If you’ve been following along our executive perspective series, you’ve heard a little from our CEO Mark Bishof about where we are investing and how that innovation is accelerating, as well as what we envision for the agentic future of experience from Sid Banerjee, our chief strategy officer. 

Now I want to share how these ideas are shaping the way we build the Medallia platform for the future, which is a deliberate continuation of what we’ve built over the years. 

Medallia has always designed for whoever is consuming experience intelligence on behalf of the enterprise. For most of our history, that consumer was a human: a manager reviewing a report, an analyst querying a dashboard, a frontline employee acting on an alert. What is changing rapidly is how those human consumers are being joined by an explosion of AI agents. And in the not-too-distant future it will be increasingly unclear whether a given action in an enterprise experience workflow is being taken by a person or by an agent operating on their behalf. 

Designing for that reality is not a new direction — it’s the natural evolution of the design principles we have always held. But it changes some important elements about how we build. 

The Near Future: Where Humans and Agents Come Together

The premise that enterprise software has always been designed for a human at the other end in its strict form is no longer accurate, and is becoming less accurate every month. The more precise framing is this: we are entering a near future where, for a growing share of enterprise workflows, the question “who is taking this action?” will not have a clear answer.  

Consider what is already starting to happen — AI agents are:

• Monitoring CRM data streams and updating account health scores in real time based on customer experience signals
• Routing and resolving service interactions through workflow automation platforms without human intervention
• Adjusting personalization decisions across marketing platforms based on behavioral signals
• And increasingly, performing complex orchestration tasks through Model Context Protocol (MCP) services — a standard that allows agents to discover available tools, access governed data sources, and trigger actions across an enterprise’s full system landscape, all within a single reasoning loop

The key distinction from prior generations of automation is agency: these systems do not simply execute a predefined rule. They reason, plan, and act. And they do so continuously, at a speed and scale that no human process could match. 

The question is no longer ‘are agents part of our platform?’ It is, ‘is our platform designed for the agents that are already arriving?’.

At Medallia, our answer has to be yes — and that starts with being purposeful about what designing for agents actually requires at the infrastructure level. And that changes almost everything about how we design the platform.

What Agent-Ready Infrastructure Actually Means

Let me be specific about what designing for an agentic future actually means, because it is not as simple as exposing a REST endpoint or adding a webhook layer on top of an existing platform. It requires moving toward standardized, context-aware protocols that allow agents to dynamically discover and safely interact with your system.

Today, Medallia serves human users across enterprises that run some of the world’s most demanding experience programs: 

• 60,000 frontline teammates at one of the world’s largest banks
• 70,000 pharmacists at multinational healthcare company
• Global hospitality companies managing hundreds of properties with thousands of frontline staff 

What makes that scale possible is not the interface. It is the infrastructure underneath it: extensive experience dealing with the world’s most complex and constantly changing organizational hierarchies with precise role-based access controls that know exactly which person in a 100,000-employee organization should see which signal — without requiring manual configuration at every level. A data governance architecture that meets the compliance standards of global financial institutions, healthcare systems, and government agencies simultaneously.

Extending that infrastructure to serve AI agents means solving a set of problems that are architecturally distinct from the ones we solved for human users. 

• Human users tolerate latency. Agents do not — they operate on real-time data, or they operate incorrectly 
• Human users apply judgment to ambiguous information. Agents propagate ambiguity into downstream actions at machine speed
• Human users are naturally constrained by the interfaces presented to them. Agents need constraints enforced at the data and permissions layer, because they do not read warning messages 

And there is a more subtle problem that is easy to underestimate: agents are only as reliable as the data they consume at query time. Unlike a human analyst who carries institutional context into every decision, an agent’s reasoning is bounded by what is in its context window at the moment of action. If that context is incomplete, stale, or ambiguous, the agent will act on it confidently — and at scale. This makes data quality, completeness, and real-time availability a must-have security feature in an agent-ready platform. 

This is one of the reasons we built Insights Assistant with traceability as a first-class requirement from day one: every response is backed by real data, with inline links to the source material that generated it. That design decision was made for human users who need to trust and verify answers. But it will be equally critical for agents, because a traceable, auditable answer chain is what allows downstream systems to act on Medallia’s intelligence with confidence, rather than assumption. 

It is also why our infrastructure investments over the past 18 months have been so foundational: 10x faster processing throughput, 4.1 billion feedback signals processed with a 30% year-over-year increase, and signals now linkable to 5x more entities across the organization. These are the architectural prerequisites for real-time, always-on, contextual intelligence that agents can actually depend on. 

An AI agent doesn’t need a beautiful dashboard. It needs governed, real-time, contextually complete intelligence it can act on. Those are different design problems.

The Importance of Guardrails: Power Requires Constraints

There is an aspect of agentic AI that does not get enough attention in product conversations: agents can go rogue. Not in a science fiction sense — in a very mundane and consequential enterprise sense. An agent with access to customer engagement tooling and insufficient guardrails can send the wrong message to the wrong person at the wrong moment. An agent with access to operational systems and poorly defined boundaries can trigger a cascade of actions that are individually defensible and collectively disastrous. 

