Episode 72 - The Real State of Agentic AI in Retention
Tom Burrell
October 24, 2025
Did someone forward this to you? If so, click to subscribe
Welcome to episode 72 of the Retention Blueprint!
80% of teams are “using AI.”
Using co-pilots and generative AI as a sidekick in diffuse operating models.
At the same time, almost none are ready for agentic AI.
Last week, I attended one of the standout conferences of the year: Elevate ‘25.
During the conference, I had heard from two leaders at the cutting edge of AI development.
On one side, Tom Mason, former CTO of Stable Dufusion, now part of the CTO’s office at Google. Tom is at the coalface of agentic AI developments.
On the other side, Thomas Husson, one of the leading researchers in AI transformation at Forrester.
The key message was this: agentic AI will transform how we design, decide, and deliver customer experiences.
But adoption isn’t about the technology, it’s about us.
It’s about process, organisation, and people.
History shows that we tend to adopt transformational technology slowly at first, then accelerate quickly.
Currently, we’re in the early-adopter phase of using agentic AI within retention/CRM/CX functions.
But to use agentic AI well, we need the foundations in place, which many brands struggle with.
In this episode, I’ll walk through my retention maturity curve and why foundations are critical to success in the world of agentic AI.
Black Friday is coming…
📰 Top Story - The Real State of Agentic AI for Retention
Agentic AI refers to AI systems that don’t just respond to prompts or assist humans, but act autonomously, take goal-directed actions, learn from context/memory, plan multi-step workflows, and integrate across systems (not just standalone chatbots).
I wrote about this in depth over the summer in a 3-part series.
In retention, agentic AI:
Identifies indicators of customer churn and factors that drive retention, then proactively initiates personalised interventions.
Coordinates across marketing, service, and product data to optimise the customer lifecycle.
Works with minimal manual input, even autonomously executing campaigns, offers, digital experiences, service interaction or workflows.
80% of companies already use AI (e.g., gen-AI, chatbots, copilots) in what McKinsey calls horizontal use cases.
But in terms of agentic capabilities (autonomous decisioning, full process integration, memory, orchestration), adoption is still minimal, and McKinsey reports that 90% of these agentic AI projects are stuck in pilot mode.
The reality is that in many businesses, the infrastructure (data, integration, process redesign) is still not where it needs to be to enable autonomous agents.
In some cases, we are talking about basic data integration, basic personalisation, and basic CRM Marketing workflows not being in place.
The problem is that if these foundations are not in place, agentic AI will operate on limited information, making autonomous decisions on weak foundations.
One subscription leader told me yesterday that his brands are struggling with inconsistent outcomes because, in a large membership base, a 1% error rate can make a huge difference.
Many brands are still wrestling with:
Fragmented data
Shallow segmentation
Patchy personalisation
Underused automation
Weak experimentation
If an AI agent operates autonomously on incomplete information, the results are not pretty.
Where we are now
Currently, we’re in the early adopter phase of using agentic AI within retention/CRM/CX functions.
In the Q&A at Elevate, I asked Thomas how long he thought it would take for agentic AI to reach widespread adoption.
The answer was 5-7 years.
Seven years is not long, given that many large enterprise brands can take up to 2 years to implement new Martech or a new CDP, and sometimes upwards of 3 years to align teams around that new capability.
Then layering agentic AI and testing, experimentation, workflow optimisation and organisational, people and process changes that are needed mean the window to gain a competitive advantage is rapidly closing.
After 26 years of consulting to major brands, and studying and speaking with some of the most experienced leaders in the fields of Customer Experience, Retention, CRM, Loyalty, Product and Personalisation, I have started to frame an opinion on what a retention maturity curve might look like. I wrote about it in episode 27 of this newsletter.
To gain a competitive advantage with agentic AI, brands need to get to level 5 before their competitors.
Level 1 – Immature: Focus on fixing the basics. Without clean data, automation, and a working CRM foundation, agentic AI can’t even start (even today, lots of brands are still here).
Level 2 – Developing: Start integrating and segmenting. Build your data layer, unify channels, and automate simple workflows - this gives AI the signals and structure it needs to learn.
Level 3 – Evolving: Shift from rule-based to insight-driven. Deploy predictive models, experiment and let algorithms inform retention actions.
Level 4 – Maturing: Move from assistive to autonomous. Introduce early agentic AI pilots, AI that plans, acts, and optimises retention journeys in real time, under human guardrails (some brands are playing here).
Level 5 – Excelling: Achieve full orchestration. Multiple agents collaborate across marketing, service, and product to deliver seamless, hyper-personalised retention at scale, self-improving, transparent, and ethical by design (virtually no one is here yet).
This year, I consulted with a brand focused on developing AI capabilities while building foundational elements in parallel.
And in a mature business like this, it is a highly challenging, complex process.
Major credit goes to the CEO, CFO, and CCO for investing in and empowering the teams to advance this project.
The advantage of establishing these foundations now, while considering a future of agentic AI, is that it allows for building with that future in mind.
In fact, it's the approach McKinsey and others advocate.
I’d love to hear your thoughts and opinions on the maturity model and this perspective on agentic AI. Please message me directly at [email protected] or comment on my LinkedIn post to share your views.
Until next week,
Tom
P.S. What did you think of this episode? |
Reply