Dreamforce 2026 brought a flood of announcements, from AIforce and Koa to new Agentforce capabilities, Slackforce and deeper partnerships with Anthropic, AWS and Google Cloud. But not every announcement needs to change your Salesforce roadmap.
In this blog, we cut through the noise and look at the developments we think existing Salesforce customers should understand, what they could mean in practice, and where it’s probably too early to take action.
What you’ll find in this blog
- The biggest announcements from Dreamforce 2026
- What AIforce could change about how people interact with Salesforce
- Why Koa and the latest Agentforce developments matter
- What existing Salesforce customers should prioritise now
Dreamforce 2026: What Was Announced and What Actually Matters?
Dreamforce is very good at making everything sound like it’s about to change immediately.
New products.
New AI.
New partnerships.
New names for things you’d only just learned the last name of.
Dreamforce 2026 was no exception.
Salesforce unveiled AIforce, introduced its first CRM-specific reasoning model, expanded Agentforce, pushed Slack further towards becoming an AI workspace and announced deeper integrations with some of the biggest names in technology.
There is genuinely important stuff in there.
But if you’re running an existing Salesforce environment, the more useful question probably isn’t:
“What did Salesforce announce?”
It’s:
“Which bits actually matter to us?”
So, let’s go through the biggest developments.
1. AIforce is probably the biggest strategic shift
The headline announcement was AIforce.
Salesforce describes it as a new live interface layer designed to make the data, workflows, business logic, permissions and governance inside Salesforce available from other AI interfaces and places where people already work.
That’s a fairly big change in thinking.
Historically, Salesforce has largely been somewhere users go to do work.
You log in.
You open a record.
You update an opportunity.
You run a report.
AIforce points towards a future where Salesforce increasingly sits underneath the experience instead.
You might interact with Salesforce through Slack, Claude, another AI assistant or a custom interface, while the Salesforce platform continues to provide the trusted data, rules, permissions and processes behind the scenes.
Why does this actually matter?
Because it could eventually change one of the oldest Salesforce challenges of all:
adoption.
Instead of constantly trying to get users to work in Salesforce, Salesforce may increasingly come to the places where they already work.
That could be hugely useful.
But there’s a catch.
Those interfaces still rely on what sits underneath them.
Bad data doesn’t become good data because an AI agent accesses it.
An overly complicated process doesn’t suddenly become sensible because you access it through a different interface.
And weak governance becomes more important, not less, when AI can take action on a user’s behalf.
Our take
You probably don’t need an “AIforce project” tomorrow.
You do need to make sure your Salesforce environment is something you would actually be comfortable allowing an AI agent to use.
That means trusted data, clear processes, appropriate permissions and well-managed automation.
2. Koa: Salesforce is building AI specifically for CRM
Another major announcement was Koa, Salesforce’s first CRM reasoning model, developed with NVIDIA and built on NVIDIA Nemotron. Salesforce says it has been designed specifically to reason through complex CRM tasks and multi-step business processes.
This is interesting because much of the AI conversation so far has centred around large general-purpose models.
Koa represents a slightly different direction.
Rather than asking a general AI model to understand how selling, servicing customers and managing CRM workflows works, Salesforce is building a model specifically around those kinds of tasks.
Why does this matter?
Because enterprise AI increasingly seems to be moving towards specialisation.
A model that understands the context of CRM, business processes and Salesforce-specific workflows could potentially make agents more reliable when they’re dealing with complex operational work.
That matters much more than whether an AI can simply write a nice email.
The long-term opportunity is AI that understands enough context to help make decisions and carry out multi-stage work.
Our take
This is one to watch rather than something most customers need to act on immediately.
Salesforce is making some strong claims about Koa’s CRM reasoning performance, but those claims are currently based on Salesforce’s own benchmarking, so real-world customer results will matter more over time.
The broader direction, though, is important.
AI inside Salesforce is becoming less about generating content and more about reasoning through actual business processes.
3. Agentforce is moving from experiments to real work
Agentforce itself wasn’t the new shiny announcement this year.
And that’s arguably significant.
Salesforce’s Dreamforce 2026 announcements increasingly position Agentforce as an established layer of the platform rather than something experimental. Its official Dreamforce coverage included new long-horizon agents, expanded AgentExchange capabilities and more examples of customers deploying Agentforce across sales and service.
The conversation is changing from:
“What could an AI agent do?”
to:
“What work should we actually let one do?”
That’s a much healthier question.
Why does this matter?
Because the value of AI agents isn’t in having one.
It’s in finding processes where automation genuinely improves speed, consistency, customer experience or employee productivity.
That might be:
- qualifying or responding to inbound enquiries
- resolving straightforward service requests
- preparing information before a human interaction
- carrying out repetitive administrative work
- supporting users through a multi-stage process
The use case matters far more than the technology label.
Our take
Start with the process, not Agentforce.
If a workflow is inconsistent, poorly documented or relies on questionable data, automating it probably isn’t your first move.
Find something repetitive, reasonably structured and measurable.
Then ask whether agentic automation could make it better.
4. Slackforce shows where Salesforce thinks work is heading
Yes, Slackforce is now a thing too.
