The language around customer data is changing, with customer data platforms (CDPs) being positioned as foundations for AI-native operations, real-time decisions, and autonomous service. The key question for customer experience leaders is whether this signals a meaningful shift in what technology can deliver.
The requirement for high-quality data has not changed, Richard Manthorpe, Product Director at Content Guru, told CX Today, as AI cannot deliver useful customer outcomes without accurate, connected, and governed context.
What has changed is the degree of autonomy that organizations are beginning to give AI systems, and the role customer data needs to play as automation moves from answering questions to making decisions.
Related Articles:
- Video Interview: Where Does the CDP Fit in an Agentic AI Strategy?
- Five Costly Mistakes Enterprises Make When Replacing a Legacy Contact Center Platform
- Enterprises Still Running On Outdated Platforms Face Risks They Can No Longer Afford To Ignore
From Deflection To Autonomy
Manthorpe sees the evolution of self-service and AI in four broad phases.
The earliest iteration of chatbots was limited in ambition and value. “The first generation was a glorified FAQ section,” Manthorpe said. “There was nothing transactional about it, very straightforward, normally quite poorly implemented.”
Those early bots were largely designed to reduce contact volumes. “It was a cost-saving exercise—'can we deflect that customer contact?’” Manthorpe noted.
The next phase was more useful, with connected bots able to follow defined processes and handle straightforward requests. “They could start to address some of the low-hanging fruit in the simple customer journeys,” Manthorpe said.
Some tasks should simply be easy to complete automatically. “If I need to have access to my bank account, I don’t want to have to wait until the opening hours of the bank. It’s something that’s simple and transactional.”
GenAI improved the front end of those interactions, making intent recognition and information capture feel more natural. But the workflow behind the interaction often remained fixed.
“We can be generative when we understand the intent—I’m capturing that information,” Manthorpe said. “But then I follow an exact process behind it. And that’s where most organizations are.”
The next stage is introducing AI agents. “We’re giving autonomy to AI, and they’re performing those processes as they see fit,” Manthorpe explained.
This evolution is amplifying the importance of the customer data layer. When AI is simply retrieving an answer, incomplete data is frustrating, but it becomes a business risk when AI is recommending or taking action.
AI Needs Context, Not Just Records
The basic data problem is familiar to customers. For example, they buy something online, then receive adverts urging them to buy it again, indicating that the organization has the information somewhere, but its systems are not integrated enough to use it at the right time.
“If it doesn’t know all the things about me that it needs to know, then I’m going to get recommended things that are inappropriate, but it has to know that to be trusted with full autonomy” Manthorpe said.
The risk is not limited to poor recommendations. As AI gains more freedom, unintended outcomes become harder to predict, Manthorpe said, pointing to recent examples of autonomous systems identifying unexpected routes through a task because they were not explicitly prevented from doing so.