Agentforce, Agentforce, Agentforce. Salesforce’s flagship AI tool dominates conversation in the ecosystem, and the CRM giant continues to go all-in on enhancements, demos, and product marketing to get users on board and sell the product successfully.
But it’s no big secret that Agentforce hasn’t lived up to its full potential so far. Adoption has been unremarkable, with skepticism around maturity and pricing preventing people from investing – despite the ongoing efforts Salesforce have made to convince the masses.
On Salesforce Ben, we’ve covered the buzz around Agentforce extensively since its launch over a year ago. From industry leaders to long-time Salesforce veterans, the feedback hasn’t exactly been glowing. So why not switch gears and look at the brighter side?
With the aim of making a more bullish case for Agentforce, I spoke to Robert Sösemann, Senior Principal Architect (AI) at Aquiva Labs. where he focuses on Salesforce architecture and AI solutions.
Robert provided an optimistic outlook on where Agentforce has already delivered, how it will change Salesforce for the better, and why the payoff for users could be huge.
“No One Wants to Be the First to Take the Risk”
While we want to make a positive case for Agentforce, it would be wrong not to take a look at what the real adoption barriers have been so far.
According to Robert, this comes down to several different factors – from AI apprehensiveness to actual product readiness.
“It’s a mix,” Robert explained. “First, there’s general fear of AI and uncertainty about how to get business value from it. That’s not Salesforce-specific.
“Then there are Salesforce-specific hurdles: the technology isn’t ready, the pricing model is unclear – even partners like us don’t fully understand costs.
“And reliability is an issue. AI is non-deterministic, but Agentforce feels even more so. It’s hard to get consistent results, even with guardrails.”
When it comes to mitigating these issues, it’s now on Salesforce to continue building on their improvements and ensure purchasers have clear guidance on how to use it.
Once they do, Robert believes that adoption rates will steadily rise, as the ecosystem can really understand the business value that comes with it.
“They need to make Agentforce more reliable and testable – they’ve announced “Agent Graph” to improve this. They also need to help customers understand how to use it. The technology is still immature, but once it improves, partners like us can show more success and business value.
“I believe Salesforce will fix this in the next year, but right now, no one wants to be the first to take the risk.”
Agentforce Will Elevate the AppExchange
Most of the success stories we’ve heard about Agentforce come directly from Salesforce, often highlighting wins within their own internal teams. But what about what’s happening on the ground?
For Robert, the best bit of proof of Agentforce performing for users is with AppExchange partners.
The AppExchange is a cornerstone of the Salesforce ecosystem, driving productivity and extending what orgs can achieve. Bringing agentic AI into that mix has the potential to massively transform how apps are designed and delivered.
“Aquiva works with many, and we’ve been able to give their apps a simple AI twist – making them conversational,” Robert explained. “Instead of clicking through many screens, users can chat with the app. With Agentforce, this was straightforward: repurposing an existing API or giving the agent a basic understanding of the app. It didn’t require much extra programming, and it worked well.”
The key aspect of Robert’s example is the simplicity of it. With Agentforce, you can completely reimagine how existing apps are used. For customers, this shift from clicks to chat is an instant quality-of-life upgrade, while for partners, it demonstrates how accessible Agentforce can be when applied creatively.
Just as importantly, it sparks an imaginative question: if an app can be made conversational so easily, what else could be possible as the platform matures?
Agents Will Co-Exist With Code
One of the big ongoing conversations surrounding Agentforce is whether it’s capable of replacing existing tools like Flow, or even reducing the need for Apex development.
Salesforce Ben recently spoke to Flow expert, Tim Combridge, who argued that “Flow is not going anywhere in the Agentforce world”, but could certainly be impacted by agentic AI down the line. This is backed up by the head of automation products, Alex Edelstein, whose 5-year vision for Flow would see it having a central role in Salesforce application creation.
For Robert – a self-described “hardliner” – structured code also isn’t going anywhere, and Agentforce will only complement Apex going forward.
“I believe Agentforce will always be a mix of AI and classical code,” Robert said. “We don’t need AI to solve problems that can be written in 10 lines of testable code. Future software will look similar to today’s, just with AI parts embedded. Sometimes AI will call code; sometimes code will call AI.”
“Structured, repeatable code won’t go away, especially on Salesforce, where reuse is important. AI should do what AI does best – bridging language with systems – and code should do what it does best – precision, repeatability, testability. The challenge now is figuring out what belongs to AI and what belongs to code.”
So rather than replacing one with the other, the future looks like a fair blend between the two, where agents can provide flexibility and adaptability, while Apex ensures reliability and overall governance.
For developers generally, this perspective reframes the conversation. Instead of worrying about AI taking over, Agentforce can be seen as an accelerator – a tool that opens new possibilities without discarding the discipline and best practices that underpin good software engineering. And that in itself is the bullish case – Agentforce doesn’t replace what developers do, but instead extends and elevates it.
