Voice AI Technology

26
Jan
Talkbot Basics  ·  Voice AI Technology
Creating authentic connections to engage customers in the digital age
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

The rise of e-commerce and social media have given businesses new opportunities as well as challenges when it comes to connecting with their customer base. Brands are now able to reach and engage with customers in innovative ways through channels like Facebook, WhatsApp, and even TikTok – and customers expect them to. In the State of the Connected Customer report by Salesforce, 88% of customers surveyed expect companies to accelerate their digital initiatives. And in that same report, 80% of customers say experiences are as important as a company’s products and services. Customer engagement has become more important than ever.

With technology, it’s become increasingly easy to automate interactions and rely on machines rather than humans. This may lead to colder, more sterile communications. On the flip side, more data enables brands to create more personalized experiences, which can build brand loyalty. Businesses need to find a balance between using technology to enhance customer engagement and maintaining the human element that is essential for building authentic connections.

Using artificial intelligence to bring that personal touch

One of the key areas where technology can bridge technology with personal connections is through the use of Artificial Intelligence (AI). AI technology can be used to create personalized, authentic connections with customers through well-designed interactions. With conversational AI, these interactions – whether text-based or voice calls – enable customers to connect with business in real-time. Thanks to natural language processing and machine learning, AI bots can be the first point of contact for common customer queries, reducing the frustration of long waits. And when fed with localized data, AI-powered voice bots like the WIZ Talkbot can adopt local accents and expressions, creating a more human-like engagement.

AI technology can also be used to improve customer service by analyzing customer feedback and providing insights into customer sentiment. This can help businesses identify and resolve issues more efficiently. And when issues are spotted before they gain mass scale, brands can take a more proactive approach to problem resolution by sending updates even before a customer calls or sends a message. All these help build an excellent and memorable customer experience.

Another way AI can be used to improve customer engagement is through personalization. AI-powered algorithms can analyze customer data, such as browsing history and purchase history, to create personalized recommendations and offers. This not only improves the customer experience but also increases the chances of conversion and repeat business. By tailoring the customer experience to the individual, businesses can create a sense of relevance and value for the customer.

Connected omni-channel interactions to delight customers

Recently, Zendesk has coined a new term – immersive customer experience (CX). The concept is anchored on something many marketers already strive for, which is seamless omni-channel communications. When brands give customers a consistent experience even when they switch channels, that’s part of building an immersive experience. It’s about building a customer support environment that’s accessible, engaging, and connected – something that would make customers want to stay.

Part of creating that immersive experience is making interactions less rigid and more natural. Customers are now looking for more conversational brand communications across the entire journey – from marketing to support. This means customer touchpoints also need to be more integrated so that brands provide one consistent message, regardless of the channel or the topic. A solution like WIZ Engage enables brands to design fluid, connected interactions across voice and text, walking alongside the customer on their journey.

The quality of the experience matters; 73 percent of customers say they will leave for a competitor after multiple poor interactions. Even after just one bad experience, more than 50% of customers will consider another brand. The pressure is on companies to provide the kind of service that customers now expect – seamless, proactive, and personalized.

Moving beyond business as usual

In Forrester’s Predictions 2023: Customer Experience report, the research firm suggests that context – more than channels – will drive experience in the coming year. This means brands will have to think harder about how and when their customers engage with them. Businesses will need to know how best to meet customers where they are, regardless of channel.

Customer engagement is a vital aspect for any business. By prioritizing authenticity and personalization through the use of AI technology, businesses can build trust and loyalty with their customers, leading to long-term success.

Wondering how AI Talkbots can boost your customer engagement efforts? Schedule a call with one of our consultants to explore the possibilities
Contact us

19
Jan
Talkbot Basics  ·  Voice AI Technology
AI technology in finance: from concept to implementation
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

In a world of increasing data volume, managing spreadsheets and numbers is becoming more and more complex. Today’s finance organization needs a new set of tools to keep up with this fast-paced development. Enter AI technology. 

In this blog post, we present insights taken from Gartner’s webinar on the AI-Forward Finance Organization

Overload of financial data

Analysts predict that data volume will double to 181 zettabytes by 2025, which is a staggering amount. But what makes going through the world’s trove of digital data more challenging is its complexity, rather than its volume. Gartner foresees that data complexity will increase 4x in three years, twice the increase in volume. At the end of the day, it’s not just how much data we will get – it’s how much harder it will be to make sense of it all.

