Gartner says AI will handle 80% of customer service by 2029. Only 14% of issues are self-served today. And 64% of customers already wish you’d stop using AI on them.
Three stats. One massive gap. That’s the entire AI-in-CX debate in a nutshell, and it’s exactly where James Cuthbertson, CRO at Relative Insight, starts his Nudgestock session.
Here’s the thing nobody wants to say out loud: the ambition is right and the execution is wrong. Everyone’s chasing the same 80% automation target. Almost nobody has worked out what actually belongs in it.
In this session, James Cuthbertson unpacks how CX teams can stop guessing and start proving which journeys deserve a bot and which ones need a talented human, fast. It includes a real example from a large US telco where two specific journeys were quietly wrecking NPS, CSAT and cost-per-interaction, and how finding them (not automating everything) turned the numbers around.
The question underneath all of it: if you can only get the AI decision right in a handful of places, which ones actually move the needle?
Duration: 15 mins | Available on-demand | Ideal for: CX leaders, ops directors, and anyone with a self-service target on their scorecard
What you’ll hear in this session
1. 1. The automation ambition is outrunning the automation reality
Gartner’s 80% prediction is punchy. The Lorikeet number, 14% self-service today, is the reality check. James opens with that gap deliberately, because it’s the gap most CX strategies are quietly ignoring.
2. Why “automate everything” is the wrong brief
Cost per interaction is still the boardroom obsession. CSAT, NPS and churn haven’t gone anywhere. What’s changed is the pressure: more reporting, more detail, more evidence, delivered by fewer people with less budget.
James argues the answer isn’t more automation everywhere. It’s knowing precisely what belongs in the 80% and what has to stay in the 20%, and using conversational analytics to draw that line with evidence instead of instinct.
3. The telco story: two journeys were doing all the damage
A large US telco used conversational analytics to find that most customer friction was mild and forgettable. But two journeys, duplicate billing and number porting, were tanking NPS and CSAT and sending average call length and cost per interaction through the roof every single time they showed up.
Hear how naming those two journeys let the telco automate the easy stuff with confidence (like handset changes) and route billing and porting straight to a human within a minute, and what happened to their numbers once they did.
4. Why insight, advice and action keep falling apart
James makes the case that gathering data and generating insight was never really the hard part. The failure point is what happens next: untrusted insights, no structure, not enough detail for an operational stakeholder to actually act on. Directional charts don’t move a contact centre. Defensible detail does.
5. What specialist analytics gets you that a general-purpose LLM doesn’t
Three concrete wins James walks through: tighter, more referenceable results with fewer holes for stakeholders to poke; a serious reduction in token spend versus running everything through general LLMs; and reporting that becomes a trusted, familiar part of people’s working week instead of a one-off exercise nobody quite believes.
Why most CX and AI investment stalls before it pays off
The pattern James lays out is uncomfortably familiar. Most CX teams are reacting, not deciding:
- A journey gets painful, someone throws automation at it anyway.
- A stakeholder asks for proof, someone hands over a high-level chart.
- Leadership wants AI everywhere, someone promises it without evidence of where it’ll actually work.
What’s missing is a way to prove, journey by journey, where automation genuinely reduces pain and where it just moves the pain somewhere else. Without that, you can’t defend the decision when a stakeholder pushes back. You can only hope you guessed right.
What you’ll learn
- Why the automation ambition (80% by 2029) and the automation reality (14% self-service today) are so far apart, and what that gap means for your roadmap
- How to work out what belongs in your “automate it” bucket versus your “put a human in the lead” bucket, using evidence instead of assumption
- How one US telco found the two journeys quietly wrecking their NPS, CSAT and cost per interaction, and what they did once they knew
- Why specialist analytics tools outperform general-purpose LLMs for this job, on trust, referenceability and token spend
- How to turn insight into advice that operational stakeholders actually act on, instead of directional charts nobody trusts
- How reporting becomes a repeatable, trusted part of the working week, not a one-off exercise people quietly ignore
Who should watch this session?
This is built for anyone with a self-service or automation target on their desk and no clear way to defend it yet:
- CX and customer experience leaders deciding which journeys deserve automation and which need a human in the lead
- Operations directors under pressure to cut cost per interaction without wrecking NPS and CSAT
- Insights and analytics teams being asked for evidence, not just anecdotes, on where AI is working
- Executive stakeholders who keep hearing “AI will fix it” and want to know what “it” actually is
If your organisation has an automation target and no defensible list of where it should apply first, this is your session.
