Before you buy anything, work out which constraint you are trying to lift. The technology decision follows from that, not the other way round.

Let us break our thoughts down for you here.

Making strategic decisions about AI.

Charities get the most from AI solutions like Agentforce when they start with a constraint rather than a product. Identify where you are losing hours, where you are short of people, or what is stopping you serving more beneficiaries, then look at whether automation, Salesforce Agentforce or a simple process change is the right fix. The tools are only worth buying once you can name what they are replacing.

Most conversations we have about AI start in the wrong place. Someone has seen a demonstration, or a funder has asked what the organisation is doing about it, and the question arrives as “should we be using AI?”

It is a hard question to answer in that form. A more useful one is: if you could give three people in this organisation five hours back every week, who would they be and what would they stop doing?

Answer that, and the technology decision usually makes itself.

1. Start with the constraint, not the product

Three constraints come up again and again in not-for-profits. Work out which one you are actually facing, because each points to a different answer.

You want to grow services but cannot grow the team.

Demand is rising, the funding for delivery staff is not, and your current processes will not stretch. This is the strongest case for automation, because the value is measured in capacity rather than savings.

You are spending skilled hours on unskilled work.

Your fundraising manager is reconciling spreadsheets. Your service lead is copying case notes into a funder template. The work is necessary, but it does not need their judgement.

You are sitting on data you cannot use.

You hold years of case records, survey responses and service notes, and every time someone asks what difference you made, it takes a fortnight to answer.

If none of those sounds like you, that is a legitimate finding. Not every organisation needs to do anything about AI this year, but there are reasons you might want to think about it.

2. Growing services without growing headcount

Say your strategy commits you to supporting twice as many people in three years, with roughly the same delivery team. The instinct is to ask where you will find the staff. The better question is which parts of the current service consume staff time without needing a person.

In practice, that is usually intake, chasing, scheduling and reporting rather than the delivery itself. A referral arrives by email, and someone retypes it. An appointment needs confirming, then reminding, then rescheduling. A monthly return needs assembling from four places.

A good deal of this is straightforward Salesforce automation rather than AI: web forms that create records directly, automated reminders, approval routing, scheduled reports. It is unglamorous, and it is often where the biggest hours are hiding.

Where AI earns its place is the work that involves reading and writing. Summarising a long referral into a case record. Drafting a response that a caseworker then checks. Sorting an inbox of enquiries by urgency and topic before a human picks them up.

Impact reporting, and where Agentforce fits

Impact reporting is the ‘proof’ of your hard work.

Impact reporting is the clearest example we see of data that exists but cannot be used.

The information is genuinely there, spread across case records, service logs, survey responses, volunteer hours and free-text notes. Turning it into a report for a funder means someone reading all of it and writing a narrative, usually under deadline, usually in the evening.

Agentforce is Salesforce’s framework for building agents: assistants that work on your own data inside your own environment, following instructions you set, taking actions you have permitted. They are not a general chatbot bolted onto the side of your organisation.

For a charity or a grant-making foundation, that makes a few things practical.

  • An agent that drafts impact reports.
    It pulls the relevant records for the period, assembles the figures, and produces a first draft of the narrative sections against a specific funder’s template. Your team edits and signs it off. The work moves from writing to reviewing, which is considerably faster.
  • An agent that reads reports.
     Useful for foundations and grant makers receiving reports from grantees. It summarises each submission, pulls out the numbers against what was promised, and flags anything missing or inconsistent, so your programme team reads with a head start.
  • An agent that answers routine questions.
     Staff asking what support a beneficiary has already received, or which fund a donation belongs to, without waiting for the one person who knows the system.

The consistent pattern is that the agent produces a draft and a person approves it. That is not a limitation to work around, it is the design. Anything going to a funder, a regulator or a beneficiary should have a named human behind it.

3. Decide whether you need AI or just automation

A quick test before you spend anything. For the task you have in mind, ask:

  • How often does it happen? Weekly or daily tasks are worth automating. Something you do twice a year is rarely worth the build.
  • How much judgement does it need? If the rules can be written down, it is automation and it will be cheaper and more reliable than AI.
  • Is there unstructured text involved? Reading, summarising, drafting and classifying free text is where AI does something automation cannot.
  • What happens if it gets it wrong? Low consequence and easily spotted is a good place to start. Safeguarding decisions and grant awards are not.

