Level three is where it stops
Eight hands went up. I asked a management team who believed they were ahead of their competitors when it came to AI. Every hand in the room. Then I asked which decision from the past quarter had gone differently because of AI. It stayed quiet for a while, and then someone named a pilot in the contact centre that had been running since spring.
That room is not unusual, and the gap in it has been measured. EXL surveyed 322 C-suite leaders. Seventy-six percent believed they were ahead of the competition. Around ten percent actually were.
The distance between the slide and the shop floor
Self-assessment is the least reliable instrument in your toolkit. IDC found that roughly 88 percent of proofs of concept never reach broad rollout. They work, they get applauded, and they stay where they were built. In Asia-Pacific, IBM and Ecosystm asked organisations whether they were data-driven or AI-first. Eighty-five percent said yes. Eleven percent could show it.
So the pilot in the contact centre was real. It was also the only thing that had moved, and it had moved nothing else with it.
A pilot proves that something can work. It says nothing about whether your company can carry it.
Five rungs, and a crowd on the third
Maturity runs from level 1, where everything is ad hoc, to level 5, where the system improves itself. About one in five organisations gets past level three. Level three is comfortable: there are use cases, there is a team, there is a budget line, and there is a slide. The work still depends on the people who happen to care about it.
Three questions tell you where you stand:
- Does an AI application keep running when the person who built it changes jobs?
- Can a team get the data it needs without asking a favour of someone?
- Who decides that a model goes live, and is that written down?
What the leaders rebuilt
The leaders in the EXL research changed the way the company works, not just the tooling. Forty-four percent of them had redesigned their operating model, against 23 percent of the laggards. They were also roughly 2.5 times more likely to have data available at the place where decisions are actually made.
That second number is the one I would put on the wall. Data in a warehouse is an asset. Data in the hands of the person who answers the customer at half past four on a Friday is a decision.
The return sits at the top of the ladder
ServiceNow found that pacesetters achieve an average return of around 160 percent. The pattern holds on the customer side too. High CX maturity is associated with up to 30 points of extra NPS, roughly double the loyalty, and up to a third lower service costs.
Those numbers belong to the organisations that climbed. They are not available at level three, however good the pilot looks in the quarterly review.
What you do on Monday
- Ask your team to place your organisation on the five rungs, with evidence for each claim, and do the same yourself before you compare.
- Pick the pilot you are proudest of and write down exactly what has to be true for it to serve every customer, not a sample.
- Name one decision that will be made with data this quarter, and make sure that data reaches the person making it.
- Write down who has the authority to put a model live, and who stops it.
The gap between what a boardroom believes and what it can demonstrate is measurable, and the measuring is the cheapest part. The CX+AI Maturity Check walks you through the five levels with the questions that separate a claim from a capability, so you know which rung you are actually standing on before you plan the next one.