A new global study from Talkdesk suggests enterprises may be entering a more difficult phase of AI adoption: moving beyond deploying individual tools and proving that AI can actually complete work across the business.
According to Talkdesk’s new State of Agentic Automation in CX report, 98% of surveyed organizations have deployed AI somewhere across the customer journey. Yet only 15% are combining agentic AI with cross-departmental orchestration to resolve customer needs from end to end, while just 5% say they can quantify AI’s impact on business outcomes.
The findings point to an emerging enterprise challenge that extends beyond customer experience. As AI agents become embedded in business processes, their effectiveness increasingly depends on their ability to move across applications, data, workflows and organizational boundaries rather than operate as isolated layers of automation.
The AI Deployment Gap Is Becoming an Orchestration Gap
The research was conducted by NewtonX for Talkdesk and surveyed 252 director-level and above leaders and influencers responsible for CX, IT, operations or AI strategy at mid-market and enterprise organizations. Respondents represented North America, EMEA, Latin America and Asia-Pacific.
While AI adoption was nearly universal among those surveyed, 85% of organizations lack the orchestration capabilities needed to connect AI agents, employees, data and workflows across enterprise systems, according to the report.
That distinction is increasingly important. An AI system may successfully understand a customer request or recommend an action, but resolving that request can require interaction with multiple underlying systems and departments.
A billing query, for example, might ultimately depend on customer records, finance systems, order information and operational workflows. Without connections between those environments, AI can automate individual stages of the interaction while still leaving employees responsible for completing the underlying process.
Tiago Paiva, CEO and founder of Talkdesk, described this as a widening divide between AI activity and the operating capabilities required to translate it into measurable results.
“As organisations shift from deploying AI tools to managing AI as part of their workforce, orchestration has become the defining capability separating experimentation from measurable business impact.”
Fragmented Systems Are Limiting End-to-End Automation
The report also highlights how existing enterprise architecture can constrain increasingly sophisticated AI initiatives.
While 64% of respondents said their organizations use specialized AI agents, only 35% retain customer context as interactions move from one system to another.
Disconnected systems were identified as a technical obstacle by 45% of respondents, while 44% cited legacy infrastructure. Nearly 80% of organizations remain limited to ten or fewer AI automations.
The operational consequences can fall back on employees. According to the research, when automation cannot complete a workflow, human agents spend an average of 28% of their time switching between systems, re-entering information and searching for customer context.
For ERP and enterprise application leaders, this raises a broader question about the next stage of AI investment. Agentic systems may be capable of taking increasingly autonomous actions, but those actions still need reliable access to the systems where customer, financial and operational processes are executed.
In that environment, integration architecture, data consistency and workflow design become as important as the intelligence of the AI model itself.
AI Agents Are Starting to Be Managed as Part of the Workforce
Talkdesk’s research also suggests that organizational thinking around AI agents is beginning to change.
Nearly one in five respondents already view AI agents more as labor than technology, while 99% believe some form of hybrid human-AI workforce can deliver value.
However, governance and operating models appear to be developing more slowly.
Trust remains a significant barrier, with 52% of respondents citing confidence in AI decision-making as a primary concern. At the same time, 94% of surveyed organizations operate without AI-assisted knowledge management, according to the report.
This combination presents a governance challenge. Enterprises increasingly want AI agents to perform work rather than simply generate recommendations, but greater autonomy also requires stronger controls around the information agents use, the decisions they make and the systems they are permitted to change.
More Mature Orchestration Is Associated With Better CX Outcomes
The study groups organizations according to their maturity in agentic AI and cross-departmental orchestration, with the results suggesting a relationship between orchestration maturity and measurable customer outcomes.
Organizations combining agentic AI with cross-departmental orchestration were four times more likely to report major improvements in customer satisfaction or Net Promoter Score, according to Talkdesk.
Higher-maturity organizations were also nearly twice as likely to automate revenue-related use cases such as churn prediction and personalized recommendations. Among organizations classified as leaders, 38% reported autonomously resolving more than 40% of customer issues. None of the organizations in the lowest maturity tier reached that threshold.
Zeus Kerravala, principal analyst at ZK Research, said the findings demonstrate the difference between simply deploying AI and operationalizing it across an organization.
“Moving from AI experimentation to real execution requires an operating model where AI, people, data, and workflows operate as a unified workforce.”
Why This Matters for Enterprise Technology Leaders
The findings reinforce a shift already becoming visible across enterprise AI strategies: the next competitive question may be less about who has AI and more about whose AI can actually complete a business process.
That puts greater emphasis on the enterprise systems underneath the AI layer. Customer experience processes rarely exist independently of ERP, CRM, finance, supply chain, commerce and other operational platforms. Giving an AI agent responsibility for an outcome therefore requires more than conversational intelligence; it requires the ability to maintain context, invoke workflows and coordinate actions across those environments.
It also changes how enterprises may need to measure AI maturity. Counting deployed copilots, agents or use cases provides evidence of adoption, but says relatively little about whether those systems are reducing process friction or producing measurable business outcomes.
That measurement challenge is particularly notable given that only 5% of organizations surveyed said they can currently quantify AI’s business impact.
From AI Adoption to Enterprise Execution
The pressure to close that gap is unlikely to diminish. Talkdesk found that 83% of respondents expect autonomous issue-resolution rates to increase over the next two years.
The next phase of customer experience automation may therefore depend less on adding another AI tool and more on building the operational foundation that allows AI, people and enterprise applications to work as a coordinated system.
For CIOs and enterprise technology leaders, that makes orchestration, integration, governance and data context central to the agentic AI discussion. AI adoption may already be widespread, but the Talkdesk research suggests that turning that adoption into end-to-end execution remains a much less common capability.
ERP News Editorial Team
The ERPNews Editorial Team covers global developments in ERP (Enterprise Resource Planning), enterprise software, cloud platforms, AI, automation, and digital transformation, providing independent news and editorial analysis for senior business and technology leaders. Our reporting focuses on market signals, strategic shifts, and enterprise impact across the ERP and enterprise technology ecosystem.
For editorial inquiries, please contact:
đź“© [email protected]
