The days of AI hype are over. Today, patience is thin, and AI has to deliver real results. Every discussion about AI in Customer Success in 2025 centers on what tools people are using, how they’re using them, and the value they’re seeing. 

Nowhere was this more evident than at Planhat Open 2025, where we gathered together CS leaders from around the world to discuss the most pressing matters in Customer Success. AI was prominent in many discussions, including a dedicated interactive workshop session. 

In the session, we discussed the current state of AI, practical use cases they are seeing value from, as well as advice about how AI is shaping CS careers. This article is the distillation of their knowledge. 

GenAI in the trough of disillusionment

The session started with a discussion about the recent Economist article on AI moving along Gartner’s Hype Cycle, which charts the progress of any new technology.

The general principle is that when a new technology emerges, there are expectations that exceed its ability to deliver, a lull, then an eventual stabilization. 

According to the Economist, Gartner has moved Gen AI to the transition point where it enters the ‘trough of disillusionment’, which means it will be 2-5 years before we reach the ‘plateau of productivity’, where AI delivers consistent value in line with expectations. 

This falls in line with the sentiment that we are seeing in the CS space, where patience has grown thin with tools and features that fail to live up to overinflated expectations. For CS leaders at our event, this was resoundingly the norm rather than the exception – with the majority reporting use of AI, but only a small minority seeing clear value.

AI use cases in Customer Success

If you find yourself using AI but yet to see value that moves the needle, or are yet to start using AI – here is a list of all the AI use cases CS leaders shared that they are using today: 

• Call recording/note taking:

  • AI tools: Fireflies.ai otter.ai

  • Approach: Many leaders reported CSMs save 2-3 hours a week by automating notes using note takers, providing summaries to Slack, sending email summaries to customers, and follow-ups for the team. Some are also using note takers to analyze Voice of the Customer, sentiment analysis, trends across customers on topics, and collate product feedback.

  
  

• Churn and opportunity analysis:

  • AI tools: Gong & Planhat

  • Approach: One of the highest impact use cases shared was integrating Gong with Planhat to auto-create opportunities and risks by identifying churn indicators and expansion potential flags.
    
    

• Content generation:

  • AI tools: Claude, ChatGPT, Google Gemini, Gamma

  • Approach: There were a number of different applications for the creation of content. Some participants used AI for written content like emails, one-pagers, presentations, and business plans. Others said they have had success creating pitches/demos, automated account plans, and even rewriting contract clauses before sending to legal, expediting the process.
    
    

• Video creation:

  • AI tools: Synthesia, HeyGen, ElevenLabs, Veed.io

  • Approach: For video content, people made AI avatar videos with Synthesia and HeyGen, used ElevenLabs for AI voiceovers, and Veed.io for subtitles.

• Strategy:

  • AI tools: ChatGPT

  • Approach: One participant shared an example of a Custom GPT "strategy bot" that allowed teams to input ideas and receive feedback on how they align with long-term strategy and roadmap.

The impact of AI on CS careers

We’re all familiar with the phrase: “You won’t lose your job to AI, but you will lose it to someone that uses AI.” But now we’re starting to see what that looks like in reality, and it contains more nuance.

While CS leaders aren’t letting people go for not using AI, it’s becoming a requirement for new hires. And because of the efficiency gains from AI, there will be fewer hires in general.

This means that those not willing or able to use AI will find their career options limited, and will get left behind at the organization where they currently work. As one participant put it: “If one CSM can manage 300 customers with AI agents, and another can only manage 30, which one is going to get the promotion?”

An important caveat is that because we’re still in the early stages of AI, there is a certain amount of education needed for people to adopt this new technology. And while many early adopters are happy to take the burden of education upon themselves, it is unreasonable to expect that of everyone.

Leaders who reported the greatest adoption and success with AI have led education sessions. For some, these took the form of expert-led workshops introducing staff to the different tools and the basic concepts like prompting and hallucinations.

For others they ran ‘science fairs’ where staff used AI for something they were passionate about – not necessarily work related – then brought their creation in to share with the team.

The future of AI

There’s an obvious parallel between these AI trainings and the training sessions from the 1980s on how to send an email. While that may sound preposterous now, this instructional video from 1984 is a perfect example of why training is needed for emerging technology. 

If AI survives the ‘trough of disillusionment’ and becomes as easy to use and productized as email, the current teething pains will seem just as preposterous. 

One final thought – from Planhat’s former CCO, the late Chris Regester – as we enter the trough is this: What is an appropriate expectation of AI? Many people expect it to be perfect, and consider any hallucination an unforgivable fault, but would not hold a human to the same standards. 

Humans get things wrong, forget things, and yes, occasionally make things up to make themselves sound good. 

We are currently expecting AI to perform tasks at a speed and scale far greater than any human can manage, while also being more accurate than any human would be. 

And we risk limiting ourselves because of it. 

So where does this leave us? 

It leaves us with a question, not just about AI and its capabilities, but also our thoughts and biases towards it. Next time you use AI to complete a task, ask yourself not just whether the output is accurate, but also if it’s better than a human would do – perhaps even better than you would do (at least within reasonable time constraints).

The real litmus test, we discovered is: ask yourself what you would be doing if you weren’t using AI. If the answer is ‘nothing’, you shouldn’t be comparing the output to an imaginary ideal of what AI should be capable of, you should be comparing it to doing nothing. 

Want more advice on how to use AI in CS? Check out our on-demand webinar Mastering AI in Customer Success.