Cautiously Optimistic: What Dreamforce Taught Me About AI and Nonprofits

Three takeaways from the nonprofit AI session at Dreamforce that made me feel a little less afraid — and a lot more informed

I'll be honest with you.

Like a lot of people this week, I've been watching the AI conversation with a mix of fascination and genuine unease. The pace of change feels relentless. The claims feel enormous. And for those of us who work with nonprofits — organizations that run lean, rely on relationships, and can't afford to get technology wrong — the stakes feel real.

So when I watched Salesforce's Dreamforce session "AI Agents Transform Nonprofit Impact," I wasn't sure what I was going to hear. Another round of breathless promises about what AI is going to do to everything we know?

That's not what I got.

What I got was three real organizations, talking about real results, with a refreshing amount of honesty about what it actually takes to make this work. I didn't leave a convert. But I left more informed than I arrived — and for the first time in a while, the conversation felt honest enough to take seriously.

Here's what shifted for me — and what I'm still sitting with.

1. The humans stayed in charge

This was the first thing that settled my nerves, at least a little.

Every organization in the session — RASA, YMCA, and Blue Star Families — was clear that their AI agents work alongside their people, not instead of them. The agent handles the bottleneck. The human handles the judgment, the relationship, the mission.

RASA closed a 5-6 week emergency safety response gap down to hours. YMCA freed up nearly 2,000 staff hours in nine months. Blue Star Families saved 400 hours weekly. In every single case, the staff were still there, still making decisions, still building relationships. The agent just removed the friction that was slowing them down.

I think a lot of the fear around AI — and I include my own — comes from imagining a future where the technology replaces the people. What I saw in this session didn't really look like that. It looked like a tool that gives people more time to do the work only they can do.

For conservation organizations especially, that reframe matters. Your mission is served by people who care deeply about wildlife, the land, and the environment. An agent that handles data entry or generates reports doesn't change that. It gives those people more time to actually be out in the world doing the work.

That said — I'm aware that three curated success stories at a company's own conference aren't the full picture. I'm holding this one with some room for nuance.

2. The best results came from listening first

This one hit close to home.

Jeanine from Blue Star Families shared something I think every nonprofit needs to hear before they touch any AI tool: the most important thing they did wasn't choosing the technology. It was listening to their actual end users.

They didn't assume what the problem was. They didn't start with the solution and work backward. They asked their staff what was actually making their days harder. And the answer surprised them — it wasn't reports, it wasn't dashboards, it wasn't analytics. It was data entry. The daily grind of getting information into the system was the real bottleneck.

That specific, user-centered discovery led to a purpose-built agent with measurable, defensible ROI.

I've been saying this for a while in a different context: you don't start a Salesforce engagement with solutions. You start by listening. What's frustrating your team? Where are the hours going? What would they fix first if they could? The answers are almost always more specific — and more solvable — than the vague sense that things aren't working.

The same principle applies to AI. If you don't know what problem you're actually solving, the technology won't tell you. And no amount of impressive demos changes that.

3. The guardrails are real — and I needed to hear that

This was the section I needed most — and also the one I'm still sitting with.

Salesforce's Lori Freeman walked through a four-layer trust framework for deploying AI agents: data ownership, application governance, agent safeguards, and multi-surface experiences. And she was clear: this isn't optional. Organizations that jump straight into agentic technology without thinking through governance are setting themselves up for adoption failure.

The framing that stuck with me was this: balanced guardrails matter. Not no guardrails. Not maximum restriction. Balanced. Strategy and culture alignment before deployment.

Thoughtful governance before turning anything loose on mission-critical work.

That doesn't resolve my skepticism entirely. But it tells me that at least some people building this technology are asking the right questions. That's not nothing.

For nonprofits who've been nervous about AI — and I suspect that's most of you — this is the part worth sitting with. What does your organization own? What can the agent do and not do? How are decisions reviewed? Who's accountable when something goes wrong?

Those aren't obstacles to using AI. They're the foundation that makes using it responsibly possible. And any vendor who glosses over them quickly is worth watching carefully.

What I'm actually taking away

I'm not going to pretend one Dreamforce session erased the concerns I walked in with. The pace of change is still crazy fast. The implications are still significant. And I think healthy skepticism about AI — especially in mission-critical nonprofit work — is not only reasonable, it's responsible.

But I came away with something I didn't have going in: a clearer picture of what thoughtful deployment actually looks like. Real organizations. Specific problems. Measurable results. A governance framework that takes data ownership seriously.

That doesn't mean every AI implementation will go well. It doesn't mean the guardrails will always hold. It means that when they do go well, it's because someone did the hard, unglamorous work first — knowing their data, knowing their users, knowing what they were actually trying to solve.

For wildlife and conservation nonprofits thinking about where AI fits: stay curious and stay cautious. Those two things aren't opposites. Ask hard questions before you deploy anything. Make sure your data foundation is solid before you build on it. And don't let anyone rush you past the governance conversation — not even Salesforce.

I'm watching this space carefully. And I'll keep sharing what I see honestly — including the parts that still give me pause.

I'm more informed about where this is heading than I was before. That feels like the right place to land.

About Spruce Bloom Consulting

Spruce Bloom Consulting helps wildlife, conservation, and environmental nonprofits get more from Salesforce — with a focus on program management, automation, reporting, and fractional Salesforce operations. Based in Arvada, Colorado. Serving nonprofits remotely nationwide.

SpruceBloomConsulting.com