Tim Hentschel, co-founder and co-CEO of HotelPlanner, speaks to Globetrender about the company's investment in AI 'staff', who can mimic Donald Trump and Santa Claus.

What inspired HotelPlanner.com to invest in building an AI-powered team of travel call agents?

"As a leading travel technology company and hotel booking platform combining proprietary artificial intelligence and a 24/7 global gig-based reservations and customer service network, the latest developments in the AI space have long been central to our expansion plans.

"In the travel-tech industry, it’s almost impossible to grow both operationally and from a customer experience perspective without embracing the ever-expanding AI boom. We’re an engineering-led company, always looking at how we can innovate space, take something that’s high touch and low automation, and turn it into high automation.

"We launched our gig economy call centre during Covid, allowing employees to process calls from their own location around the world instead of being based in a call centre. This not only gave our gig economy agents the desired flexibility in the hybrid world that we now live in, but also the opportunity for customers to be connected to agents with firsthand knowledge of the areas they are selling.

"It’s been incredibly successful with over 7,000 active salespeople, but the number of available agents limited our growth. We wanted to find a way to remove the limitations that come with staff time and capacity, and with AI there are no such barriers. The same AI 'character' could be processing 1,000 calls at the same time, for example.

"The AI booking agents can be seen as a significant evolution of the online chat bot, not only answering queries but offering tailored booking suggestions to suit independent needs."

Could you tell us more about the journey that led to the launch of HotelPlanner.ai?

"HotelPlanner.ai has been in planning for several years now as we looked to evolve at the same speed as technology and customer demands. Our mission was to find something that could enhance the customer experience from a speed and convenience perspective.

"We used data from over 8 million calls to train AI Voice assistants to not only look after our customers many queries but sell rooms in a streamlined, enjoyable and efficient experience that has never been seen before."

How did you go about gathering, processing, and analyzing such a vast amount of data for this initiative?

"We looked back over the highest rated calls from the past four years and programmed our AI agents based on those. This platform wouldn’t be possible without working with the best partners in the business – for example the best partners for large language modules, voice recognition and more. Getting these companies to work together seamlessly enabled us to put the data from our gig agent calls to the best possible use.

"It was important that we kept our data as clean as possible, in order to ensure that the AI technology could continue to develop, running like a well-oiled machine. We use customers behaviours on the site pre-call, and then during the call we use the audio and transcription for analysis. There are cleanup phases to each of those to make sure it is accurate."

Can you explain how the AI agents handle two-way conversations and adjust to the nuances of different languages and dialects?

"Linguistic AI systems use a combination of technologies to understand the intention of the user, maintain stable context throughout, and ultimately generate a human-like response. Through the analysis of those 8 million gig economy call transcripts and audio recordings, these systems have learnt how to handle various queries, accents and dialects. They’re adapted to deal with ambiguity, understand natural language, and provide highly personalised experiences in response.

"We’re now seeing over 45,000 calls taken per day via the agents, and with each call the platform learns more and more about human interaction and speech within this booking context. The ability for the AI assistants to engage in emotionally intelligent conversations is revolutionary. Our booking agents speak fluently in 15 languages [including English, Hindi, Mandarin, French, German Spanish, Portuguese, Greek and Turkish], all of which have been programmed to offer humanised conversations with customers using cutting-edge algorithms, and delivering tailored recommendations for every type of traveller."

"Customers have no idea they aren’t talking to a human."

How are the AI agents able to personalise interactions with callers, and what has the feedback been from customers so far?

"The personalisation of the booking process is something often registered as missing from our competitor platforms – with HotelPlanner.ai, the much-loved travel agent of the late twentieth century meets cutting-edge tech to provide a world’s-first, industry-leading online service.

"Users are able to select an agent of their native language, and even choose if they’d prefer to speak with a male or female agent. The feedback from consumers has been superb. We do offer the customer the chance to switch to a human operator at any point, but we don’t have record of that being the case in many instances.

"We actually don’t even introduce the agents as AI on the call, and recorded transcripts point towards come customers having no idea they aren’t talking to a human. We’ve had users ask if the agent themselves will be at the hotel for check in, or whether they can give a five-star review, to which the response from the AI agent has been that they aren’t in human form! It’s so fascinating.

"The agents are even trained to recognise the tone of the customer. For example, if they are seemingly in a hurry or less interested in small talk, there will be less conversational dialogue and quicker process to the end booking. Whereas, if the customer appears to be more laid back and inquisitive, the agent will offer more detail about the surrounding area of the hotel and things to see and do.

