Dive Brief:
- Only 6% of U.S. decision makers say all relevant systems are connected enough for employees and AI to coordinate actions across the customer journey, according to a SAP survey released last week. SAP surveyed 4,800 CX practitioners and CX decision makers in the U.S. and Germany.
- Eight in 10 CX leaders said customers experienced stalled or failed handoffs between teams or channels, get transferred to a person when the automation fails and then end up repeating themselves.
- When serving customers, three-quarters of CX leaders said their organizations’ employees must then search multiple systems, manually reconcile information or wait for another team to act before they can solve the customer issues.
Dive Insight:
Many companies are already using AI within major CX functions, from marketing and sales to fulfillment and customer service, but few have integrated those siloed AI deployments with the AI in other systems.
Companies need to integrate those silos to truly connect and minimize friction across the customer journey — but there are multiple challenges to accomplishing that goal.
“When organizations deploy isolated AI tools across departments without an integrated data and journey architecture, they don’t create an intelligent enterprise they create automated silos that pass context amnesia directly to the customer,” Behnam Behzadifar, a CX expert and senior strategist with Mofid Securities, told CX Dive in an email.
This integration applies at various steps in the journey, not just when the customer calls in for help, Jessica Keehn, CMO at SAP CX, told CX Dive in an email.
“An AI shopping assistant might recommend a specific kitchen appliance, not recognizing the item isn't available for local installation,” Keehn said.
Integrations across the tech stack are complex on multiple levels. Each departmental AI is trained on its own key performance indicators, Behzadifar said.
“Sales AI optimizes for conversion speed. Ordering AI optimizes for fulfillment efficiency. Service AI optimizes for ticket deflection,” Behzadifar said. “Without a shared cross journey orchestration layer, one AI can promise a delivery window or concession that another AI immediately violates.”
Beyond establishing integrations by application programming interfaces or, increasingly, via the Model Context Protocol — which is how AI agents access data and tools to accomplish tasks — companies also need to standardize at the data semantics level.
“Even when APIs connect systems, departments often define core entities differently — what constitutes an active dispute or a high risk churn account, for example,” Behzadifar said. “Connecting AI models on top of unaligned data taxonomies simply accelerates contradictory decisions.”
Beyond integrations, companies should also consider how their existing human-centric workflows might remap to AI-centric workflows. This might involve automating routine tasks or passing context along with the call so customers aren’t repeating themselves.
“Organizations should look at where AI can reliably handle tasks independently and where connecting those tasks into an end-to-end workflow would deliver greater value, with human involvement ready when needed,” Keehn said.
Behzadifar suggested companies adopt a risk-tiered approach to figuring out what to automate — and whether that automation should be task-centric or end to end. High frequency, low ambiguity journeys — routine, simple tasks like order status updates — are likely safe to automate end to end.
But, if a customer calls to dispute a bill or wants to cancel service, end-to-end automation may not be the best solution.
Instead, the company should consider automating “the micro tasks like data gathering, eligibility verification, compliance checks and summary drafting, while keeping a human in control of empathy, judgment and final exception approval,” Behzadifar said.
Introducing AI-powered automation at either the task or end-to-end level shouldn’t be about automation for automation's sake or shaving a few seconds off each interaction.
“Ultimately, success should be measured by whether AI resolves customer problems, reduces friction and strengthens the relationships that drive loyalty and retention,” Keehn said.