Salesforce is increasingly positioning Slack as more than a collaboration or messaging tool. Its Dreamforce coverage describes Slackforce as part of its move towards a workspace where employees and AI agents can work together.
This connects directly to AIforce.
If Salesforce becomes less dependent on a traditional CRM interface, Slack becomes one obvious place for employees to interact with Salesforce data and agents.
Why does this matter?
Because for a lot of organisations, the biggest obstacle to Salesforce adoption isn’t functionality.
It’s friction.
If somebody can ask a question, approve something or trigger a process from the place they already spend their day, that could make Salesforce much easier to use.
Our take
Don’t start planning around the word “Slackforce”.
Look at how your employees actually work.
If your organisation already lives in Slack, these developments could become very relevant.
If it doesn’t, there’s little value in forcing another interface into the mix simply because Salesforce has given it a new name.
5. Claudeforce and the partnerships might be more important than they first appear
One of the interesting themes around Dreamforce wasn’t Salesforce trying to build absolutely everything itself.
Salesforce announced Claudeforce with Anthropic before Dreamforce, alongside deeper integrations and partnerships with AWS and Google Cloud. Its official Dreamforce announcements also highlighted expanded infrastructure and agent integrations with both cloud providers.
This links back to AIforce.
Salesforce increasingly seems comfortable with the idea that customers may use another company’s AI model or interface, while Salesforce provides the underlying trusted business context.
Why does this matter?
For customers, potentially quite a lot.
It points towards greater choice rather than a future where every company has to use one Salesforce AI model in one Salesforce interface.
That could allow organisations to choose different AI tools for different use cases while keeping Salesforce as the system providing trusted customer and operational data.
Our take
This is a direction worth watching carefully.
The future enterprise stack may be much less about buying one giant AI platform and much more about connecting the right models and interfaces to trusted business systems.
Which makes your underlying architecture, integrations and governance increasingly important.
6. The “headless” Salesforce idea is getting much more real
Before Dreamforce, Salesforce had already been talking about Headless 360, allowing Salesforce functionality and data to be exposed to external agents and interfaces.
AIforce takes that idea much further.
Salesforce is essentially saying:
the Salesforce interface doesn’t have to be the only front door to Salesforce anymore.
Its own Dreamforce resources link the Headless 360 expansion directly with the broader move towards agentic experiences outside the traditional Salesforce UI.
Why does this matter?
Because over time it could change how Salesforce projects are designed.
Instead of asking:
“What Salesforce screen should the user see?”
we may increasingly ask:
“Where does this user already work, and how should Salesforce support them there?”
That’s potentially a much better starting point.
So, what should existing Salesforce customers actually do?
Probably less than the Dreamforce keynote makes you feel you should.
You don’t need to implement every new product.
You don’t need to rewrite your roadmap around every announcement.
And you certainly don’t need to adopt something simply because the word AI has appeared in front of it.
But there are a few sensible things worth doing now.
Check your data
AI agents need reliable information.
If duplicate, incomplete or inconsistent data already causes problems for your human users, it will also create problems for AI.
Review your processes
Ask whether your Salesforce automation and workflows still reflect how your organisation works today.
There’s little point automating an outdated process more efficiently.
Review permissions and governance
The more agents can access and act on business data, the more important governance becomes.
Make sure people and systems have access to what they genuinely need.
Identify worthwhile AI use cases
Start looking for repetitive, structured work where faster execution could create measurable value.
Avoid starting with:
“Where can we put Agentforce?”
Start with:
“Where are our teams wasting time?”
Get your Salesforce foundations ready
This may be the least exciting Dreamforce recommendation.
It may also be the most valuable.
The organisations that will get the most from Salesforce’s newer AI capabilities are unlikely to be those that simply adopt them first.
They’ll be the organisations whose data, processes and architecture allow them to use those capabilities well.
Our biggest takeaway from Dreamforce 2026
Dreamforce 2026 wasn’t really about one new product.
It was about Salesforce becoming less dependent on the traditional idea of Salesforce as a piece of software you log into.
Data 360 provides trusted data.
Salesforce provides the underlying processes, permissions and business logic.
Agentforce provides agents capable of carrying out work.
AIforce allows that capability to appear in more places.
And models such as Koa are being developed to reason about the business processes sitting underneath all of it.
That direction is genuinely interesting.
But the practical advice for most Salesforce customers remains surprisingly familiar:
Get the fundamentals right first.
Because whatever interface, agent or AI model you eventually use, it will still depend on the Salesforce environment underneath it.
And that’s the bit you can start improving today.
Not sure whether your Salesforce org is ready for what comes next? Get in touch with us today and we will be happy to run through your orgin more detail.

Samantha Mathie is a Senior Digital Marketing Executive at Xenogenix and has been part of the business for four years. Before joining Xenogenix, she spent more than eight years working independently as a marketing consultant.
With a background in IT, telecommunications and video conferencing, Samantha specialises in B2B technology marketing, digital campaigns, content and connected customer journeys. She is particularly skilled at turning complex subjects into clear, engaging communications.
As an author for Xenogenix, Samantha shares practical guidance on digital marketing, campaign planning, content and communicating technology effectively.











