Customer Success, AI or die?

Over the last several years, companies have done what the playbook told them to do. They hired more CSMs, built out onboarding and renewals teams, and invested in the tooling to support them. Customer success headcount has grown.

Net revenue retention has not followed. According to Bain's recent report on the state of the function, NRR is declining even as investment in customer success roles continues to rise. More people, more process, and less to show for it.

Bain frames customer success as being at a crossroads: evolve with AI, or fade away. The headline is dramatic, but the underlying point is a fair one. The traditional model of scaling post-sales by adding headcount has hit its limit, and the teams that recognize that early will pull away from the ones that don't.

Why more headcount isn't fixing retention

The uncomfortable finding in Bain's survey is where CSM time actually goes. Roughly 65% of a CSM's time, close to two-thirds, is spent on low-value tasks that AI could automate. That means pulling data from multiple systems, preparing for account reviews, writing updates, logging notes, and chasing information that already exists somewhere else in the stack.

This is the core scaling problem. When the majority of a CSM's week is spent assembling context rather than acting on it, adding another CSM adds another person doing the same assembly work. Capacity grows, but impact per account does not. The accounts that need real attention still wait behind the administrative load, and risk still gets spotted late.

It is the same pattern we described in our piece on why Customer Success can no longer operate as a silo. When customer data is spread across the CRM, the ticketing system, and the communication tools, humans become the connective tissue. That is expensive, slow, and impossible to scale linearly.

Everyone sees the potential. Few have moved past the pilot.

If the problem is that obvious, the answer should be adoption of AI at scale. In practice, that is not what is happening.

Bain's survey of customer success practitioners found that while most leaders see the potential of AI, nearly 70% have not moved beyond scattered pilots or low-level use cases. A meeting summarizer here, a draft-email assistant there. Useful, but nowhere near enough to change the economics of the function.

The report points to a handful of reasons why teams stall:

• Chasing micro-efficiencies. AI gets layered onto existing, human-centric processes, which saves a few minutes per task without changing how the work gets done.

• Unclear priorities. Without a sharp understanding of what customers actually need, it is hard to know where AI should be applied first.

• A fragmented tool landscape. There are few AI tools built specifically for customer success, so teams end up juggling several vendors, each with narrow context on the customer.

• AI treated as an IT project. When AI sits with IT rather than with the business, it rarely becomes a strategic priority.

• Limited executive sponsorship. Customer success often lacks the size and visibility of sales or engineering, which makes it harder to win CEO-level backing for a transformation.

Several of these come back to the same root cause. AI is only as useful as the context it can see. A tool that only knows what happened in the CRM, or only what happened in the inbox, produces shallow insight no matter how good the model is. Narrow context in, narrow value out.

Turning the tide: what Bain says it takes

The report is clear that incremental effort will not close the gap. It points to four moves that separate the teams making real progress.

1. Set a bold ambition with measurable targets. Decide what the AI investment is actually for. Serving more customers with the same team, spending more quality time with strategic accounts, or delivering a better experience with fewer people are all legitimate goals, but they lead to different decisions. Vague ambition produces scattered pilots.

2. Focus on two or three processes. Spreading AI thinly across every workflow yields small gains everywhere and a transformation nowhere. Concentrating on a small set of related processes, such as renewal preparation or risk detection, creates bigger efficiency payoffs and a clearer story for leadership.

3. Rethink the process, not just the steps. Applying AI to each step of an existing workflow gives modest improvements. The bigger gains come from starting with a blank slate and asking how AI and automation could achieve the business outcome in the first place, even if the resulting process looks nothing like the current one.

4. Embed it deeply. New processes need internal champions, and the old alternatives need to be retired. Bain notes that when the legacy dashboards go away, teams are forced to adopt the new tools to get their work done. Optional AI stays optional, and optional rarely wins.

What this looks like in practice

Taken together, these points describe a specific shift: from AI that helps a CSM do the same job slightly faster, to a system that changes how the job gets done.

Consider the sequence a CSM goes through today when an account looks shaky. They check the CRM for the deal history, open the support queue to see recent tickets, scan Slack and email for tone, review usage data, and then try to piece together what is going on and what to do about it. Each step is manual, and each depends on the CSM knowing where to look.

Now consider the alternative. Every new signal, whether a ticket, a call, a message, or a file, updates a single account view automatically. The risk is assessed with the full evidence behind it. A recommended playbook is attached, and the follow-up work is already assigned to the right owner. The CSM starts from a decision, not a research task.

That is the design principle behind Aitelligent. By connecting the CRM, ticketing, and integrations into one post-sale operating layer, it gives AI the full customer context that narrow point tools lack, and it turns that context into assigned action rather than another dashboard to interpret. It is built around the two or three processes where the gains are largest: detecting risk early and executing the right response without the manual assembly in between.

So, AI or die?

The framing is stark, and reasonable people will argue about whether "die" is the right word. What is harder to argue with is the trend. Investment in customer success is rising, retention is not, and the majority of teams are still experimenting at the edges of AI rather than rebuilding around it.

The teams that come out ahead will not be the ones with the most pilots. They will be the ones that pick a bold target, focus on the processes that matter most, redesign the work instead of decorating it, and commit to the new way of operating.

The crossroads is real. The question for every post-sales leader is which direction their current investment is actually pointing.

Source: Bain & Company, "Customer Success at a Crossroads: Evolve with AI or Fade Away," based on the Bain Customer Success Practitioner Survey, August 2025 (n=235).

---

Aitelligent connects CRMs, tickets, and integrations into one post-sale operating layer, turning customer signals into assigned action. Get in touch to see how it works.