Limits and Known Gaps

What GrantLoop doesn't do, or doesn't do well yet. We'd rather you find this on a page than in week three.

It drafts; you submit

The agent researches, decides what to pursue, and writes the application. A person still confirms the ask amount, attaches the documents, and presses submit. If a program officer emails back, that's you as well.

This is deliberate. Submitting on your behalf without review is a bad trade for something with your organization's name on it.

The agent can browse funder sites and pull application questions, and it often gets the whole set. But some funders put the real application behind an account, and there it can only see what's public before the wall.

When that happens, the fastest path is to copy the questions out of the portal and paste them to the agent — it will draft against them normally.

It only knows what it's been told or can find

If your Reference Materials are empty, the agent works from what's publicly findable about your organization, and drafts come back generic. This is the single most common cause of disappointing output, and the fix is uploading a handful of past applications.

It also doesn't know about work that happened outside GrantLoop unless you say so — an application you mailed, an award that landed, a funder who told you no. See Tracking Applications and Deadlines.

It doesn't watch your inbox

Louellen has her own email address — she writes to you when a scheduled task finishes, and you can reply and have her pick the work back up. What she cannot do is see your inbox. A funder's reply, an award letter, a rejection, a request for more information: all of it is invisible to her until you forward it to her or say so in chat. There's no bank feed either.

It thinks before it answers

The agent is tuned for quality over speed, so a real request takes time rather than returning instantly. That's the right trade for "write this application" and the wrong one for "answer this while I'm on the phone with a program officer." If you need something mid-call, expect to wait or ask something smaller.

Complex budgets need a steer

Straightforward budgets are fine. Multi-year budgets — especially ones splitting funder contribution against in-kind or applicant match across years — often need you to show the shape you want. Upload an example budget or paste the funder's template and ask it to match; output improves sharply.

Supporting materials are weaker than the application

Funders sometimes want extras: a draft agenda, a survey instrument, a logic model. The agent will produce them, but they're less polished than the application itself, and you should expect to edit.

Giving-history data lags

Grant records come from tax filings, which are published a year or two after the fact. Use them for patterns — who a funder backs, typical check size — and rely on live web research for what's open right now. See Where Our Data Comes From.

It won't tell you your odds

Whether a grant lands depends on your programs, your relationships, your track record, and the funder's priorities that year. What GrantLoop reliably changes is how many well-targeted, well-written applications you can get out the door — and volume with fit is most of the game.

For what that looks like in practice, see the Healing Meals case study.

The organizations that do best already have some grant capability and use this to multiply it. If you've never written a grant, this helps, but it isn't a substitute for the networking and judgment the work still needs.

No individual-donor pipeline

This is grant fundraising. It has nothing to say about major donors, events, or annual giving.

Usage is finite

Plans include a monthly amount of usage. A heavy research sweep or a busy autopilot week draws on it faster than everyday chat. If you're regularly running out, a larger plan is the answer rather than asking for less.