Dreamforce 2026: What Salesforce Businesses Should Actually Prepare For Before the Event

Salesforce is changing the role of the CRM itself, and Dreamforce 2026 is where that shift gets made official

For most of its history, Salesforce has been a system you log into — a place to record what already happened. That’s shifting. Agentforce, Data 360, AI agents, deeper automation, and a widening web of connected applications are reshaping the platform. It no longer just stores your business activity — increasingly, it acts on it.

That makes September 15 more than just another date on the Salesforce calendar.

When Dreamforce 2026 opens at Moscone Center, businesses will be looking at more than a new round of product announcements. They’ll be looking at where Salesforce is taking the CRM next. And they’ll be asking whether their own Salesforce environment is ready to follow.

The question isn’t what Salesforce will announce. It’s whether your organization is actually ready for what comes next.

Not in theory. In your org, with your data, your automations, your integrations, and your team.

Salesforce Is Entering a Different Phase of CRM

Salesforce has spent the last two years building toward what it calls the “Agentic Enterprise.” It’s a model where AI agents don’t just draft a follow-up email or summarize a case history — they take action inside the CRM directly. That means updating a record, triaging a support ticket, qualifying a lead, or resolving a case without a human ever opening it.

The numbers behind that push are real. In its first-quarter fiscal 2027 results, Salesforce reported that Agentforce annual recurring revenue crossed $1.2 billion, up over 200% year over year. More than half of Agentforce and Data 360 bookings came from existing customers expanding what they’d already deployed.

Salesforce’s Chair and CEO Marc Benioff has been equally direct about where the company sees the opportunity. In that same earnings release, he described agentic AI as the biggest growth opportunity in front of both Salesforce and its customers. The statement matters because it makes one thing clear: Agentforce isn’t being positioned as another feature sitting alongside the CRM. Salesforce is increasingly treating AI agents as part of the platform’s core direction.

That number tells you something specific: this isn’t a side project. Salesforce has restructured its entire platform narrative — Sales Cloud, Service Cloud, Data 360, Slack. All of it now revolves around AI agents that act, not just assist. It doesn’t tell you that your organization is ready to hand an agent that same level of autonomy. Those are two separate questions, and Dreamforce 2026 tends to answer the first one a lot louder than the second.

The Agentforce Question Isn’t “Can It Work?”

Most Agentforce demos answer a question nobody’s actually asking anymore. Can it work? Yes — Salesforce has shown that convincingly, across sales, service, and increasingly regulated industries like banking and healthcare. Multi-agent orchestration is a genuine technical step forward — a coordinated set of agents now hands off context across an entire workflow instead of one bot answering one question. Agentforce Builder’s newer Agent Script layer adds to that, giving admins tighter, more auditable control over what an agent is actually allowed to do.

The real question is different. Can your organization run it reliably, with the right guardrails, without it becoming someone’s unmanaged side project?

That’s where the adoption data gets more honest than the keynote slide. Research from S&P Global Market Intelligence and McKinsey backs that up. The share of enterprises with an AI agent actually live in production sits at around 31% as of early 2026. Gartner has gone further, forecasting that more than 40% of agentic AI projects will be cancelled by the end of 2027. Typically, that’s not because the underlying model failed — it’s because nobody defined who owns the agent, what it’s scoped to do, or what success was even supposed to look like before it launched.

The technology is moving quickly. Business adoption is a different story. The gap between the two is exactly where most Agentforce projects either take root or quietly stall.

Salesforce’s push is also becoming more industry-specific. Agentforce is increasingly being positioned around use cases in areas such as financial services and healthcare. Workflows, compliance requirements, and customer interactions in those industries look very different from a generic CRM scenario.

That’s an important shift to watch at Dreamforce 2026. An agent designed around a real industry workflow is much easier to evaluate than a generic AI demo.

Before AI, There Is a Salesforce Problem Most Companies Need to Solve

Here’s the part that rarely makes it into the pre-event coverage ahead of Dreamforce 2026: the biggest risk to an Agentforce rollout usually isn’t the AI. It’s what’s already sitting in the org underneath it.

Every long-running Salesforce instance carries some version of the same baggage. Duplicate Flows built by different admins who never compared notes. Custom Apex nobody fully documented. Integrations wired years ago that nobody’s touched since. Permission sets that quietly grant more access than anyone remembers granting. None of that is unusual. It’s normal Salesforce entropy, and most orgs carry a manageable amount of it.

It stops being manageable the moment you put an autonomous agent on top of it. A messy report just hands someone a wrong number to double-check before they act. A messy automation layer under an autonomous agent hands that agent bad instructions to act on. And it acts fast, at scale, without pausing to ask if something looks off.

AI doesn’t remove Salesforce complexity. In some cases, it exposes complexity faster, and with real consequences. The agent doesn’t know to be suspicious of a broken process the way a human eventually learns to be.

Data 360 Is Where the Agentforce Story Gets Serious

An agent is only as good as what it can see. That’s arguably the more important half of the Agentforce story. It gets far less stage time than the agent itself.

Data 360 is Salesforce’s unified layer for structured CRM data and unstructured content like call transcripts, PDFs, and knowledge articles. It’s what determines whether an agent is reasoning over accurate, current context — or a fragmented, half-updated picture of your customer. Salesforce has reported ingesting tens of trillions of records through Data 360 in recent quarters. A growing share of that moves through Zero Copy connections, which pull data directly from source systems instead of duplicating it.

That’s genuinely useful architecture. It also doesn’t fix bad data on its own. If a customer’s account record is duplicated across three systems and nobody owns data quality as a job function, that’s a problem. Data 360 will faithfully unify the mess and hand it to an agent as ground truth.

