Agentforce in Amsterdam: Beyond the Fridge and into the Future

Salesforce pros have practical goals for AI agents -- not what was showcased on stage

Last week, Salesforce brought its Agentforce world tour to Amsterdam — and it delivered exactly what you’d expect: high production value, a bustling crowd, over 40 sessions packed into a single day, and a laser focus on “AI-driven, agent-first innovation.”

The Valo team was there on the ground, connecting with Salesforce architects, admins, ops leaders, and platform owners from across Europe. Everyone came with big expectations — and big questions. Because while the term “AI agent” is flying off every keynote slide and Gartner deck, the Salesforce ecosystem is still wrestling with the same question:

What the heck do we do with these things?

Let’s talk about it.

The Keynote: Hype, Fridges, and a Missed Opportunity

During the keynotes there were a few examples that Salesforce featured on AI, however, there was one example that seemed to really stick in everyone’s mind. And unfortunately, it was a problematic example.

But first, let’s give credit where it’s due: Reinier van Leuken, an AI product leader at Salesforce, delivered an engaging keynote on AI. He spoke with genuine energy, and Salesforce pulled out all the stops to make sure the room was full. Salesforce is a master of the keynote spectacle and they were top of their game in Amsterdam.

But here’s the thing:

Look, I appreciate the effort to find a simple analogy. But measuring a fridge? Is that really the killer app for powerful, autonomous AI agents interacting with our Salesforce data and processes? It felt... underwhelming. My immediate thought, and I bet I wasn't alone, was a mental image of Khaby Lame silently pointing out how much simpler it is to just use a tape measure. It wasn’t just underwhelming. It was borderline meme-worthy.

In fact, during the talk, I whispered: “What is it with fridges and new tech promotion?” It took me back to the early days of IoT. Remember the fridge that auto-orders milk? Or the one that plays Spotify and syncs your family calendar? Those use cases rarely materialized at scale because they weren't addressing genuine pain points. The most impactful IoT applications turned out to be far more industrial, logistical, or integrated into larger systems.

Yeah. We’ve seen this movie before.

The risk with examples like the fridge is that they inadvertently make this powerful technology seem trivial, or worse, like a solution desperately searching for a problem. It felt like a disconnect from the real challenges and opportunities within our complex enterprise environments.

And that’s the problem. Because while Salesforce was trying to spark imagination, what they ended up doing was unintentionally telling an audience of seasoned professionals: “AI agents are a toy.”

What People Really Want from The Agentforce World Tour

Talk to people outside the keynote — in the demo booths, the breakout rooms, the lunch lines — and a different story emerges.

The audience wants Agentforce to succeed. These folks are genuinely excited. They’re already running proofs of concept. They’ve got AI agents surfacing insights from research archives, generating draft replies for both simple and complex support tickets, monitoring internal systems for anomalies — the real deal.

So what’s missing?

Examples. Clarity. Confidence.

People aren’t asking, “What is Agentforce?” They’re asking, “How do I use it responsibly in my org?” That’s a huge shift. It means we’re past the hype curve. It’s time for substance.

Through conversations with fellow attendees at the event, it became clear that Salesforce professionals have much more practical goals for AI agents than what was showcased on stage:

• Knowledge Management Insights: A global chemical processing company described to me how they're using internal AI agents to save significant time answering questions related to their massive archive of research PDFs.
• Numerous Variations of Sophisticated Support Case Agents: Multiple organizations are deploying agents to handle various tiers of support cases—not to measure fridge widths, but to resolve issues where AI agents can respond faster than humans.
• Data Analysis Insights: Teams are exploring how agents can help business users extract insights from their Salesforce data without requiring report-building expertise, indeed, sometimes skipping the report building all together and going straight to insights.

These are the kinds of examples that resonate. And, that’s because they are real world. They address tangible business problems, leverage the unique capabilities of AI (like understanding natural language and complex data relationships), and demonstrate clear value within a Salesforce context. They aren't about replacing a tape measure; they're about augmenting human capabilities and streamlining complex workflows.

