The Dreamforce 2026 Conversation Financial Services Leaders Weren’t Having on the Mainstage

What did you miss if you didn’t walk through our doors at Dreamforce this year? Besides a genuinely great spot, good food, some of the best people in this business, a trip out to Sonoma and Napa, panels that made people think instead of just nod along, and enough good jokes to make three panels over two days not feel like work. Honestly, quite a bit.

We had people check in for the welcome reception and never really leave, straight through panels the next day, into dinners, back to more panels, all the way to Dreamfest. The Atrium Lounge was, once again, the place to be at Dreamforce.

Every outlet covering Dreamforce ran a recap this year. We did too. AIforce as the new interface layer, Claudeforce in open beta, Koa as Salesforce’s first CRM reasoning model, Slackforce turning Slack into a real CRM surface, Guardian and Agent Fabric for governance, three simplified editions, Benioff comparing the whole shift to the jump from DOS to the web. If you want every announcement in one place, we already wrote that piece.

Here’s what those recaps won’t tell you, and what you missed if you weren’t in our Lounge: what came out of our three panels, four financial institutions, and a room full of women in financial services working through something harder than a product announcement.

1. AI has a permanent seat at the table.

I didn’t ask any of our panelists to agree on anything, but they did anyway: AI is here, and it has to be part of the plan. What struck me is how every one of them still put a person at the center of it. Amir Madjlessi talked about keeping a human in the loop, or better, above it, watching the whole system instead of rubber-stamping every step. Marc Sylvain from Liberty Bank put it more bluntly: human in the lead, period. And Jim Collins from Salesforce reframed the ROI question entirely: the value isn’t optimizing tasks people already do, it’s redefining the workflow so people get their time back for the parts of the job that actually matter. That’s a very different starting point than “how fast can we automate this.”

2. Starting later has one real advantage.

Four institutions in one room made the timeline impossible to ignore. Fulton Bank is six years into its data and process work and already ten agents deep, real production numbers to show for it: 80,000 hours saved, over $389 million in new loans and deposits. That’s a genuine market advantage, and it didn’t happen by accident or overnight. Conventus is mid-rebuild, already looking at where AI can layer in from day one and using it to move faster through the build itself, while the more complex AI-driven underwriting console is deliberately sequenced for later, once the full data foundation is in place. Liberty Bank just went live with CRM in June and is only now turning toward Agentforce. Institutions earlier in that timeline aren’t wasting time, but they also can’t afford to treat this as optional or wait for a perfect moment to start. The honest read is that being behind has a real cost, and the way you offset it is by moving now and moving smart, using everything Fulton and the others already learned instead of relearning it yourself. That’s the actual advantage of starting later: not less urgency, just less guesswork.

3. The gap in AI adoption isn’t capability. It’s proof.

Aubree Kendall at Conventus has a name for the mismatch financial institutions are feeling: the “CSI effect.” On crime shows, DNA results solve the case in an hour. Real labs take weeks. AI creates that same expectation gap, the sense that it should be able to do anything instantly, when what it actually needs is validation against real cases, real compliance requirements, and real edge cases before it touches production. Her line stuck with me: AI was never out of reach because the models weren’t good enough. It was out of reach because the data underneath it wasn’t trustworthy yet. That reframes the internal conversation. The hard question isn’t “can we use AI,” it’s “what do we need to prove before we trust it here,” and that’s a much easier case to make to a skeptical leadership team than a vague promise of speed.

4. The foundation work sets your ceiling.

We already saw what Fulton’s six years bought them. Dollar Bank is proving the same principle from the opposite direction. Heather Maples grew their Salesforce footprint from 30 users to over 850 in under a year, and the DevOps processes, agile methodologies, and documentation standards she stood up alongside that growth weren’t overhead, they were what made the growth possible at all. Her advice to other banks was simple: be methodical, crawl, walk, run. Two banks, two different starting points, one answer: the foundation isn’t the boring prerequisite to the AI work. It is the work.

5. Every institution invested in change management on purpose.

Dollar Bank talked about it. Liberty Bank talked about it. Fulton Bank talked about it. Conventus talked about it. We didn’t set out to find four institutions that happened to care about change management, but every one of them raised it unprompted as core to how they got this far. And when we ran our Women in Financial Services event separately, with a completely different room and a completely different agenda, change management came back as the most common theme people wanted to talk about, without anyone steering it there.

Every single person across both rooms agreed on one thing underneath all of it: the pace of change right now is unlike anything most of them have dealt with before. And when change is moving that fast, change management stops being a nice-to-have and becomes the thing that determines whether people get left behind. That means being transparent about how roles are actually shifting, showing people where they can upskill, where they might get repositioned into something new, and how they could end up leading the institution in a different way than they expected to. People don’t lean into a system they don’t understand, and they don’t stay bought in if they find out how their job is changing secondhand instead of from their own leadership.

That’s not a coincidence worth glossing over. Clean data and a sequenced rollout plan get you only as far as your team is willing and able to go with you. Every institution ahead on the technology was, without exception, also deliberate about bringing people along, and the room built entirely around leadership and change confirmed it’s the piece most likely to get shortchanged everywhere else. If you’re planning your own AI work and change management isn’t already a named workstream with its own budget and owner, that’s the gap worth closing before the technology gap.

The real takeaway

Here’s what four institutions at four completely different stages, and a full room of financial services leaders, actually said.

• AI is a race worth being in, and starting a little later than Fulton just means you get to build on everything they and the others already figured out, instead of learning it the hard way yourself.
• The real gap to close is proving your own data can be trusted with something real, not what the technology is capable of.
• The foundation work you do now, whether that looks like Fulton’s six years or Dollar Bank’s first committee meeting, is what decides how far and how fast you get to go later.
• This is still a people business first. Every one of these institutions treated their people, and how they’d lead them through the change, as part of that same foundation from day one, not something to figure out after the platform was live.

None of it is flashy, and it’s exactly the kind of work that makes everything else possible. It’s what we talked about behind our doors this year, and it’s the conversation we’d genuinely love to keep having with you. Reach out to your Atrium team, or come find us at the next event.