Bad data doesn’t become intelligent because AI can read it. Before any serious agent deployment, data quality isn’t a nice-to-have — it’s the prerequisite everything else depends on.

Salesforce Is Also Changing How Applications Interact With the CRM

There’s a quieter architectural shift happening underneath the agent story, and it matters most to businesses with existing integrations. Salesforce has been expanding support for the Model Context Protocol (MCP). It lets AI agents — Salesforce’s own and third-party ones — query connected tools like Tableau directly. It also lets external applications interact with Salesforce data without a human manually clicking through the UI for every task.

Per Salesforce’s own Q1 FY27 earnings release, Slack’s MCP implementation passed one million active users within six weeks of launch. That’s a fast curve for infrastructure most end users never see directly. Headless 360, Salesforce’s API-first distribution layer, has followed a similar pattern of rapid early usage since its launch this year.

If applications and agents can interact directly with Salesforce, the CRM stops being just a destination employees log into. It becomes a service other systems can work with. That’s a meaningful shift for how you think about authentication, permissions, and governance. An org that’s been lax about connected app hygiene is exactly the org that feels the most friction here. This is becoming the default way of working, not the exception.

Some of the Most Important Salesforce Changes Won’t Be the Flashiest

It’s easy to fixate on the agent demo that gets the loudest applause on stage. In practice, admin and developer productivity tooling tends to have a bigger day-to-day impact on whether Agentforce actually gets used.

Salesforce’s own Admins blog reported that response to Setup with Agentforce was overwhelmingly positive among early testers at TDX 2026. The feature lets admins create objects and troubleshoot access issues directly inside Setup, instead of hunting through documentation. The same logic applies to guided configuration tools. They let non-developers tune agent behavior in plain language, without a trip into Agentforce Builder.

The feature that saves an administrator two hours a week may create more practical value than the flashiest AI demo. It just doesn’t get the applause. Neither shows up in a highlight reel. Both determine whether what you buy at Dreamforce 2026 actually gets configured, maintained, and used six months later.

The Enterprise Demo Isn’t Always the Mid-Market Reality

Most flagship Agentforce case studies come from enterprises with dedicated AI teams and mature data infrastructure. Their implementation budgets run in the seven figures. That’s a real story — it’s just not the story of most Salesforce customers.

A mid-market business running Salesforce is working from a genuinely different starting point. Think a smaller admin team, limited development resources, and years of legacy customization. Fewer people are available to manage AI governance day to day. Fewer people are available to catch a misconfigured agent before it causes a problem. That’s not a reason to sit out the shift — it’s a reason to be honest about pace.

What does Salesforce’s AI roadmap actually look like for a company with 200 or 500 employees, not 20,000? The honest answer is usually one narrow, well-scoped use case, proven against a real baseline. Only then does anything wider get built on top of it. Phased scope beats platform-wide ambition for a team without a dedicated AI function. It’s slower, yes — but it’s also far more likely to still be running, and trusted, eighteen months from now.

What Businesses Should Check Before Dreamforce 2026

Rather than a list of sessions to bookmark, here’s the filter worth bringing into Dreamforce 2026:

Data. Is your customer data actually clean enough for an agent to reason over accurately?

Automation. Which processes are still fully manual — and which ones are automated in ways nobody’s reviewed in years?

Integrations. Which systems depend on Salesforce today, and what breaks if that connection changes?

Technical debt. What in your current org could quietly block or complicate a new Agentforce implementation?

AI use cases. Where could an agent genuinely save time on a well-defined, rules-based task — not an ambiguous judgment call?

Governance. Who controls what an AI agent can access, and what it’s allowed to do once it has that access?

Business priorities. What are the three problems you actually want Salesforce to solve this year?

That list isn’t a demand to fix everything before September. It’s a filter sharp enough to separate what’s exciting on stage from what’s actually useful for your business.

What Amroar Will Be Watching at Dreamforce 2026

Here are the signals that will tell us whether Salesforce’s AI strategy is becoming genuinely easier to implement in real organizations. Not just easier to demo:

Agentforce moving from pilots to production. Case studies that describe an agent running unattended for months, not a proof of concept from a single quarter.

Data 360 maturity. How much of the roadmap is production-ready today versus still-in-pilot messaging dressed up as availability.

AI governance. What Salesforce says about accountability, audit trails, and control when an agent gets something wrong.

MCP and connected applications. Whether this architecture gets easier or harder for a mid-sized IT team to implement safely. Not just for Salesforce’s flagship enterprise partners.

Admin and developer productivity. The unglamorous tooling that decides whether Agentforce gets properly configured or left half-set-up in a sandbox.

Realistic mid-market adoption. Whether Salesforce shows a credible path for a 200-person company, not only Fortune 100 logos.

The Real Test Comes After Dreamforce 2026

The value of Dreamforce 2026 isn’t the number of announcements you leave with. It’s what your business can actually implement afterward.

Don’t chase every new feature. Skip any technology that doesn’t have a defined use case behind it. Avoid stacking AI on top of a process nobody’s reviewed in years. Data quality shouldn’t get ignored just because the demo made it look clean. And the architecture changes happening quietly underneath the agent headlines deserve just as much attention.

If you’re evaluating Agentforce, Data 360, Salesforce automation, or your broader CRM architecture before Dreamforce 2026, that’s exactly the conversation Amroar has with clients every week. What’s realistic for your organization right now — and what should stay on the watchlist for another release cycle.

Talk to Amroar before the keynotes drop →