2025: Looking back at the year AI became a teenager
If 2024 was the year AI toddled into everyone’s workflows and daily life, 2025 has been the year of realizing the need to do something more with it. Not just for search or summarization but harness some of the power for outcomes in business settings.
And it’s this shift from curiosity to mandate, which has defined everything in the world of customer experience and contact centers this year.
AI has become accessible in a way we couldn’t have predicted
One of the most surprising and exciting things about 2025 has been the accessibility of AI. This is the first time in decades that transformational technology is available without needing a full-stack transformation to be useful.
If you’d asked me five years ago whether companies could simply plug a bot right into their CRM despite still working within 2010’s call routing infrastructure, and I would have laughed at you. But this year? That happened.
The good news: You can adopt AI-driven automation without having the perfect infrastructure.
The bad news: There’s a lot of unintended consequences happening that impact everything from cost to revenue to customer loyalty.
Most bots will work better with stronger infrastructure and a thorough training, testing and implementation process instead of “plug and pray”.
I recently met with an executive who was still using call routing methods which I hadn’t seen in fifteen years. Still, they implemented a booking bot. Bravo! That’s bold, exciting and a step in the right direction. Their bot could schedule appointments perfectly when it was a straightforward type of appointment and customer data was clean in the system.
Here’s the thing – while it did everything it was supposed to do, it was also the year of unintended consequences. When there was lack of data hygiene, technicians went to the wrong address and sometimes weren’t scheduled for the right appointment or the right amount of work time. And the icing on the cake was that upsells tanked – because the organization didn’t realize how often agents were educating customers about additional services or discounts and then taking their orders on those scheduling calls.
This leads me to the paradox of 2025: The deployment of AI is more accessible than ever before, leading organizations to mistake accessibility for operational usability. There is always a hidden level of complexity to each AI tool that meant many organizations started the deployment process with technical feasibility but made the operational teams struggle to implement even passable automations.
The biggest lessons of 2025
Across dozens of conversations with CX and operations leaders, here are the three major themes that I’ve observed.
Yes, you can implement AI quickly. Skip foundational work at your peril.
This excitement about low-friction AI created a misconception that went something like: “If we can deploy fast, then we should deploy fast.”
The truth is: without good data, understandable processes, and sound testing practices, your advanced AI system will just execute poor decisions quicker.
Organizations just can’t afford to skip laying the groundwork for implementing AI, which is what makes AI implementation successful. The companies that took time to understand call drivers, edge scenarios, escalation routes, exception processes, and the like were the ones that were able to scale. The companies that skipped this step are currently paying for this in terms of customer dissatisfaction and process rework.
The players who won in 2025 weren't always AI first movers; they were the ones that who took the time to understand their business foundation first and then moved quickly while testing thoroughly with controlled rollouts.
2. AI can do far more than we can even imagine. Pilot does not mean thinking small – it’s a stepping stone when you’re thinking big.
AI can already perform certain tasks more than 100 times faster than humans. But most of the applications that we viewed this year included bolt-on automations (“Can the bot do this task?”) or attempts to recreate existing human workflows through bots.
Meanwhile, the firms seeing strongest results were the ones asking bigger questions:
Where is human touch important to my brand or process?
Will customers trust recommendations from a bot or should that only be a human process?
How do people and bots collaborate for the best outcome for customers?
How do I reinvent this entire process, not just automate it as is? Can I do it in fewer steps or totally differently?
Can I let consumers choose between bots and people because they can both perform the same sets of tasks?
Are there things where a bot will just do the task better in terms of accuracy and so that function should only be performed by the bot?
Can I find some settings to practice with the MVP (minimum viable product) and develop a risk tolerance for a new operating model?
These companies accepted a change in basic assumptions that allowed them to open their minds to even more possibilities with AI.
2025 was the year of discovery that AI can definitively handle more complex tasks, provided that the way the work is done changes.
Doing more with less has changed the talent landscape
This year has uncovered a growing divide between those who can operate in an AI-enabled world, and those who can’t.
Experienced professionals felt that AI worked as a force multiplier
For entry-level employees, it was considered an obstacle to entry
I saw countless instances: from junior marketers unable to find entry-level roles to frontline supervisors struggling with new quality programs being automated behind the scenes.
This created a ripple effect that changed teams in subtle ways. Job descriptions changed in a quiet manner. The bar was raised. Fewer people needed to be involved in complex and high-leverage activities. Organizations started realizing that training, upskilling, and role clarity would become important factors beyond headcounts.
This is not an anomaly or a temporary blip; this is just the beginning of a generational skills shift.
The decline in satisfaction surveys
If you’ve heard me speak recently, you know how strongly I feel about this: customer satisfaction surveys as we know them are dying. And that’s a good thing.
In 2025, response rates dropped below 1% across industries.
At the same time, AI-powered transcription and interaction analytics can analyze 100% of calls and chats in seconds. That means:
You know what the customer actually said
You know how they felt based on tone, sentiment, and topics
You know exactly what the agents or bots did during the conversation
Why rely on self-reported data from the 1% of survey respondents when you can rely on objective data from 100% of your customer service interactions?
This is another paradigm shift happening in real time: rather than asking our customers to tell us how we did, we are now capable of observing it firsthand. This makes it possible to have a more honest, measurable, and scaled feedback loop system than could ever have been possible with surveys alone.
Surveys aren’t dead just yet, but 2025 was the year we unplugged them from life support.
A very big year – but still only the beginning
2025 felt monumental not only because of what was happening but because of the speed with which it was happening. Faster than the Industrial Revolution or the dot-com boom or even TikTok trend shifts. Changes in the workplace are happening globally in days, weeks and months, not years or decades.
The organizations that thrived in 2025 went beyond a mandate to use AI – it was the ones that were ready to challenge assumptions, test and pilot, and lay the groundwork to scale that wound up accelerating. I don’t think 2025 was the first major year in AI, but it was the year we realized there’s no going back.