A healthy nonprofit needs both a CRM and an intelligence layer. Salesforce Nonprofit is a great CRM. Gratefully is the AI-native intelligence layer that sits on top of it - preserving the relationship context your CRM was never built to hold.
In one sentence: Salesforce Nonprofit is where your donor records live. Gratefully is the layer that reads those records alongside everything your team knows that never made it into a field.
CRMs and intelligence layers are different categories of software. This table maps which capabilities live where.
| Capability | Salesforce Nonprofit | Gratefully |
|---|---|---|
| Donor gift history tracking | via CRM connection | |
| Unstructured data ingestion (staff notes, emails, documents) | without heavy customization | |
| Natural language query interface | Agentforce agents | |
| Answers cited to source records (no hallucinations) | ||
| Donor relationship context preserved through staff transitions | ||
| PII redaction for public content | ||
| Sits on top of existing CRM (no migration required) | is the CRM | |
| Institutional memory as a permanent org asset | ||
| Audit trail on all AI interactions | requires additional products | |
| Proactive churn risk scoring (overnight) | Einstein, configured | Action Center |
| Autonomous donor signals (wealth, life, engagement events) | ||
| Hidden revenue discovery from your own data | ||
| Setup time | Weeks to months | Under 60 minutes |
Salesforce runs two different nonprofit products, and the distinction matters here. Agentforce Nonprofit, renamed from Nonprofit Cloud in December 2025, is the current one. NPSP, the Nonprofit Success Pack, is the older managed package that installs on top of Sales Cloud, and it is a genuinely different data model rather than an earlier version of the same one. For larger organizations with the technical resources to configure and maintain it, Salesforce offers depth and customization no mid-market tool can match.
Salesforce's own AI, Einstein and now the Agentforce agents generate intelligence from structured Salesforce data - the records that live inside Salesforce fields. The institutional knowledge your team has accumulated over the years - in emails, documents, conversation notes - generally isn't there. It never was. That's not a Salesforce limitation; it's a category limitation. CRMs were built for structured data.
Gratefully is the AI-native intelligence layer designed to fill that gap. It connects to Salesforce as a data source and pairs your structured records with the unstructured knowledge that lives everywhere else - making both queryable in plain English, with every answer cited.
Any comparison written before December 2025 argues against Einstein. That is no longer the whole picture and it would be misleading to leave it there.
Salesforce now ships prebuilt AI agents with Agentforce Nonprofit. A Prospect Research agent assembles background on a prospective donor. A Participant Management agent handles programme intake and updates. Volunteer coverage and donor support agents followed in beta.
If you are already running Agentforce Nonprofit, use them. They are built for the data model you are already on, they are supported by Salesforce, and for structured work inside Salesforce records they are the shortest path available. We are not going to pretend otherwise.
The boundary is not how good Salesforce's AI is. It is what Salesforce holds. Those agents reason over the records in your Salesforce org. The reasoning behind a cultivation decision, the email thread where a donor mentioned their late husband served on the board, the handover memo from the officer who left in 2023, the note from a coffee that never got logged: none of that is in a Salesforce field, and no amount of AI on top of Salesforce can read data that was never put there.
That is the gap. It is not a criticism of Salesforce. It is a description of where the information actually sits.
Gratefully sits above Salesforce in three layers, and none of them change your CRM.
Reads from Salesforce, and from email, documents and notes, without disrupting the CRM as the system of record. Read-only by default, no managed package, no schema changes.
Resolves "Margaret Chen", "M. Chen" and "Maggie" across structured and unstructured sources into one canonical donor, carrying provenance on every assertion so you can always see where a claim came from.
A plain-English interface that answers questions, drafts outreach and surfaces proactive intelligence, always cited back to the source record or message.
Salesforce stays exactly where it is. Nothing migrates, and nothing changes about your reports, your Flows or your AppExchange integrations. The technical detail of how the connection is built, including OAuth scopes and which objects are read, is on the Salesforce integration page.
CRMs hold your structured records. Gratefully holds the relationship context that lives between the rows.
Your gift records, contact history, and engagement data are well-organized. But the relationship context - the conversation from the gala, the note about a donor's late father, the planned gift conversation that was six months in - lives in emails, notes, and the heads of the people who logged them.
David moves on. Every conversation, note, and email he logged is already part of the organization's permanent knowledge graph, sitting alongside Salesforce's records. The next person on the team asks: 'What was David working on?' - and gets a prioritized list of every active cultivation with context, history, and recommended next steps. Margaret's planned gift continues without missing a beat.
In principle you can. With enough Apex, enough custom objects, an external language pipeline, a vector store, entity resolution logic, integrations into your email systems and an audited model gateway, plus somebody to maintain all of it, the thing can be built.
The reason most organisations should not is that a system of record and a system of intelligence optimise for different things. A system of record is judged on integrity: the number in the field is correct. A system of intelligence is judged on inference quality: the answer is useful and traceable. The data shapes differ, structured tables against a resolved entity graph. The interaction differs, forms against conversation.
So the real question is not whether it can be built in Salesforce. It is whether building it is the best use of your fundraising team's time and your IT budget. For most nonprofits it is not, and the ones who try usually discover the maintenance cost rather than the build cost is what stops them.
Gratefully is built to sit on top of your existing tools - no migration, no rip-and-replace. Most organizations are set up in under 60 minutes.
Get Started →