Gratefully Review: Pricing, Fit and Limits

Gratefully is an AI donor intelligence and stewardship platform built for nonprofits. It connects to a nonprofit’s existing donor database and answers questions about it, through an assistant called Grace. It does not replace your CRM. It reads it, along with your documents and email, and builds a searchable model of what your organization collectively knows about its donors.

That distinction matters more than anything else in this review, because the most common mistake is evaluating it against Bloomerang or Salesforce. It is not an alternative to either. It is something you run on top of one of them, and if you do not have one, there is nothing here for you yet.

What it costs

Confirmed against the vendor’s own pricing page on the review date below.

Every account starts with a 14 day free trial of Advanced, the top plan, with no card required. When it ends the account moves to the Free plan rather than locking, so the trial is a way of seeing the ceiling rather than a countdown. Extra seats on Professional and Advanced are $100 a month each, or $80 billed annually, and the vendor states that all integrations and real time syncing are included on every plan including Free. Annual billing saves 20% against monthly on each paid tier.

Publishing a price at all is worth noting in this category. Of the four dedicated AI fundraising products we looked at, two publish pricing and two do not.

What it actually does

Several things, and they are worth separating because the marketing runs them together. The list grew in the second half of 2026: it now sends approved emails, researches foundations and builds board decks, which it did not do in the summer.

It ingests. Live CRM connectors for Salesforce for Nonprofits, both Nonprofit Cloud and NPSP, Blackbaud Raiser’s Edge NXT, Bloomerang, Little Green Light and HubSpot, each syncing records, gift history and interactions. Mailchimp, Constant Contact and HubSpot supply email engagement signals, and Gmail, Google Workspace or Outlook can be connected for sending. Everything else arrives by CSV import or document upload, in PDF, DOCX, XLSX, PPTX and plain text. No plan caps donor records, connectors or syncs. The uploads matter more than they sound: the case notes, the board memos and the old strategy documents are usually where the context lives that the CRM never captured.

What it builds from that is a knowledge graph rather than a second database. It models relationships between records, so donors connect to other donors, to board members, and to the staff conversations that mention them. The vendor is explicit that nothing is migrated and that it reads your existing systems in place. That matters practically: the failure mode of most donor data projects is a migration nobody finishes.

It ranks. The assistant, named Grace, produces a prioritised portfolio each morning, with a stated reason attached to each name. The vendor describes seven signal categories behind the ranking: relationship risk, moves management progress, commitment health, giving trajectory, stewardship moments, hidden revenue and deadlines.

It answers and drafts. Natural language questions about donors, with a row of sources under every answer that opens the gift records or document behind it, and a plain statement when it cannot back something up. It drafts briefings, thank yous and handover documents in the organization’s voice, which it learns from your website at setup.

It sends, with a person’s approval. This is the largest change. Plays are outreach programmes switched on by default, such as a thank-you after a first gift and a check-in with lapsing donors. Each night Grace drafts the emails each Play calls for into a Ready to Send queue. A person reviews, edits and approves each one, and it goes out from their own Gmail, Google Workspace or Outlook mailbox with their signature. Nothing is sent without approval, and drafts honour do-not-contact and channel opt-outs. For bulk appeals it pushes a segment to Mailchimp as a draft campaign that you send from Mailchimp. Reminders work the same way: ask it to remind you in three months to thank a family, and on the day the reminder arrives with a fresh draft.

It researches foundations. Ask about a foundation or organization and Grace builds a profile from its IRS 990 filings, through ProPublica, and a bounded web search: giving, assets, leadership and focus areas, each fact cited. It does not research individuals, so it is not a wealth screening tool.

It reports. Answers can be exported as a branded Word or PDF document, a list of donors as a CSV, and a chart or summary as a PowerPoint deck built only from what is in the conversation. Board reports cover retention for any period, major gift totals, lapse risk counts and staff portfolio coverage.

It cleans. Every plan, including Free, gets a data health report: donors who appear more than once and how much giving is split across their duplicate records, donors you cannot reach, and gifts with missing dates or zero or negative amounts. Duplicates are merged by answering a few questions about how matches should be handled, and every merge can be undone.

It works inside Claude. Gratefully can be added to Claude as a connector, so a fundraiser who works in Claude can ask about donors, priorities and at-risk lists there, with answers cited to Gratefully’s records. Through Claude it can log a note, create a reminder or save a draft to Ready to Send, and it never sends email or changes the CRM. Our Claude for nonprofits guide covers the Claude side.

The claim that matters, and how to test it

The vendor’s central technical claim is that numbers are calculated and auditable rather than generated, and that every answer cites the record it came from. If true, that addresses the single largest risk in applying language models to donor data, which is a confident and wholly invented figure presented to a board.

Do not take it on trust, and do not test it on questions you cannot check. Take three donors whose history you know in detail and ask about each. Verify the answer, then follow the citation to the record. Then take one reported total and reconcile it by hand against the CRM.

The trial is 14 days of the top plan without a card, and the account moves to the free single user plan afterwards rather than locking, so the verification can continue past the fortnight on one seat. An assistant that is accurate on the cases you can verify has earned provisional trust on the ones you cannot. One that is not, has not.

Two of those seven signals, moves management progress and stewardship moments, assume you are running a defined donor cycle in the first place. If that is not a phrase your organization uses, start with moves management, which is the practice this product is built to accelerate, and our free tracker template if you want to run it before buying anything.

What each plan includes

The pricing page itemises the tiers, and the differences are in daily workload and alerts rather than in which integrations you get.

For most small organizations the real choice is between Free and Essential. Free is enough to find out whether the ranking matches your own judgement on the donors you know; Essential is the tier where one fundraiser can use it every day. The team plans make sense when several people share a portfolio and the knowledge between them is the thing at risk.

