Dreamforce 2026 runs September 15 to 17 in San Francisco. The theme is “Becoming an Agentic Enterprise.” Roughly 1,600 breakout sessions. Keynotes stream free on Salesforce+, so if you are not planning on being in person, you have a great follow-along option online.
Salesforce is going to announce a mountain of things next week, and we’re excited to hear what’s coming next. The Belmar team is going to be in San Francisco in force this year (no pun intended), and we’ve been scheming over the past few weeks on what will be announced.
So here goes our annual Dreamforce prediction blog.
First, what we already know
Three facts frame the week.
Nonprofit Cloud is now called Agentforce Nonprofit. That rename happened in October 2025, and it was not cosmetic. In December, Salesforce shipped purpose-built agents for fundraising, program management, volunteer coordination, and donor support. Every dollar of nonprofit product investment now flows through that line. NPSP has not received new features since March 2023.
Salesforce and Anthropic announced Claudeforce on August 26. Claude moves into Salesforce as a reasoning engine. Salesforce moves into Claude as a plugin with 37 prebuilt sales skills. Dario Amodei and Marc Benioff share a keynote stage next week. It is the first time Salesforce has attached its “force” suffix to another company’s product, which tells you how big a bet this is.
And there is a dedicated nonprofit keynote on the agenda: “Maximize Impact as an Agentic Nonprofit.” That is where the sector news lands. Put it on your calendar.
Now the predictions.
Prediction 1: Claudeforce expands past sales, but nonprofits do not get their own skill pack
Confidence: high
The Salesforce in Claude plugin launched with 37 sales skills. Meeting prep. Deal health review. Pipeline review. Salesforce has already said more integrations are coming across Claude, Salesforce, and Slack, and the plugin is expected to move from pilot to open beta this month.
So next week you will hear Claudeforce framed as a platform capability rather than a sales tool. Expect service skills. Expect Slack integration demos. Expect at least one line about industry clouds.
What you will not get is a nonprofit skill pack. Not in September.
Here is why that matters. Those 37 skills are shaped like a sales cycle. Opportunities, stages, forecasts, quotas. Your development team does not have a pipeline. It has a moves management process, a grant calendar, a soft credit tangle, and a board member who wants to know why the year-end appeal underperformed. Those are different objects and different logic.
The honest version of this prediction: a major gifts officer will be able to point Claude at Salesforce and get real value on day one, because prospect research and meeting prep translate cleanly. A grants manager will not, because grant compliance does not map to a sales skill.
What to do: If your team is already using Claude or another AI assistant on the side, this is the moment to get ahead of it. Decide who gets connected access to your Salesforce data through an AI tool before someone connects themselves. Permissions and field-level security are the control point. Audit them this month.
Prediction 2: The next Agentforce Nonprofit agents target grants and program outcomes
Confidence: high
The first wave of nonprofit agents went after fundraising and donor support. That was the obvious move. It is also the easiest to demo.
The next wave goes where the pain actually lives for mid-size and large nonprofits: grant lifecycle management and outcomes reporting.
The signal is in the Winter ’27 release notes, which added meaningful updates to the Fundraising and Grantmaking sections of Agentforce Nonprofit. Salesforce does not usually invest in an object model right before ignoring it. The nonprofit keynote description points at fundraising, program delivery, and stewardship as a set, not fundraising alone.
Expect an agent that monitors grant agreements, flags reporting deadlines before they become emergencies, and drafts a compliance report from live program data. Expect an outcomes agent that turns your indicator data into something a program officer can read.
If that lands, it is the most consequential nonprofit announcement of the week!
What to do: Grant compliance agents are only as good as your grant data. If your award terms live in a shared drive and your reporting deadlines live in someone’s Outlook calendar, an agent has nothing to reason over. Before you buy anything, ask a harder question: could a new hire find every reporting obligation your organization has this quarter using only Salesforce? If the answer is no, that is your Q4 project, and it is worth doing whether or not the agent ships.
Prediction 3: Salesforce leads with nonprofit ROI numbers, and you should read them carefully
Confidence: very high
Two years into Agentforce, the conversation has moved from what agents could do to what they have actually done. Dreamforce 2026 is a proof-point year across the board.
For nonprofits, the proof points are already queued up. There is a session on the agenda about how the YMCA saved thousands of staff hours with Agentforce. Salesforce has been publishing staff-hour figures from its Agents for Impact accelerator. Expect a headline number in the nonprofit keynote, probably expressed in hours returned to mission.