One warning that applies to both. Automating a broken process only gets you to the wrong answer faster. If nobody can explain why a step exists, fix the process before you build anything on top of it.

Your data decides what is possible

Any agent is working from what you already hold. If your service records are inconsistent, if half your outcomes live in a spreadsheet, or if the same beneficiary exists three times, an agent will summarise that faithfully and confidently.

This is the honest reason we often recommend consolidation and data work before AI rather than alongside it. It is also the cheaper order to do things in. Organisations that have already brought their data into one place tend to find the AI part surprisingly quick.

Questions worth asking before you invest

“Just because you can do something, it doesn’t mean you should” – a line we use with all of our clients.

Start with the hours, not the technology

  • Whose week are we trying to give back? Name the roles. “The fundraising team” is too vague to act on; “the two people who spend Monday mornings reconciling donation exports” is something you can test.
  • What would they do with that time instead? This is the real return. More conversations with major donors, faster grant turnaround, less weekend catch-up. If you can’t picture the better use of the hours, the saving won’t show up anywhere that matters.
  • Which of these tasks repeat? AI tends to earn its place on work that happens weekly rather than annually, even when the annual thing feels more painful.

    Check whether you need technology at all
  • Would a simpler fix get you most of the way? A cleaner form, one fewer approval step, a shared template. It’s worth knowing this before you buy, because AI laid over a tangled process usually makes the tangle faster rather than smaller.

    Make sure your data can answer back
  • If you asked a question of your CRM today, would you trust the answer? Most charities sit somewhere between “mostly fine” and “we know about the duplicates.” That’s a normal starting point, and it’s useful to know which parts of your data are solid enough to build on now and which need tidying first.
  • Where is the information you’d want it to draw on? Spread across inboxes, drives and someone’s head is common, and it shapes what’s realistic in year one.

    Decide who stays in the loop
  • Who checks the output before it reaches a donor, a beneficiary or a funder? Naming that person early is what makes the tool usable rather than a risk. It also tells you which tasks are a good first candidate: drafting, summarising and sorting are easy to review; anything that goes out unseen is not a starting point.

    Think about the whole year, not the demo
  • What does it cost at your actual volumes? Ask for the numbers at your list size, your case load, your inbox, not at demonstration scale.
  • Will this still fit if your strategy shifts? Charity priorities move with funding. Tools that do one narrow job well tend to travel better than platforms that assume you’ll keep working exactly as you do now.

Questions we get asked

What is Agentforce, in plain terms?



It is Salesforce’s platform for building AI agents that work on your own data inside your Salesforce environment. You define what an agent can see, what it can do and when it must hand over to a person. For charities it is most useful for summarising records, drafting reports and answering routine internal questions.

Do we need AI, or would automation be enough?

If the task follows rules you can write down, automation is cheaper, more predictable and easier to maintain. AI is worth paying for when the work involves reading or writing free text, such as summarising case notes or drafting a report narrative.

Does this work on NPSP or only Nonprofit Cloud?

Agentforce can be added to most Salesforce orgs, including those running NPSP on a supported edition. The prebuilt nonprofit agents and actions are designed for Agentforce Nonprofit (formerly Nonprofit Cloud), so how much you get out of the box depends on which data model you are on. We confirm exactly what is available for your org during pre-discovery.

Is our data good enough to use AI?

An agent can only work from the records you hold. If your data is spread across systems, duplicated or inconsistently recorded, the answers will reflect that. Most charities benefit from consolidating and cleaning their data first, which is worth doing on its own merits.

What does Agentforce cost a charity?

Pricing for Salesforce AI capability is consumption based, so what you pay depends on how much the agent is used rather than how many staff you have. It is generally licensed separately from the donated licences available through Salesforce’s nonprofit programme. Confirm current pricing and what your existing agreement covers before you plan a budget.

Is it safe to use AI with beneficiary data?

It can be, with the right controls. An agent should only be able to see the records the user in question is permitted to see, and special category data such as health or safeguarding information needs particular care under UK GDPR. Complete a data protection impact assessment, keep a human in the approval path for anything consequential, and document the lawful basis. Treat it as a governance question as much as a technical one.

Not sure which of your constraints is the real one?

That is what our free technical pre-discovery is for. We map your current systems, processes and data before you commit any budget, and we will tell you plainly if automation would solve the problem more cheaply than AI.

Send us a message if you’d like to chat.