The tech startups doing something different with customer service AI

(and scaleups!)

Mar 14, 2025

It’s safe to say that Generative AI is now well-established in customer service. There are numerous examples of successful implementations of AI assistant tools which help Customer Service Representatives (CSRs) by taking notes, looking up information and providing process guidance. In addition, there are a small, but growing number of Generative AI powered voice- and chatbots that are serving customers directly.

But those use-cases alone still don’t get to the heart of what makes customer service so complex for many organisations. I’m talking about things like:

Customer needs can be ambiguous, with many exceptions that don’t fit the standard “happy path” process
Customers can act in ways which may seem irrational, and are difficult for probability-based AI models to predict
Multiple back-end IT systems which are complex to operate and not well integrated, meaning the process to resolve a customer’s need is more complex than it should be

This week I want to spotlight three companies that are building solutions to these more knotty problems. This is not an endorsement of any of these products, but will hopefully provide inspiration for the next wave of innovation in customer service.

SentioCX is a smart orchestration tool that gets a customer from a bot to a human at the right moment

Sentio doesn’t replace chatbots, voicebots, or CCaaS (Contact Centre as a Service) platforms, but it provides enhanced orchestration between those tools.

Put simply, it enables companies to dynamically match a customer with a human CSR, whilst they are interacting with a bot, based on many factors including: what the customer needs help with, sentiment analysis of the interaction, customer value, and recent customer activity. This can be put to use for practical real-life applications like:

Connect a customer with a human CSR if it there is a high risk that they will cancel their service
Provide an enhanced service to a customer showing signs of vulnerability - important in financial services and other regulated industries
Avoid the need for customers to say “I want to speak to an agent” by proactively offering this only at the point it seems necessary

Koios is an AI model for understanding a customer’s personality and mood

Some companies are starting to use sentiment analysis to offer more personalised service to customers. In other words, to create bots that adapt their response based on the customer’s mood. An early challenge they find is that most sentiment analysis algorithms focus only on the words used by customers, and it can be hard to extract an accurate sentiment using words alone. For example, what is the sentiment of this statement:

I was surprised by the quality of the service

The team at Koios have built a model, which they say, can extract deeper insights about a customer’s underlying personality and feelings, based on a combination of what a customer says and how they say it - their tone of voice, loudness, speed etc.

Customers can be profiled against what psychologists call the Big Five personality traits:

Openness
Conscientiousness
Extroversion
Agreeableness
Neuroticism

These profiles can be used to personalise a response to a customer. For example, a customer with a high level of conscientiousness may expect a solution to be explained in more detail than to an average customer. Of course, companies would need to be extremely careful about how this data is used. There are clear pitfalls to labelling your customers as “highly neurotic”.

Kore.ai have built an agentic bot builder platform designed for complex customer service processes

Kore.ai is a well established conversational AI company, providing voice- and chatbots and AI assistant to financial services, healthcare and other industries.

Kore has recently launched a platform called AI for Process. It’s a low-code/no code bot builder tool that connects to chatbots and voicebots (including Kore’s own) to create bots that are able to form complex tasks off the back of customer conversations.

For example, a bank could set up a fraud management process where a customer can call in to describe where they think a card has been used without authentication. The back-end bot can block the card, order a new one, and investigate through the history of the customer’s transactions to identify any other potentially fraudulent activity.

We can expect to see a lot more of these low-code/no code bot builder tools over the next few years. In fact, Google has recently announced something similar for the Customer Engagement Suite.

The power of a new generation

I’m excited about new technologies like those listed above, as they start to scratch at the true complexity of running a customer service operation in a way that chatbots and voicebots on their own are not able to cover.

Please donate here to directly support the vital work done to support homeless young people by the Centrepoint Helpline

Any donations made up to 31st March will be matched
In addition, I will personally match all donations made up to 31st March, up to a total value of £500 (plus Gift Aid)
For any donations over £50, I will promote your organisation in a future newsletter (read by over 3000 industry professionals each month)

This means, if you are a UK taxpayer and select Gift Aid, then every £50 you donate will be worth an incredible £170 to the Centrepoint Helpline.

Please send a screenshot of your donation to servicematters@substack.com for me to match it and give you a shout out.

Latest perspectives from BCG

How many jobs will AI eliminate? Nobody really knows, and here’s why

The age of artificial intelligence has been full of predictions of mass technology-driven unemployment.
A 2013 report by the Oxford Future of Humanity Institute posited that nearly half of U.S. employment at the time was “potentially automatable” over the next “decade or two.” A decade later, however, there were 17 million more jobs in the U.S.
But will this also be the case in the new age of AI? Nobody knows for certain. There are still too many unknowns to take forecasts of employment doom too seriously.