No fridges. Just real, measurable value.

The Risk of Over-Hyping the Wrong Things

When Salesforce showcases a fridge-measuring agent, it doesn’t just miss the mark — it actively creates confusion. It makes AI agents look trivial. Worse, it sends platform owners and security leaders scrambling for clarity.

We're a savvy audience; we've seen tech hype cycles before. We need substance to back up the sizzle.

I genuinely believe AI agents have enormous potential within the Salesforce ecosystem. But we need to be realistic. It's another powerful tool in our toolkit, not a magic wand for every conceivable problem. Salesforce would do well to pivot their messaging, grounding the hype in these more practical, high-impact use cases that are already emerging.

Identifying Genuine Use Cases for Salesforce Agents

All this begs the question - where do AI agents make sense?. Like many things in tech, a good framework for answering this question is by looking at use cases. AI agents aren't magic—they're tools designed to automate specific tasks that follow predictable patterns. The most successful implementations target high-volume, repetitive processes where human involvement adds limited value.

Rather than trying to reinvent your entire operation around agents, look for the low-hanging fruit: processes that are:

• Well-documented
• Rules-based
• Time-consuming
• Repetitive
• Not requiring complex judgment calls

The Agentic Era Needs Agentic Thinking

Here’s the core truth: AI agents are not just fancy chatbots.

They’re autonomous actors with real power. They can read records, make decisions, trigger workflows, send emails, access APIs, and more.

AI Agents are different enough that they require a new way of looking at availability and security. And legacy tools for availability and security? They were never designed for this.

Traditional user management assumes human context — managers, org charts, training, behavior norms. AI agents? They don’t clock out at 6. They don’t have gut instincts. They don’t raise their hand when something’s wrong.

Which is why the Salesforce ecosystem needs a new kind of readiness. Questions AI agents raise for Salesforce platform owners and admins include:

• How do you distinguish agent activity from human activity for accurate monitoring and auditing?
• How do you manage permissions effectively for agents, which often need different access levels than human users?
• How do you monitor what an agent is actually doing in real-time, beyond its static configuration? What data is it accessing? What actions is it taking?
• How do you detect when an agent goes rogue, is misconfigured, or starts leaking data in unexpected ways? These patterns of failure are different from human errors.
• What happens when the agent’s behavior drifts from what it was trained to do?
• Can I control or shut one down if it goes rogue?
• How do you quickly contain a problematic agent before it causes significant damage?
• What will these agents really do in my environment?
• How do I prevent agents from overconsuming resources and hurting availability.

These are not academic questions. These are operational imperatives in the agentic era.

Bridging the Gap: The Need for Agent Readiness

Perhaps the most overlooked aspect of adopting AI agents is environment readiness and availability. If you're introducing autonomous actors into your complex Salesforce system, your existing governance and security frameworks will be challenged in new ways.

This is where the conversation needs to evolve. We need specific capabilities – let's call it Agent Readiness – to safely integrate, operate, and oversee these non-human actors.

Based on the challenges we're seeing, true Agent Readiness requires solutions that provide:

• Visibility into Non-Human Users: The fundamental ability to clearly distinguish, monitor, and analyze agent and API activity separately from human users. Knowing who is active and how they're connected is step one.
• Misconfigurations of Agent Permissions: Spotting overly permissive setups or configurations likely to cause errors.
• Agent Posture Evaluation: Evaluating how agents impact your overall Salesforce security and compliance.
• Behavioral Monitoring for Agents: Going beyond static configurations to see what agents are actually doing moment-to-moment – including data access, data transmission, actions performed, and resources consumed.
• Agent-Specific Risk Detection: Identifying patterns unique to agent problems:
• Anomalous Activity: Detecting unusual data access, unexpected actions, or resource spikes that signal a malfunction or compromise.
• Potential Data Leakage: Monitoring interactions to flag pathways for unauthorized data exposure.
• Tailored Management & Control: Tools specifically designed for managing agent permissions (which often differ significantly from user permissions) and providing mechanisms for intervention (alerts, containment).
• Agent Containment: The critical ability to take swift action, like isolating or disabling a risky agent, before it causes significant harm to your data, processes, or business reputation.

