Hard Service Problems, Better AI Economics
"And then the harder they come / The harder they fall, one and all" —Jimmy Cliff ("The Harder They Come")
It's basically an article of faith in the conversational artificial intelligence (now agentic AI, I guess) world: You start by automating the easy stuff. You identify the high-volume, low-complexity tasks that are gunking up the arteries of your customer service organization. Then you set AI loose to handle those password reset, "Where Is My Order," scheduling, and product availability inquiries. Seems like every webinar, conference presentation, and tech vendor pitch offers that same advice for how to kick off your efforts in AI automation.
Inherent in that approach is the idea that companies still need to convince executives that modern agentic AI works and provides value. You could see this approach as a modern gloss on the idea of getting some quick wins to grease the wheels for further investment. In addition, companies already know how to measure the easy stuff. They've already worked out most of the kinks in creating a baseline and demonstrating containment/deflection (and, yes, I chose those customer-unfriendly words deliberately). Finally, starting with the simple use cases makes companies feel like they can limit the blast radius of their experimentations. When the automation of a password reset or order-status query goes pear-shaped, the service recovery is usually straightforward, after all.
That all sounds sensible, I guess. But AI has a striking ROI problem. For example, in PwC's "2026 Global CEO Survey," 56 percent of the executives surveyed said their companies have seen no significant financial benefit to date from their AI investments. So, neither cost removal nor revenue increases. That's a pretty dismal showing.
And, I'll argue, that the lack of imagination in where conversational and agentic AI gets deployed is one of the big obstacles in the way of that magical ROI and its attendant business transformation. At this point, I see start with the easy stuff as less a strategy than conversational AI's version of "Have you tried turning it off and back on again?"
To see why ROI might be easier to achieve with AI applied to harder problems, we don't need to look much farther than the labor unit economics. Automating a password reset generally replaces the least expensive labor in the company: the highly scripted work of an offshore level-one contact center agent. There are savings to be had here, especially when you scale up, but each small victory isn't going to add up to huge changes on its own. On the flip side, tougher issues usually tie up more experienced staff, like level-three agents, claims specialists, underwriters, and technicians, all of whom come with higher costs.
And AI doesn't necessarily have to automate the whole task to produce a meaningful return. If it can shave 15 minutes off a complex, 30-minute process or handle the research, documentation, and system-hopping while a human makes the consequential decisions, the economics usually still work. Partial automation of expensive work might be worth considerably more than complete automation of cheap work.
Additionally, when a customer experience team starts with the easy things, they prime executives to expect AI projects to be relatively simple and fast (and inexpensive, ideally). That, however, leads to a lot of questions when that team wants to expand to more involved use cases, least of which are why this will cost so much and take so long when the first project was a breeze. Start with the complex use cases and you've anchored expectations at needing more effort but for much greater reward.
Maybe most important, harder problems are also where companies can differentiate on customer experience. Nobody becomes loyal to a particular brand because a password reset worked. Resolve a denied claim or billing mess, though, and customers might actually remember. So, maybe it's time to quit using AI to nibble around the edges of your business and start looking for the big, hairy, high-value problems worth taking a proper bite out of.
Ian Jacobs is vice president and lead analyst at Opus Research.