I’m seeing more and more comments suggesting that the day of the CSM is rapidly ending, that the profession must transform or become obsolete and be replaced by artificial intelligence. I asked ChatGPT for a list of typical CSM activities for a B2B software firm and for it to indicate which of those activities could be done partially or in full by an ai. (Note that the question specified “could be” and not “should be.”) ChatGPT identified 5 activities where an ai could replace humans, two where it would augment human efforts, and only one where a human was required.
Which is Better for What: AI or Human?
First on the replacement list was Customer Health Scoring, with ChatGPT asserting that an ai could do this better than a human because it could analyze more data. Okay, but that assumes that you have the data in one of the 20 or so CS platforms (Gainsight, ClientSuccess, ChurnZero, etc.) What if you don’t? Even if you do have a CSP, Ed Powers of Service Excellence Partners challenges the notion that monitoring health scores could more effectively be relegated to an ai. “Having more data isn’t necessarily more accurate. Having the right data is, and only humans can determine if the data are right or wrong.” Ed continued by noting that “AI has some advantages over machine learning and regression models, but it also has disadvantages. AI models are opaque, for example, so while it makes a prediction, any information about which variables are being used in the model is hidden. That includes trending, by the way. And the juice isn’t always worth the squeeze—a few more points in predictive accuracy doesn’t make a practical difference unless you’re dealing with extremely large data sets, like Netflix or Amazon.”
Second on the list was Customer Onboarding Automation, including sending user emails, suggesting help articles, tracking progress on checklists, and escalating if things look like they’re stuck. Again, that assumes that you have one of the 20+ onboarding systems implemented (Rocketlane, GuideCX, OnRamp, etc.) If not? Keep in mind that a human had to design and define that onboarding process for the ai to manage.
Email drafting, transcribing meeting notes, and forecasting all were on the replacement list as well — but note the “drafting” on emails. Trusting an ai to write and send an email on its own is a risk I’d not want to take, having seen some ai generated output that included painful errors. I use an ai to transcribe meetings all the time, but I always review that transcript. Voice recognition isn’t infallible, and there are nuances that are only recognizable when you personally know the other party well.
Activities on the augmentation list, where ai can assist but not replace CSMs included preparation for Customer Value Reviews (the proper name of a “QBR”) and for drafting Success Plans. Again, the prerequisite is having a CSP to provide quality data, and we’re talking only about prep for the session, not the delivery.
Finally, we came to a category where ChatGPT admitted that humans had the advantage, where emotional intelligence, deep product knowledge, and strategic alignment were required. Strategic roadmap discussions, building trust-based relationships with champions, escalation management and complex negotiations, and securing customer advocacy and case studies are not something an ai can do. Two points come to mind. First, this sort of activity is exactly where the joy is for me. How about you? Caterina Canaletti insists that “AI is a powerful assistant, but the magic of CS lives in trust, nuance, relationship building and the ability to read a room. That’s not something you can automate.” Second, to build the kind of trust-based advisory relationship where these activities are possible takes time and consistent human interaction. The domain expertise that is the foundation for speaking with authority doesn’t come overnight or from an algorithm.
AI and Customer Data Quality
Lurking under all of the assertions and concerns is a major assumption. While invisible, it’s the elephant in the room: data quality. In the early days of the technology industry, there was an apt acronym for the issue. GIGO = Garbage In, Garbage Out. In all the years since computers arrived on the scene, this challenge remains serious. Now, with ai, more questions arise. What data? From what source? With what built-in biases? Slanted how? How was it selected and what was left out of the analysis? Why? And how will you get answers to these questions?
As an industry, we can’t afford to just assume that the output of an ai is automatically better than what humans could produce or that the ai is necessarily correct. If the data is in some way flawed, so will the conclusions the ai delivers. And so will the decisions that we make using that report be flawed.
Looking Ahead
In the discussion of this topic in The Customer Success Forum on LinkedIn, Jason Noble summed it up neatly. “We need to stop talking about AI like it’s plug-and-play magic. Without the right data, systems and structure in place, you’re not replacing anything, you’re just adding noise. Where AI can help is saving us time on the repetitive stuff, so we can lean into what really matters – building trust, having strategic conversations, driving outcomes.That’s the work I love too – and it’s not something an algorithm can replicate.
Resources
If you’re looking for Customer Success technology tools and ai, see The Customer Success Directory pages on Customer Success ai Vendors and the Customer Success ai TechMap.
.For discussion of this article. please join us on LinkedIn in The Customer Success Forum