The reality is that non-human actors are becoming increasingly prevalent in our Salesforce environments. These might be sophisticated AI agents built with Agentforce, third-party AI tools, complex low-code/no-code automations, or even simple scripts.

Crucially, it's also becoming easier to build powerful custom agents using external AI tooling and integrate them directly with Agentforce data and processes. Regardless of their origin, these agents access data, perform actions, interact with APIs, and consume resources, often autonomously.

This shift fundamentally changes the landscape of Salesforce management, security, and governance. The legacy tools and processes we rely on? They simply weren't designed with these autonomous, non-human actors in mind.

Introducing Valo AI: Making Agentforce Safe and Real

At Valo, we’ve been thinking deeply about the agentic future — not in abstract terms, but in code, dashboards, and deployments.

Valo AI is purpose-built to help Salesforce teams:

1. See What Agents Are Actually Doing

We provide visibility into non-human users so you can distinguish AI agents and integrations from people. You’ll know which agents are live, how they’re connected, and what they’re touching.

2. Monitor Behavior in Real-Time

Static configs are not enough. We track what agents actually do — what data they access, what actions they take, what resources they consume. Think of it like a Fitbit for your agents.

3. Detect Risks Early

We use intelligent detection to spot:

• Over-permissioned agents
• Unusual activity patterns
• Potential data leaks
• Misconfigurations that could open the door to exploits

4. Contain Problems Fast

If something goes sideways, Valo can intervene. Contain the agent. Limit its access. Notify the right people. All without pulling the plug on your whole environment.

5. Assess and Strengthen Your Security Posture

We help you evaluate how agents impact your overall security and compliance. Because the minute you introduce AI agents, you’re also changing the surface area of your org.

Final Thought: Hype Isn’t the Enemy — Misuse Is

We’re not against excitement. We love the energy Salesforce is putting into Agentforce.

But we also know that hype without substance creates confusion — especially among smart, experienced practitioners who are trying to do the right thing for their orgs.

AI agents are not a gimmick. They’re a paradigm shift. Let’s make sure we match the fireworks with foundation.

The most successful Salesforce teams won't be those who adopt agents the fastest—they'll be those who adopt them the most thoughtfully, with proper controls and realistic expectations. True Agentforce Environment Readiness & Availability, as we see it at Valo, is about having the specific functionalities needed to address these unique AI Agent challenges. It's about equipping your Salesforce environment – and the teams managing it – with the right surveillance, risk detection, and control mechanisms tailored for these powerful, new, non-human actors.

At Valo, we’re here for the long-term, and focused on making sure your environments are always available by building the rails and safety nets that make it possible. We’re helping companies evolve their Salesforce ecosystems to be agent-ready — with the tools, visibility, and controls needed to manage this new reality.

The agentic era is here.

Let’s lead it together — no fridge jokes required.

About Hannu

Hannu is Co-Founder and engineering leader at Valo, driving the development of an AI-powered Salesforce platform to enhance workflow efficiency and system security. He focuses on delivering scalable, secure, and innovative solutions by mentoring teams and streamlining product development. Hannu has held leadership and engineering roles at Spotify, Infrakit, and F‑Secure, where he built robust distributed systems, optimized data pipelines, and ensured high availability and security across diverse technology environments. His expertise spans software development (Python, Java, Go), cloud computing, databases, and cybersecurity, all relevant to today’s mission to deliver impactful solutions in the Salesforce ecosystem.

• 

  Hannu Varjoranta