Guardrails are not a constraint on what agents can do. They are the condition that makes it safe to give agents meaningful capability in the first place. 

This is an area where Medallia’s existing enterprise architecture is genuinely difficult to replicate. The organizational hierarchy modeling, role-based access controls, and data governance frameworks we have built over the years to serve human users translate directly into the constraint layer that agents require.  The goal is that when an agent queries Medallia’s intelligence through an MCP service, it inherits the same permissions, the same data residency rules, and the same organizational boundaries that apply to the human user it is acting on behalf of. Those constraints can not be bolted on after the fact. They must be structural. 

In practice, this means that as we would expose Medallia’s intelligence to agent ecosystems — whether existing CRM, workflow automation, or customer’s own agentic infrastructure — the governance model travels with the data. The agent can do what the authorized user can do, and nothing more — by design. 

From Frontline-Ready AI™ to Agent-Ready Infrastructure

The platform work we are doing right now is happening in two simultaneous timeframes.

In the near term, Frontline-Ready AI is our current expression of AI that acts at the human layer.  

• Intelligent Summaries that compress hours of analysis into seconds 
• Root Cause Assist that identifies systemic issues before they escalate 
• Smart Response that enables frontline teams to respond to customer feedback at a scale no manual process could match
• Smart Topic Builder that replaces days of manual text analytics configuration with minutes of AI-powered setup

These capabilities are live, they are in production, and they are already changing how the world’s most complex organizations operate.

Over 650 of the world’s leading brands are now using Frontline-Ready AI features — from zero to enterprise-wide deployment in under a year, with some exciting outcomes already. AI processing requests on the Medallia platform tripled in that same period. The adoption curve is steep, and it is accelerating.

In parallel, we are building what the next generation of the platform requires. An always-on, real-time signal processing architecture that never sleeps. A data model designed for agent consumption, not just human navigation — one that surfaces the full context of a customer relationship in a single, governed retrieval, so that an agent has what it needs to reason correctly without requiring multiple round trips or additional enrichment. Integration architecture that exposes Medallia’s intelligence to the CRM workflows, workflow system processes, marketing platform engines, and custom agentic systems that thousands of our customers are building every day. 

Insights Assistant, our first conversational AI agent, deliberately sits at the intersection of both timeframes. It is designed to work as a natural language interface for human users — allowing anyone from an executive to a frontline GM to ask questions and get answers grounded in real data with full traceability. But it is also designed to be invokable programmatically by agents. That dual-use architecture is purposeful. It is the bridge between the human-facing intelligence layer we have built and the machine-facing infrastructure we are building. Every design decision we make for Insights Assistant has to work for both users simultaneously. 

Frontline-Ready AI is what we built for today. Agent-Ready Infrastructure is what we are building for tomorrow. Supporting both simultaneously is no small task — and far harder than a slide deck makes it look.

The Strategic Direction We Are Committed To

I am so excited about what this architectural commitment means and how game-changing I believe it will be for the industry. And I want to be clear about why we are making it now.

The conventional approach in enterprise software is to build for current users and extend from there. Optimize the dashboard. Improve the analytics. Add more features to the interface people already use. That is a reasonable strategy for a stable category. But this category is not stable. 

Within a relatively short timeframe, we expect the majority of value flowing through an enterprise experience platform to be consumed not by a human sitting at a screen, but by AI agents operating continuously, at machine speed, across every system that touches the customer relationship. The platform architected for that reality now will have a structural advantage that compounds for years.

The investment behind Medallia’s next chapter — backed by Blackstone, KKR, and Apollo, with $150 million in new capital and a $500 million commitment to AI innovation — is what funds this innovation approach at the scale it requires. Not because the capital creates the architecture. But because the architecture requires sustained investment, a long-term commitment to build for a future that is not yet fully visible, and the kind of enterprise trust that only comes from years of delivering at the highest standard.

We already have the trust. Now we have the investment. The platform we are building is the result of both.

What This Means Right Now

If you are a Medallia customer, here is the practical implication.

The AI capabilities you are using today — Root Cause Assist, Smart Response, Intelligent Summaries, Smart Topic Builder, Insights Assistant — are designed to deliver value to your human teams right now, while simultaneously establishing the infrastructure foundation that your AI agents will depend on tomorrow. You are building both simultaneously.

The organizational intelligence that Medallia has embedded in your deployment — the role hierarchies, the data governance, the signal routing logic, the permissions architecture — is the same infrastructure that your agentic systems will inherit through MCP services and our integration layer.

The work you have already done to embed Medallia in your operational fabric is the foundation of your agentic future. And if you are not yet where you want to be on any of those dimensions, now is the time to build those — and our team can help. In my next post, I’ll share more about the importance of having the context from across your organization to make the most of the agentic future. 

The question is not whether your organization will eventually operate with AI agents consuming and acting on experience intelligence. It will. The question is whether the platform infrastructure you are building on today is designed for that world. Ours is, and that’s an advantage Medallia customers will uniquely enjoy in the years to come.