As your front office increasingly consists of AI agents handling a growing portion of customer queries, how long can you still call it a contact center? Ziptone spoke with Anne Jenkins of Teneo.ai, a company that isn’t shy about using the term “agentless contact center.”

 

Anne Jenkins doesn’t like to talk about “chatbots.” Because when you say chatbot, most people think of a text-based application of GenAI and solutions that don’t work well. That’s largely because of conversation designers, Jenkins believes. “Most of them should be ashamed of themselves, because even when the technology was pretty good, a lot of what ended up in production was pretty bad.”

Anne Jenkins is VP Solution Architecture at Teneo.ai. The (originally Swedish) company has been active in the field of AI for 20 years and was involved in the development of the chatbot for Ikea, for example. “Our largest customer is one of the three largest tech companies in the world and processes an estimated 120 million calls per year in 50 languages, spread across 100 countries.”

'Incorrect answer from our AI support bot'

Examples of bots that are malfunctioning regularly go viral. Think of DPD, Air Canada, and recently Cursor, an AI company whose chatbot gave incorrect information about the terms of use of a software package. Customers then raised the alarm on Reddit and cancelled the product. Cursor had to correct the situation: 'incorrect answer from our AI support bot'.

The chatbot and voicebot market is very crowded, Jenkins admits. “Anyone can deploy an LLM, put a layer on top of it and say there’s a chatbot available. The trick is to manage the interactions well. That could mean making corrections during the interaction, even before the output reaches the customer. Or it could mean continuously monitoring usage so that you can catch prompt hacking – which every chatbot and voicebot is subject to – in a timely manner,” Jenkins says.

But what if those bots do their job well? Enter the idea of ​​an “agentless contact center” — a relatively new term in customer contact land. According to Jenkins, it’s simply a contact center that’s largely run by virtual agents instead of human employees; and that’s true for the vast majority of inbound contacts.

'Always need a human being'

We put it to Jenkins that – at least in the Netherlands – the standard response often comes down to: 'there will always be a need for human support'. In other words: an agentless contact center is impossible. "Let them call me for a 'proof of concept'," Jenkins laughs, "because those people are probably basing their idea on the wrong chatbot technology. That would be my first response. Of course, it takes time to work towards an agentless contact center. It's a journey. But one of our customers started at 40%, is now at 60% and keeps increasing that percentage." Examples of other brands that are deliberately working towards more and more automation include Telefónica Germany, Swisscom, Ikea, CSG and HelloFresh, says Jenkins.

Agentless contact center – The term 'agentless contact center' was likely coined by Indian IT company Wipro in 2020. In a whitepaper on 'Providing quality customer service during challenging times' (referring to the coronavirus pandemic), it states: “…with the ultimate vision of an 'Agentless contact center' service.” Wipro is referring to the model in which a future contact center functions as much as possible without human agents; human employees then work only behind the scenes.

Prejudices

Is it possible that contact center leaders themselves are biased, rather than the technology? “It’s reminiscent of the early days of OCR,” says Jenkins. “That didn’t work well at first either, but now OCR is faster and more accurate than humans. Many people get stuck on the idea that AI is only good for simple queries. But we can now handle complex cases, even if that requires back-end integrations. Voicebots are perfectly capable of handling interruptions, topic changes, and multiple pieces of information in a row. Consumers don’t speak in neat, measured, perfectly worded sentences, even when responding to a voicebot query.”

Stop after automating simple questions

Many companies are stuck with text-based bots because voice bots are seen as riskier, Jenkins said. “But an agentless contact center is not impossible. Not every company will pursue it, and for those that do, it’s a lot of work to get there. But many companies take the first step—automating simple questions—and then stop. Or they don’t take that first step at all. There are still many organizations where you have a button-driven IVR followed by wait time. If you call that a good customer experience, you’re kidding yourself.”

Jenkins sees continuous automated testing as an essential part of deploying AI agents; important because AI operates in environments that are constantly changing. There are also situations where you don’t want to give the solution freedom, but want to keep it to a strict script. In that case, you would rather not use a GenAI application, but a more deterministic flow – if this, then that.

Escalation: critical transition moment

Another key point is handover during escalation. ‘Warm handover’ is a matter of good integration between the AI ​​bot and the contact center platform: not only do you need to have customer data visible, you also need to take into account everything that was shared during the chat or voice bot conversation, says Jenkins. “Think about intent, urgency, mood and sentiment of the customer. Based on that kind of data, you can give the customer specific treatment – ​​for example, placing them in a certain queue. You can also provide targeted information there. In the meantime, you can give agents a pop-up screen with a summary of the interaction. It all depends on how far you want to go in designing that kind of solution.”

“We are seeing a growing trend among large enterprises to embrace agentless contact center solutions. This shift is driven by the desire to improve efficiency, reduce operational costs and increase customer satisfaction. In general, as technology advances so rapidly, enterprises are realizing that full automation is a possibility.”

Speed

In other words: even during the journey towards 'agentless', escalation to an agent remains essential - even if you are working towards 80 or 90% 'agentless'. But in the latter case, the question arises whether 'agentless' is still a fair term. After all, you could say that bots form the first line of contact. This also changes the classic contact center: from large numbers of seats to an escalation team with a handful of specialists.

Jenkins responds: “The speed at which this develops depends of course on the complexity of your processes. Most companies can take the first step towards tens of percent automation within a few months. After that it depends on your further ambitions.”

There are experts who claim that within five years three quarters of the agents will have been replaced through AI solutions. Jenkins: “I don’t think those experts are far off the mark. For contact centers with large groups of agents, this investment pays for itself very quickly, even if you’ve already moved customer contact to low-wage countries. There’s no other software with such a high return. Companies that don’t do anything will eventually be outcompeted by companies that do take these kinds of steps. I think we’re going to have to expect drastic changes within the next five years.”

(Ziptone/Erik Bouwer)

Featured, Human Resources