What Companies Need Before They Start
When it comes to implementing Agentforce, “org readiness” is a big conversation. What is actually required from companies to get started with Agentforce? What’s the biggest issue here, and what’s the best advice?
According to Robert, the answer starts with something simple that is often overlooked – clarity.
“You can only make your processes AI-enhanced when you actually understand them,” he explains. “If everything is ad hoc and everyone works differently, how can you expect an agent to follow? You need processes that are consistent and ideally documented, but the key is being able to describe them clearly to an AI.
“I don’t believe in the story some companies tell you that you need perfect data or consultants to write everything down. The important part is knowing what you do and being able to explain it.”
For companies with well-defined processes – even if they aren’t written down in minute detail – the path to experimenting with Agentforce is much clearer.
The second key ingredient is tools. Agents will struggle to operate in isolation and need something to call on that actually performs. “That could be APIs, flows, or Apex classes,” Robert says. “Without those building blocks, the agent can’t act.”
This is where technical readiness really matters – companies that already have modular, reusable automations in place will find it easier to plug Agentforce into their workflows.
In essence, Robert boils readiness down to two essentials: instructions and tools. If a company has a handle on its processes and the technical hooks in place for an agent to act, then they’re in a strong position to begin experimenting.
Can Agentforce Really Cope With Large and Complex Salesforce Orgs?
Another common sticking point with Agentforce is whether it can realistically cope with large, complex Salesforce orgs.
With thousands of objects, messy data models, and overlapping processes, the fear is that an agent will simply get lost in the noise, no matter how many updates or guidelines Salesforce delivers.
Robert agrees that this is a real challenge, especially for companies attempting to use Salesforce’s out-of-the-box agent actions.
“Agents are only as good as they understand the world,” he says. “And understanding the world in an org means you need to understand the data schema. Some orgs have thousands of objects with bad names and no descriptions. Even if they had perfect descriptions, there would be overlap and ambiguity. I think it’s really hard for any AI to figure out what you mean.”
However, Robert was also quick to point out that this doesn’t, by any means, make Agentforce unusable. “When the use case is small – when you are really just having a topic which gets something done for three custom objects and one flow – that should work even in a big org.”
In other words, complexity doesn’t have to be this massive blocker – messy orgs have always been a problem, no matter what the technology is. “Restructuring them so that you can make them AI-enhanced, yes, that will be hard. But that was always the case,” Robert explains. “In times of AI, it actually became easier because AI can help with even understanding mess.”
The key takeaway here is that big orgs may be more complicated, but Agentforce isn’t automatically excluded from them. Starting small and scoped can still deliver results, and in some ways, AI can actually make sense of complexity better than past tools ever could.
The Bullish Case: Why Optimism Is Warranted
Despite the hurdles around pricing, reliability, and adoption, Robert is clear about why companies should remain optimistic about Agentforce. For him, history shows that skepticism often comes before transformation.
“It’s often better to be an early adopter when new technology comes around,” he says. “At least in my lifetime, all the technologies that came around where people were skeptic actually became really big. I grew up in times when people said the internet is just for websites.”
Adoption speed, in his opinion, is a lot less about getting early wins with Agentforce and more about preparing your organization for what’s ahead, saying: “As speed is important in the business world, the adoption speed of AI might be important even just to prepare your organization to be AI-first.”
Robert also frames the bullish case in terms of Salesforce’s own direction. “If you think about Agentforce, you already made a decision for Salesforce,” he explains. “When you decide to go with Salesforce, then you just have to buy their story. And Salesforce, I think, is reinventing themselves into an AI company.”
And for Robert, this pivot is not optional. With the rise of competitors in the CRM market, such as Oracle offering AI-enhanced alternatives to new startups, Salesforce had to adopt an AI strategy or risk falling behind in the CRM market in the years to come. Or worse, not existing at all.
“If Salesforce did not do Agentforce, in three years there would be no Salesforce anymore,” he says plainly. “CRM is not so hard to rebuild. If they ignored AI, Salesforce would shrink. They will strive with AI or die.”
That urgency is paired with a sense of momentum. Robert sees Salesforce moving faster and becoming more agile than it has in years. “I’ve seen more change at Salesforce recently than in the last 10 years”, Robert explains. “I really believe they’re becoming less of a slow cruise ship and starting to feel agile again.”
Final Thoughts
The bullish case for Agentforce, then, rests on two points: the inevitability of AI adoption and Salesforce’s commitment to leading that shift. Companies that align with this trajectory – even if they start small today – will be better prepared for the future.
Agentforce may not be perfect today, but as Robert points out, its potential is clear, and Salesforce’s direction is warranted. From small wins with AppExchange partners to the broader shift towards AI-first business processes, the foundation is being laid for something much bigger.
A huge thank you to Robert Sösemann of Aquiva Labs for sharing his perspective. His insights highlight not just the challenges still ahead, but also the reasons to be optimistic – as well as why now might be the right time to start experimenting with Agentforce.