Gratefully sits on top of a donor database rather than replacing one. What those databases cost at your size, and which of them connect to it directly, is in donor management software.

Who it is for

The fit is narrower than the marketing implies, and being clear about it is more useful than a list of features.

Where it is wrong

It cannot fix your records. Everything here is built from what your CRM and documents already contain. Thin history produces thin answers delivered confidently, which is worse than an obvious blank. If your data has gaps, fix that first. It costs nothing.

It assumes a portfolio. The core output ranks who a gift officer should contact today. An organization with no defined portfolios has nothing for it to rank, and will be paying for a daily list nobody owns.

The team plans are not for the smallest organizations. At $4,788 a year for five seats, an organization under about $1m is spending a programme line, and part time development help will do more at that size. The single user plans change the arithmetic: Free costs nothing, and Essential is $948 a year billed annually for one fundraiser, which is within reach of most organizations with a development lead. The question at that size is not the price but whether the records are good enough to be worth reading; a file of 200 donors with patchy history is still better served by a spreadsheet and a calendar.

Bloomerang customers should compare it with Bloomerang’s own AI first. Bloomerang now lists predictive donor insights, which flag supporters ready to give more or at risk of lapsing, and an assistant called Penny, in beta, among its AI features. Gratefully still adds the reading of email engagement and uploaded documents and the cited answers, but the overlap is real, and the clearest fits are Little Green Light, which has no AI features of its own, Salesforce for Nonprofits, HubSpot and, now that it connects live, Raiser’s Edge NXT.

Integration coverage is good but not universal. Salesforce for Nonprofits, Raiser’s Edge NXT, Bloomerang, Little Green Light and HubSpot are covered natively, with Mailchimp, Constant Contact and HubSpot for email engagement. DonorPerfect, Neon, Virtuous, Keela and the giving platforms are not, so those organizations are in CSV import, which works but does not stay current on its own; a Givebutter connector is listed as in development. Confirm your specific system before trialling.

Its security certification is not issued yet. The site lists SOC 2 Type II with an expected date of Q4 2026, and GDPR compliance. Until the SOC 2 report exists, the controls are described in unusual detail but not yet externally attested, and a board that requires a completed audit before connecting a donor database should wait for it or ask for the auditor’s engagement letter. Ask directly about audit status and timeline, alongside the rest of the questions worth putting to any AI vendor.

It is a young product. There is less independent track record here than for the established names, and less published evidence from long deployments. The vendor reports having analysed more than $280m of giving, 440,000 gifts and 115,000 donors; those are its own figures, not audited ones. The trial exists precisely because of that, and the answer is to use it on real data rather than to rely on case studies.

If Salesforce is the system underneath, our review of Salesforce Nonprofit Cloud covers what that platform costs and where it goes wrong, which decides whether a layer on top is the right next spend.

Data handling, and what to establish in writing

Connecting any system to a donor database means handing over names, addresses, giving histories and, in the notes, a good deal that donors would consider private. This is the part of the evaluation most organizations skip, and it is the part a board will ask about afterwards.

Automated PII redaction is the control to examine first, because the term is used loosely across the industry: is personal data removed before anything leaves your instance, or only redacted in outputs while the underlying processing still sees it. Gratefully now answers that in public. Its security page says every prompt passes through a gateway that replaces names, gifts, contact details and family or health details with tokens before any language model sees it, restoring them locally in the answer, and that redaction runs before data enters any index or embedding. That is the stronger of the two designs. It is the vendor’s description, so test it: ask during the trial for the audit record of one query and check which fields were tokenised.

One newer feature deserves its own question. Donor Pulse, the feed of signals on each donor, includes job changes detected through People Data Labs enrichment, which Gratefully lists as on by default, and which also fills in employer and capacity context. Every external signal is labelled with its source. That is useful, and it is also donor data being matched against a third-party dataset, which some donor privacy policies do not contemplate. Read your own policy, and ask how to switch enrichment off if it conflicts.

Four things to get in writing, from this or any vendor in the category. Whether your data is used to train models, and whether you can opt out. Where the data is stored and processed. Who at the vendor can access your instance, and under what circumstances. And what the retention period is after cancellation.

Ask also whether your existing obligations reach this. Some state charitable registrations, some grant agreements and most donor privacy policies contain commitments about disclosure to third parties. A donor intelligence platform is a third party. If your published privacy policy promises donors that their information is not shared outside the organization, connecting any tool of this kind requires you to look at that promise first.

The strongest argument for it

Not the time saving, which every vendor in the category claims and which is hard to verify. The strongest argument is that the knowledge graph belongs to the organization and outlasts the people who built it.

Research from the Lilly Family School of Philanthropy, surveying 1,663 US fundraisers, found median tenure in the current job of 2 years, mean of 3.6, with 20 percent intending to leave their organization within the year. Major gift relationships routinely take longer than that to mature. The gap between how long relationships take and how long the people holding them stay is the actual problem, and it is structural.

Note that this is not exclusively a software problem. Written contact reports, a defined handover process and a rule that donor context goes in the record rather than in someone’s inbox address the same gap and cost nothing. Software helps because it lowers the effort of doing that, not because it substitutes for it.

The verdict

For an organization with a populated Salesforce, Raiser’s Edge NXT, Little Green Light, HubSpot or Bloomerang database, at least one person doing donor work, and knowledge concentrated in too few heads, this is the most reachable product in its category and the trial is genuinely low risk. The published price is a third of the nearest comparable floor, and the 14 day trial of the top plan is enough to verify the accuracy claim on your own data, and the free plan afterwards means the checking does not have to stop when it ends.

For organizations below that line, the honest answer is not yet. Get the CRM current, decide who owns donor relationships, use a free general purpose assistant for drafting, and revisit this when there is something for it to work on.