AI Writer

Open any application and the AI Writer tab gives you a co-author for proposal text, LOIs, budget narratives, and full deliverables. It writes one section at a time, grounded in your organization, the funder, and the specific RFP — so the output is contextual, not generic.

Plan note. The AI Writer is a paid feature. On GrantHub Lite (the free tier) the AI Writer tab is hidden — Lite still gives you discovery, matching, a pipeline, and deadline alerts. See GrantHub Lite.

How it's organized

An application is built from bundles (e.g. "Letter of Inquiry", "Full Proposal") and each bundle holds sections. You pick a template — Foundation Full, Federal SF-424, Letter of Inquiry, Corporate / Civic Community Grant, Research Grant, and more — and it seeds the right sections for that funder type. Each section has a type (executive summary, statement of need, evaluation plan, budget narrative, etc.) that tells the AI what to write and in what voice.

What it can write

Every section type has its own specialist persona. A few examples:

• Executive summary, statement of need, program impact — the core narrative.
• SMART objectives, logic model, evaluation plan — the measurable, structured sections funders score hard.
• Budget narrative — pulls your actual budget line items and justifies each one.
• Letter of intent sections — opening hook, project summary, the ask.
• Federal SF-424 sections — project abstract and the four project-narrative sections.
• Corporate / civic sections — community connection, anchor-institution economic impact, and multi-year growth plan, for funders like Bank of America's Empowering Communities grant.
• Letters of support and biosketches — generated as standalone documents from a few details you supply.

If a funder asks a question that doesn't fit a standard section, add a custom section, pick the closest type, and paste the funder's exact wording into the Funder's exact question field — the AI answers that question directly.

How the AI knows your context

Each draft is assembled from several sources, automatically:

• Your organization profile — name, mission, NTEE codes, target populations, geographic focus, annual budget, year founded, EIN, past funders.
• Org Impact & Background — a free-text box at the top of the AI Writer's left panel where you keep your concrete impact in plain language (people served, outcomes, partnerships, awards, multi-year targets). It's tenant-level, so it's shared across every application, and it feeds into every section. This is the single best thing you can fill in — real numbers here turn vague prose into a compelling case. It works even if you never touch the structured Impact module.
• Your Impact module data — if you track programs, metrics, and beneficiaries on the platform, those measured figures are pulled in too (beneficiaries are summarized as counts by category — never names).
• The opportunity — funder, summary, ask range, geographic scope.
• Parsed RFP criteria — eligibility, required documents, and the funder's stated priorities (used heavily by the Priority Alignment section).
• Funder intelligence — for alignment-type sections, the AI surfaces what this funder actually funds (grant counts, average grant size, where they give) from IRS 990-PF filings, so it can speak to their giving pattern.
• Your own past winning sections (RAG) — once you submit bundles, your strongest sections become few-shot examples. New drafts learn how your organization writes this kind of section. Strictly your own tenant's history — never another org's.

Drafting, streaming, and polishing

1. Open a section and click Generate. The draft streams in live, token by token, so you see it form.
2. Below each draft the AI proposes 2–4 concrete improvements a reviewer could apply.
3. Use the Polish presets to refine without a full rewrite — Make it more specific, Shorten by 20%, Add an outcome statistic, Warmer tone, Replace abstractions with concrete examples. Polish shows a before/after preview; nothing is saved until you click Apply.
4. Bulk generate drafts every remaining section in a bundle at once.

A short generation that comes in under the target word count is automatically expanded once with grounded detail.

Quality guardrails

• Cliché scanner — the AI is told to avoid "passionate about", "cutting-edge", "leverages synergies", etc., and any that slip through are flagged on the draft.
• Word-count check — sections below or above the target range are flagged so you can tighten or expand.
• Version history — every regeneration snapshots the previous text. Open the history popover to compare or restore a previous version (the newest 20 are kept per section).

What to always verify yourself

• Specific dollar figures and outcome statistics — the AI only uses numbers present in your context, but always confirm them against your records.
• Names of partners, board members, staff — verify; don't ship a name the AI inferred.
• Promises of metrics or deliverables you can't actually meet. Read every sentence twice.

Cost & metering

On paid tiers the AI Writer is included — we absorb the LLM API cost; use it freely. Every generation is metered internally for cost tracking. GrantHub Lite does not include the AI Writer at all.

Common patterns

• Fill in Org Impact & Background first. One paragraph of real numbers improves every section you'll ever write.
• Start with the LOI even when the funder doesn't require one — it forces you to articulate your case before committing to a full proposal.
• Paste the funder's exact question for any non-standard prompt instead of fighting a generic section.
• Read aloud. Final drafts should sound like a real person wrote them. If a sentence trips you, polish or rewrite it.