"There is also the capability to remember a previous caller, meaning the agent can begin the conversation in a personalised way by recalling the previous customer booking and offering them a repeat reservation, having already stored their payment details."

In what ways do AI agents improve the booking experience compared to human agents?

"The AI agents have instant access to the most up-to-date information, which could take a human agent much longer to process and review. In seconds, the AI agents have the capability to complete a tailored search, based on the caller’s requests and queries, and present their best suited options in a user-friendly, highly personalised way. This time efficient, and inexhaustible search capability means users of the AI agent will get immediate access to their most individualised travel needs. It’s all about saving time and providing the best results.

"For this AI platform, or any AI platform for that matter, to work – it needs to be better, faster and more efficient than using your thumbs on an iPhone or picking up the phone to a human. Otherwise, customers just simply wouldn’t buy into it."

HotelPlanner.ai fields over 10,000 calls daily. How does this compare to pre-AI call volumes, and how are you managing this scale operationally?

"We’re now seeing over 45,000 calls come in per day and our aim is to double that within the next two weeks. Our pre-AI call volumes sat at around 25,000, so we’re seeing the opportunities afforded by these AI developments."

Are there specific types of bookings or requests where the AI agents perform better than human agents?

"The AI agents are constantly learning and bettering themselves, but at present they are best suited to take the simpler booking requests, or less highly rated calls. This is why we have our gig-agents, and why I often refute claims that AI is bad for the planet from a job perspective. Our agents take the higher value calls, allowing them to make more money per conversation, with the AI agents able to place the simple single room or family room reservations at a greater volume."

With AI now handling a significant portion of calls, how has the role of human gig-economy agents evolved within HotelPlanner.com?

"The AI agents are currently competing against our regular agents, converting bookings at around a fourth of what our human agents are – but the goal is to have the AI agents converting bookings at a similar average rate to our gig-agents. It’s not quite ready to be as efficient as our gig-agents but we are certainly optimistic that it soon will be.

"With AI now able to handle the majority of the legwork and less advanced requests, it means that our human agents are much less stretched, and able to pick up a greater volume of more advanced booking queries and requests. This means that with the combination of the AI agents and human gig-economy, our booking platform is running more efficiently than ever, serving the highest number of customers we’ve ever seen, in the shortest about of time. It’s revolutionised the way we work at HotelPlanner, and means that our employees are able to use their time more effectively than ever."

Do you see a fully automated future for your call centres, or will there always be a need for human agents to manage specific tasks?

"I think there will always be a need for human intervention behind these complex technological systems. I don’t see an immediate future where we would rely completely on AI agents, but with the rate that AI is developing, you can never say never."

How do you think AI-driven booking assistants will shape the travel industry in the coming years?

"I think this technology is something we will see spreading across the majority of major platforms over the coming years. Seeing the difference it has made to our way of working, and the effectiveness of our employees, it’s hard to imagine other major booking platform not following in our stride. The increasing integration of AI technology into everyday life, from Google AI to chatbots and Siri, has proven how processes can be streamlined and simplified with the technology.

"It’s also proven that members of the public are more than happy to converse with AI. AI-driven recommendation engines are refining the art of personalised travel. Platforms such as Google Travel and TripAdvisor employ machine learning to curate itineraries aligned with individual preferences."

With AI handling sensitive personal and payment information, how does HotelPlanner ensure privacy and data security for customers?

"We don't record or log anything during the sensitive parts of the call. It is more secure because you won't ever have an AI agent maliciously asking for your information and keeping it for personal use."

What protocols are in place to prevent issues or misunderstandings when AI agents interact with customers?

"We are integrated with a product to provide observability into AI which means we can track metrics on responses and detect things like hallucinations and overall satisfaction."

Do you have plans to expand the AI assistant capabilities further, possibly into other languages or services?

"Absolutely. Plans are in place to see the platform evolve beyond the current 15 languages. Eventually, I’d love to see HotelPlanner serving users in every country in the world. As far as other services go, watch this space."

What are your next steps for integrating AI more deeply into HotelPlanner’s operations, and where do you see this technology five years from now?

"Customer service and personalisation are the next two big areas for our AI project."

What advice would you give to other CEOs and founders looking to integrate AI into customer service?

"Don’t cut corners with the service providers needed to make a platform like this function to its optimal capacity. It would be easier to find a single partner who provides the data analysis for each different aspect of the AI programme, but you get out what you put in – and if that means investing in the best in the business across every element of the programme, then so be it."