Take the number seriously. Then take it apart.
Staff hours saved is a real metric and a slippery one. Hours saved only convert to mission delivery if you actually redeploy the person. If your development coordinator saves six hours a week on gift entry and spends those six hours on a different administrative task, you have improved nothing except the story you tell your board.
The organizations that get value from this are the ones that decide in advance what the reclaimed capacity is for. Twenty more stewardship calls a month. Two more grant applications a quarter. A named, countable outcome.
What to do: Before you evaluate a single agent, write down the three tasks in your organization that eat the most staff time and produce the least mission value. Get actual estimates, not guesses. That list is your business case, and it is also your defense against buying an agent for a problem you do not have.
Prediction 4: Some form of AI consumption relief for nonprofits, and this is the one to watch
Confidence: medium
This is the prediction I am least sure about and most interested in.
Here is the problem Salesforce has to solve. The Power of Us program gives eligible nonprofits ten donated Enterprise Edition licenses. Genuinely generous, and it has been the on-ramp for thousands of organizations. But agent usage is metered separately. Flex Credits run $500 per 100,000 credits. Common actions consume about 20 credits each. Per-user Agentforce pricing runs well north of $100 per user per month depending on the model.
So a nonprofit can have a free CRM and still find agentic AI out of reach. That is an awkward position for a company whose entire nonprofit narrative is now built on agents, and whose 1-1-1 model is central to its brand.
I think Salesforce addresses this next week. The most likely form is a credit grant rather than a price cut: a donated Flex Credit allocation for Power of Us organizations, or an expansion of the Agents for Impact accelerator with funded implementation attached. A smaller possibility is nonprofit-specific rate card treatment.
I am putting this at medium confidence because Salesforce has been visibly reworking agent pricing all year, and companies rarely announce sector discounts while the underlying model is still moving. It may also come as a pilot for a handful of large organizations rather than a broad program.
What to do: Whatever gets announced, do not let it drive your architecture. A credit grant is a trial budget, not an operating budget. Model what your agent usage would cost at full rate before you build a workflow that depends on it.
Prediction 5: The NPSP conversation gets more direct
Confidence: medium to high
Roughly half the nonprofits I talk to are still on NPSP. It works. It is paid for. Nobody wants to touch it.
Salesforce has been careful here. NPSP remains supported. There is no migration mandate. But the direction has been unambiguous for three years, and Winter ’27 tightened it further by bringing automatic household creation and naming to Agentforce Nonprofit. That closes a real parity gap. Household naming was one of the practical reasons to stay on NPSP, and it is now gone.
I expect Dreamforce to sharpen the message. Not a sunset date, that would be a shock announcement, and Salesforce does not do those to nonprofits. More likely: improved migration tooling, a formal migration program with partner funding, or a clearer statement about what agentic capability will and will not reach NPSP orgs.
Read that last one closely. The real cost of staying on NPSP is not a support cliff. It is that the AI capability Salesforce is building will assume the Nonprofit Cloud data model. If agents cannot reason over your data structure, you are opted out of the roadmap by default.
What to do: Stop treating this as a yes or no decision and start treating it as a timing decision. Get a real number on what migration would cost your organization: data complexity, integrations, reporting rebuild, staff training. You do not have to act on it. But an executive who knows the number can make a decision in a board meeting. An executive who does not has to defer.
What I am not predicting
A few things I expect people to speculate about that I do not think will land next week.
A nonprofit-specific Claude product. The partnership is eight months from that at best.
A free agent tier. Salesforce is trying to establish agents as premium digital labor. Free undercuts the whole pricing story.
Anything that makes a messy Salesforce org work better. Every capability announced next week assumes clean data, defined processes, and a data model that reflects how your organization actually operates. That has been true of every Dreamforce announcement for a decade, and it will be true of this one.
How to work the week
You are not going to watch 1,600 sessions. Here is the short version.
Watch the nonprofit keynote. That is the one that translates to your roadmap.
Watch the Claudeforce keynote, but watch it as a strategy signal rather than a shopping list. It tells you where the platform is headed over the next two years.
Then do the one thing that actually determines whether any of this helps you: pick a single process, measure what it costs in staff time today, and decide what you would do with that time back. Nonprofits that get value from agentic AI in 2027 will not be the ones that bought the most. They will be the ones who knew exactly which problem they were solving before the keynote started.
We will publish a recap next week with what actually got announced and what it means for nonprofit teams. Including the predictions I got wrong.