February 09, 2026
Connectivity is Key to Agent-Driven CX, says Salesforce
AI agents are rapidly becoming central to customer experience strategies, but many organisations are struggling to turn adoption into impact. Salesforce’s 2026 Connectivity research highlights the core problem that while AI agent usage is accelerating, around 50% of agents still operate in isolated silos, rather than as part of a coordinated, multi-agent system. As well as customer confusion and lost revenue, this fragmentation leads to disconnected workflows, duplicated automations, and inconsistent decision-making.
Crucially, intelligence is not the limiting factor. Modern AI agents are capable of reasoning, summarising, predicting, and acting. The breakdown occurs when agents operate independently, without shared data, memory, or coordination across systems.
Siloed Agents Break Customer Experience
When agents lack shared context, customers feel the gaps immediately via repeating information across channels, encountering contradictory responses, or experiencing delays when automation hands off to humans. In CX, these moments erode trust faster than any single failed interaction. It is essential to combat this, particularly considering Salesforce believes AI agents are already running customer experience.
This challenge is consistently reinforced across the wider market. Cloudera reports that 96% of enterprises are expanding their use of AI agents, confirming that momentum is real and accelerating. Its findings also highlight persistent integration challenges alongside concerns around data privacy and governance. Gartner’s survey reaches a similar conclusion, finding that 45% of marketing technology leaders say AI agents offered by existing vendors fail to meet promised business performance. The root cause, Gartner argues, is not the tools themselves but a lack of underlying infrastructure and data maturity required to support agent-driven initiatives.
The Integration Issue
At the same time, AI has firmly established itself as the dominant force shaping CX strategy. The Global State of CX 2025 report shows that 35% of CX practitioners rank AI-powered technologies as the top trend influencing their work in 2025, rising from second place the year before. Yet enthusiasm is tempered by execution challenges. Demonstrating ROI, securing budget, and integrating AI with existing tools and platforms remain among the most common obstacles, mirroring the same concerns reported in 2024. Connectivity is now the hurdle.
Salesforce’s Connectivity Layers
Salesforce’s response is to frame agent success as a connectivity problem across three distinct layers, which are systems, data, and action. At the system level, MuleSoft provides the foundational plumbing, including APIs, event-driven integration, and orchestration across legacy systems, ERP, billing platforms, data warehouses, and hybrid or multi-cloud environments. This answers the first and most basic question. Can systems reliably talk to each other? However, Salesforce is careful not to claim that APIs alone resolve CX fragmentation.
That role falls to Data Cloud, now called ‘Data 360’. Salesforce describes it as a real-time, unified data layer with identity resolution, zero-copy access to external data, and governed permissions. This layer ensures agents operate with consistent, trusted customer context, rather than fragmented or stale data. Without a shared data model, even well-connected systems still produce disjointed experiences.
Agentforce adds the final layer of operational connectivity. Positioned as the execution layer above APIs and data, Agentforce enables AI agents to reason across systems, coordinate with other agents and humans, and trigger actions through workflows. Salesforce’s argument is that most CX stacks fail here, where insights exist but cannot be translated into coordinated, cross-system action.
Architecture to Outcomes
This connectivity-first approach is backed by Salesforce’s research. 96% of IT leaders agree AI agent success depends on seamless data integration, while 94% say future success requires architectures to become more API-driven. Already, 33% of teams use APIs to accelerate integration, and 50% are using APIs to connect and govern AI today.
Beyond Salesforce, the lesson is universal. Integration platforms reduce system silos, but AI agent silos persist unless organisations add shared context, orchestration, and governance on top of connectivity. Only when those elements work together do agents stop acting alone and start delivering coherent, end-to-end customer experiences.
