AI: The making of a multitasking chatbot – lessons from Mercury

Pronounced ‘he-core’, the name Hiko is reo Māori for lightning, power, and means ‘to flash

Pronounced ‘he-core’, the name Hiko is reo Māori for lightning, power, and means ‘to flash

Already, the energy company is preparing Hiko for additional roles – answering simple administrative queries from internal staff, such as ‘how many leave days do I have left’?

“There are opportunities to make a chatbot available in different channels, to support people to get [answers to] their queries faster,” says Tim Aynsley, head of ICT at Mercury.

According to him, chatbots help lighten the load in contact centres so employees can focus on valuable conversations with customers.

He adds that Mercury is also looking at its applications around employee experience.

“AI is built into it,” explains Aynsley, on the technology needed to make this happen. “Over time, we expect it to be sophisticated.”  

He says making a chatbot is a journey that starts with building the foundation on a bigger arena for “intelligent automation”.

“We invested in some key platforms, moving stuff to the cloud with a focus on the customer experience,” says Aynsley. “We want to inspire, reward, and make it easy for our customers.”

The next phase, he says, is about analytics and innovation. 

Tim Aynsley

This strategy sets us apart in an industry that’s dominated by incentive programmes aimed at new customers 

“The intent here is not just to apply technology for technology’s sake,” he says. “So, we’re focusing on applying chatbots or robotics for how to work smarter.”

“We started with an MVP (minimum viable product) for simple queries.

“Hiko is designed so it will integrate and undertake digital activities for [both] customers and employees,” he states.

“We looked at how to integrate tools, not just robotics, but also KPIs to digital assistance. That triggers a series of orchestration events, and we have a platform designed with that in mind.”

Leading through intelligent automation 

The experience at Mercury is an example of how organisations are preparing for the rise of AI and emerging technologies such as chatbots into the workplace.

In five years, AI and emerging technologies such as chatbots will replace almost 69 per cent of the manager’s workloads – Gartner

In five years, AI and emerging technologies such as chatbots will replace almost 69 per cent of the manager’s workloads, predicts analyst firm Gartner.

Moreover, a recent Gartner survey of more than 3,100 CIOs indicate chatbots are the top, most-used AI-enabled application across enterprises today. 

The top AI-based applications are linked with improving customer experience, according to the report by Gartner analysts Alan Antin, Suzanne White, and Jen Singleton.

At Mercury, Aynsley says new roles are also evolving across the technology and digital teams through their work with Hiko.

“You see the emergence of roles that continue to support the growth and development of chatbot intelligence,” he says. “That is a work in progress.”

Soft launch

He notes that Hiko was made public through a ‘soft launch’ on Mercury’s social media sites early this year.

“We want to focus on quality, and make sure it will deliver the right experience, and around the value of meaningful customer experience,” he adds.

Mercury has a customer base of nearly 373,000 and provides 17 per cent of New Zealand’s energy generation, according to the company.

The company also says it is heavily focused on customer retention, advocacy and ensuring customers remain with them for life. “This strategy sets us apart in an industry that’s dominated by incentive programmes aimed at new customers.”

According to Ansley, Hiko is part of Mercury’s goal to make it easy for their customers to engage with them on their choice of channels. “We are creating a multichannel experience.”

You see the emergence of roles that continue to support the growth and development of chatbot intelligence… That is a work in progress 

As to what worked well with the deployment of Hiko, Aynsley says it was getting “broad executive support around the longer term value of intelligent automation”.

“You need to ensure you get support around resourcing, not just delivery of the base capability but ongoing investments,” he says.

He credits his predecessor, Roxanne Salton, now chief digital officer at Southern Cross Health Society, with paving the way for the development of Hiko as part of a wider “intelligent automation” programme.

“She got the endorsement of the board and executive leaders for intelligent automation as a key part of our future.”

Object lessons

Various business units were involved in the creation of Hiko, says Aynsley, on another key lesson from the initiative.

“It was very much a cross-functional team, it’s not a technology-centric led project.”

Hiko is a perfect example of how we used operational data and experience data to create a product our customers want, because it gives them greater control to do business with us on their terms

“The ICT, engagement and marketing teams learned new things along the way as we worked together on launching Hiko,” adds Mohammed Abbas, head of customer engagement at Mercury.

“Hiko is a perfect example of how we used operational data and experience data to create a product our customers want, because it gives them greater control to do business with us on their terms,” he explains. “It also ensures our people can focus on adding value through meaningful conversations, as Hiko can manage administrative enquiries such as billing, joining and moving.”

“The prospect of doing this excited and unified the teams around a common goal of ‘making things easy’ for our customers’.”

Mercury NZ

Chatbots: Not just ‘create and deploy’ 

Analyst firm Gartner says various sources – including ICT providers and media – have caused enterprises to believe chatbots are easy to create and deploy.

Some small development shops and framework providers lead organisations to believe they can input a couple of hundred Q&A pairs and have a chatbot up and running. 

Organisations should only attempt to create their own chatbots if they have the right data science and machine learning assets

“While this is technically true, you will end up with a chatbot that users will not be happy with due to very poor performance,” writes Gartner analyst Van Baker. “This will reflect poorly on your company and should be avoided.”

The analyst firm lists best practices for the development of chatbots:

• Avoid solutions that perform poorly or overlap: 

Gartner says many vendors are unable to deliver enterprise-grade chatbot solutions. Thus, it advises: “Screen vendors carefully, avoid providers that cater to single use cases and ensure that chatbot guidelines are defined.”

• Reduce the risk of failure by sourcing chatbots from external providers:

Organisations should only attempt to create their own chatbots if they have the right data science and machine learning assets. Some questions they need to ask include: Does the enterprise have data scientists on staff? Does the business have expertise in deep learning frameworks and machine-learning?

“If not, look to third-party providers that specialise in data preparation or providers that build and host chatbots,” says Gartner.

• Secure funding and monitoring resources for ongoing model maintenance: 

Gartner notes that chatbots need ongoing operational assets that can periodically evaluate the performance of the model and add domain-specific expertise. It calls on organisations to devote resources to model management on an ongoing basis and ensure they have access to all the required data management skills.

• Prepare for voice-enabled chatbots: 

Gartner also points out that although few chatbots currently support voice-enabled features, demand for such features is increasing. “Be prepared to meet this demand by specifying voice support in your solutions.” 

• Incorporate tone, emotion, personality, and other soft features: 

Soft features are critical to the success of chatbots, although most solutions don’t include those, states Gartner. 

Organisations must ensure the chatbots they deploy reflect the values of the company and brand, and have a “pleasant tone”. The latter increases the tolerance for mistakes or additional questions. (Source: Smarter with Gartner)

Credit: Divina Paredes, CIO.com