify it. We made the fuller case for this tool in our piece on why NotebookLM is the research assistant solo owners have been missing, and Maya’s use of it is that argument in practice.
Part two was the blank page. With her research notes in hand, she used Claude to turn a structured brief into a rough first draft. She never asked it to “write a grant proposal,” which produces generic mush. She fed it her actual notes, the funder’s priorities, and the client’s real numbers, then asked for a first pass at each section in a plain, specific voice. What came back was not sendable. It was a scaffold, a draft with the bones in the right places, which is exactly the thing that used to take her three hours to build from nothing. She got the same result from ChatGPT on the days she preferred it, the tool mattered less than the habit of giving it real material to work from.
What the change actually produced
The numbers moved in the direction that matters. Maya went from a comfortable ceiling of about three proposals a month to five, without adding hours. The twelve hours of pre-work per proposal dropped to roughly five, because the reading and the blank page, her two slowest steps, were now assisted. That freed capacity is what let her say yes to the clients she had been turning away.
The quieter win was quality. Because the mechanical part was faster, she spent more of her remaining time on the part only she could do: the narrative, the framing, the specific reason this nonprofit deserved this grant. Her drafts got more tailored, not less, which is the opposite of what the template era had done to her. Her Sunday nights came back too, a pattern we have seen again and again, most recently when a solo travel advisor handed her research to AI and reclaimed her evenings.
What you can take from Maya’s setup
Three lessons generalize beyond grant writing to almost any solo service business:
- Find your real bottleneck before you automate anything. Maya’s was not writing, it was research and the blank page. Time yourself honestly for a week. The step you dread is usually the one worth handing off, and it is rarely the step you assumed.
- Feed the AI your material, never a vague request. The difference between generic slop and a useful draft is the quality of what you put in. Real notes, real numbers, real guidelines. The tool is only as good as the brief, a point we make in our guide to writing better instructions in your first hour of prompting.
- Protect the part that is yours. Maya let AI handle the grind and kept the judgment, the voice, and the client relationship for herself. That boundary is why her work got better instead of blander. The same one source, many outputs logic drives our walkthrough of turning one blog post into a week of content.
You do not need Maya’s exact tools or her exact trade. You need her move: name the slow step, hand the mechanical part of it to AI with a genuinely detailed brief, and pour the time you save back into the work only you can do.
What is the twelve hour step hiding in your own week? Name it in the comments, and we will suggest where an AI assist would help most.
