Observability–the New Must-Have Feature for Your AI Agents?
This week on Funnel Frontier: Salesforce’s new observability tools that show you why your agents act the way they do and how to keep them in line.
This week:
- Observability–the new must-have feature for your AI agents
- Your contact center data might be doing you more harm than good…
Stat of the Week
34% of businesses using generative AI report exceptional customer service compared to those that don’t. CRM.org
Observability–the new must-have feature for your AI agents
You can’t scale what you can’t see. And with enterprise AI adoption up 282% since 2024, Salesforce thinks it’s time to turn the lights on.
Their new observability tools—Agent Analytics, Optimization, and Health Monitoring—bring full visibility into how AI agents perform, reason, and improve across the Agentforce 360 platform.
It’s not enough to know what your agents did. You need to know why, and whether they’ll do it right again.
What’s new inside Agentforce 360
These new observability tools show how your AI agents operate, reason, and recover. This way, you won’t just be monitoring performance; you’ll be steering it.
Here’s what the new observability tools bring:
- Agent Analytics: Spot patterns, identify weak flows, and surface performance trends. Get insight into what your agents are doing and how that’s changing over time.
- Agent Optimization: See how your agents think. Follow session flows, decision chains, and reasoning paths. Fix misfires by retraining models, tightening prompts, or adjusting rules.
- Agent Health Monitoring: Think uptime and stability dashboards for your AI fleet. Get real-time alerts when agents slow down, flake out, or fail to act.
Together, these tools give you end-to-end visibility into every interaction, guardrails to catch problems before they escalate, and governance at scale—powered by MuleSoft’s Agent Fabric and Data Cloud’s session tracing.
Why this matters
Most companies have built their first agents. That’s not the hard part anymore. The challenge now is running a whole ecosystem of agents you can trust.
As Adam Evans, EVP of Salesforce AI, put it: "As AI adoption accelerates, the biggest enterprise challenge will no longer be about building an organization’s first agent; it will become how to best manage a fleet of agents that are making real-world business decisions."
What this means for Salesforce users
If you’ve got agents in the wild, you’ve probably asked yourself at some point:
- Is it working?
- Why did it say that?
- Who’s monitoring this?
This release gives you receipts. Every conversation is traceable. Every misstep is diagnosable. And every success? Repeatable.
Here’s what you can do next:
- Turn on session tracing. Make sure you’re logging every agent interaction in Data Cloud.
- Audit your agents with intent + quality scoring. Look for patterns in poor performance and batch-retrain.
- Benchmark agent health. Set uptime and responsiveness thresholds, and create alerts for failure spikes.
- Train your team to read decision traces. Knowing why an agent flubbed is the fastest way to fix it.
If you’re already deep in the Agentforce ecosystem, start wiring these tools in now. For every other AI team: let this be your blueprint. Don’t ship agents you can’t see, measure, or retrain. Observability isn’t a feature. It’s the foundation.
The Week @ CRM.org
Best Marketing CRM. Thinking of jumping on the marketing CRM bandwagon? Sounds like fun, but you might wanna check our roadmap first.
The Best Project Plan Templates. 22 project plan templates. Because who has time to work from scratch?
🌸 Weekly Bloom 🌸
“The Good Life” and Self-Actualization. Life according to humanistic psychologist Carl Rogers (1902-1987). He was a bit like an alien trying to explain emotions to robots: “No, feelings are not an error in your programming; they’re a fundamental part of finding your place in the world.”
Your contact center data might be doing you more harm than good…
It wasn’t just the free pens and awkward badge scans lighting up the Contact Centre Expo in London this year. The real buzz? A hard truth: customer service teams are data-rich, insight-poor, and flat out of time.
“There’s an overload of data and a very short amount of time,” said Natalie Higgins of NiCE. Managers spend hours piecing together dashboards, but by the time the insight is ready… the moment’s already gone.
The problem isn’t too little data. It’s too much—poorly connected, poorly surfaced, and stuck in cycles no one knows how to break.
Layer on AI, and things get messier. Every vendor promises smarter service, but the foundation is cracked.
Kim Baker of L&Q didn’t mince words:
“There’s no point launching AI if your data is not right in the first place, because where’s it going to look to answer these questions?”
Data silos and strategy gaps: the remix nobody asked for
Everyone knows the data exists to answer their big questions—what’s working, what’s broken, what’s driving churn. But turning that into a roadmap? That’s the part where everything breaks.
As NiCE’s Andrew Tucker put it: “There’s a massive time lag between ‘I want to achieve something’ and ‘here’s how you actually do it.’”
Why? Because data isn’t the problem. Connection is. And if your AI initiative is just another bot layered on top of disjointed systems, you’re just automating the confusion.
What AI can actually fix
You can’t sprinkle generative pixie dust on chaos and expect customer love. But if you give AI clean, connected data, it can:
- Digest massive, messy datasets in seconds
- Surface early signals before they become escalations
- Help ops teams stop firefighting and start forecasting
What this means for your CX stack
AI won’t save your CX if you’re still saving reports as PDFs. Before you roll out another virtual agent, fix the foundation.
Start connecting, not just collecting, your data. Link up interactions, outcomes, and context so your team isn’t stuck playing Where’s Waldo with customer history.
And then, let AI do what it’s actually good at: turning chaos into clarity. Parsing trends. Flagging outliers. Suggesting the next best action.
AI doesn’t replace human judgment. It gives it a fighting chance.
Galactic Gourmet
CRM blips from around the web
Healthcare CRM Market to Hit $27.6B by 2033. The healthcare CRM market is projected to grow 12.2% annually. AI-powered platforms now help providers personalize outreach, automate workflows, and manage virtual care at scale.
Freshworks Adds Command Center, Vertical AI Agents to Freshdesk. Freshworks launched tools to centralize customer conversations, automate industry-specific workflows, and surface real-time service insights. Early users report up to 80% query resolution by AI and major gains in agent efficiency.
Five9 Unveils New Genius AI for Routing, Quality, and Analytics. Five9 launched AI tools for 100% interaction scoring, smarter agent matching, and unified dashboards. New features like Agentic Quality Management and Genius Routing aim to turn contact centers into self-improving CX engines.
If you'd like weekly CRM news like this delivered to your inbox, subscribe to Funnel Frontier!