More and more Nordic boards have approved AI investment. Most are still waiting for the return. The gap rarely sits where leaders think it does – and the questions executives ask before signing off on the next phase are usually the wrong ones.
When enterprise AI projects underdeliver, the post-mortems tend to look the same. The model worked. The tools were capable. The pilot landed. Somewhere between pilot and scale, the business case quietly stopped being met – and no one is sure where it slipped.
At boost.ai, we see the same pattern repeatedly. The slippage rarely lives in the technology. It lives in the operating model wrapped around it. Three things, in particular, tend to be quietly broken: ownership of the customer journey is divided across teams who optimize for different metrics; handoffs between systems and channels are inherited rather than designed, so context drops at every transition; and the learning generated by every interaction has nowhere to land, so the same problems recur. Each one looks like a contact-center issue and is actually a governance issue.
When those questions lack clear answers at the executive level, no platform can compensate. “We are teaching our customers that we care more about protecting handle time than solving their problems,” analyst Justin Robbins of Metric Sherpa said in a recent live conversation. “They figure that out.” That, he argues, is not a contact-centre problem. It’s a board-level retention problem dressed up in operational clothing.
The pattern is most visible in voice AI deployments. Treated as procurement decisions – pick a platform, layer it onto existing routing, measure containment – they underdeliver. Treated as an operating model change – sequence the rollout, design the handoffs, build the feedback loops – they compound.
The companies that get this right rarely automate everything at once. They start narrow – i.e., getting routing right, so callers reach the right person with their context already in hand. They add light information next – an invoice due date, a delivery status – so customers stop being handed to a human for things that do not need one. Then personalised context – ‘I see you recently updated your service plan, is this what you’re calling about?’ – so the interaction stops feeling like a phone tree and starts feeling like a service. The same logic applies to what happens after the handoff. Where the system passes real context across – who is calling, what they have tried, what they are entitled to – the agent stops opening every call with five minutes of diagnostics. By the time the more complex automation arrives, frontline teams welcome it because they can see it working, and customers welcome it because they have already felt the difference.
“AI ambition is no longer the differentiator,” says Emelie Szybanow, SVP Sales Nordics, boost.ai. “The work done around the technology is. Without that, more investment produces more friction, not more return.”
For executives evaluating the next phase of AI investment, three questions are more revealing than any platform demo:
-
Where in the customer journey does ownership pass between teams – and who is accountable for what happens in the gap?
-
When customers reach your business, what do they have to repeat? And what is that costing in trust and retention?
-
What is the organisation actually learning from every interaction, and where does that learning sit?
A practical audit
Justin Robbins suggests a more tactical version of this for leaders who want to test their organisation directly. Pull ten recent voice interactions, he says, and review them with a cross-functional team. For each one, ask where the customer had to work too hard, where the employee worked around the system, and where the metrics said things were fine when the experience clearly wasn’t. Walk out with an owner – and a fix – for at least one friction point. If you can’t, the system isn’t ready for pressure. It is just surviving it.
If the answers feel uncertain, the next investment is unlikely to solve what the last one didn’t. Get them right, and the compounding that AI was supposed to produce starts to happen: every interaction becomes input rather than overhead, frontline teams gain capacity rather than friction, customer trust accumulates rather than leaking quietly through poor handoffs and forced repetition.If the answers feel uncertain, the next investment is unlikely to solve what the last one didn’t.
The technology is ready. The differentiator is whether the organisation around it is.
Hear how voice AI sounds when the system is designed to support it: try the boost.ai voice experience.