Many organizations today rely on spreadsheets for its operations, including financial functions. While these tools were powerful in their time, spreadsheets – even macro-enabled ones – are continuously being strained by high data volumes and complex data interconnectivities. It’s not uncommon to find broken links within spreadsheet models, or calculations that fail to present expected results due to technological limitations. The increase in data volume and complexity needs a smarter, more flexible solution.

Artificial Intelligence in finance

The rise of artificial intelligence (AI) in recent years has led to solutions that answer the data accessibility and navigability question. Enterprises are recognizing the power of AI and as of 2022, 59% of organizations surveyed by Gartner Research have started an AI initiative or adopted AI into their processes. And while this is a promising sign, it also speaks of risk to AI laggards. Businesses that fail to embrace AI technology to streamline processes will fall behind even faster in today’s accelerated market.

To embark on an AI journey, the key thing is to embrace change – to think differently and learn new things.

How to begin your AI journey as a finance organization

Start top-down – digital transformation, especially for finance-related activities, gets stronger buy-in when top management initiates the change
Understand that AI deployment is non-linear – adopt an agile working mindset and be willing to have more cyclical deployment patterns
Be open to experimentation – most organizations adopting AI take five iterations to get right things right or succeed

How to pitch a pivot to AI

Many leaders see the value of AI but need to get buy-in from upper management or other stakeholders. Here are some points that can help bring people into the fold and accelerate your company’s AI adoption.

Take the quick adoption route by buying software or subscribing to services. 
Position it as a proof-of-concept exercise where further adoption is dependent on results gained. Taking a packaged solution is one of the most frictionless ways for finance companies to dip their feet into the AI pond. For organizations that require debt collection or payment reminders, WIZ Talkbots are an easy way to introduce AI technology and quickly reap the benefits of AI-powered automation

Optimize human–machine collaboration.
Some of the pushback when it comes to AI stems from the fear of losing jobs. To overcome this, AI advocates should find ways where AI will complement human efforts. Best examples are process automations where humans are still needed for exception handling or complex cases. Remember, Humans are great at strategy and handling exceptions to rules, as well as seeing the big picture and drawing insights. Machines or AI are great at calculating, analyzing, executing processes, sending warnings at critical points, and enforcing rules or guidelines. Find the right balance between human and machine, and communicate how AI can empower humans to do more in less time or with less cost.

30% of businesses with advanced AI adoption report seeing better results than expected
– Faster implementation for new projects
– Greater business impact
– More process efficiencies
Significant AI adoption increased likelihood of financial benefit by 5x

Whether you’re a cutting-edge fintech or a well-established traditional, AI solutions should be in your future-proofing arsenal. And adopting a pre-built solution is one of the easiest ways to get started on your AI journey.

Looking to quickly deploy an AI solution into your debt collection operations? Our consultants would be happy to help
Contact us

13
Jan
Talkbot Basics  ·  Voice AI Technology
How to transform voice data into actionable insights

Many businesses, particularly in the financial services industry, have a treasure trove of data that’s just waiting to be uncovered. “Businesses which have high call centre demand typically store voice data in a system for seven years,” shares WIZ.AI Chairman and co-founder Jianfeng Lu, in an interview with Startup SG. “However, the data is unstructured and not stored in a uniformed way that allows for business, product, and customer improvement.”

Enter artificial intelligence (AI), specifically voice AI. By extracting information from call recordings, voice AI can help companies identify customer trends, detect customer sentiment and provide valuable insights into customer behavior.

Unlocking the power of voice AI

The first step in getting value out of voice data is in transcribing the calls into text. Once the calls are transcribed, the data can then be organized more intelligently. “Voice AI digitalises unstructured voice data by hashtagging and categorising calls,” Jianfieng explains. This newly structured data then becomes valuable to companies that have years of saved recordings.

One of the great advantages of AI is the speed with which machines can process data. Instead of having a human listen to hours of recordings and draw insights manually, companies can use an AI engine to transcribe, categorize, and extract useful data in just minutes. What’s more, AI can capitalize on the power of big data – having large volumes of information to draw meaningful conclusions from. These types of insights would be impossible to find with just a smaller data set, or when doing spot checks and listening to random recordings.

The automated data collection capabilities of AI-powered systems brings data analytics to the fore, allowing teams to leverage data-driven decisions for long-term success.

Continuous improvement from data analytics

With this vast computing power, voice AI can empower businesses to gain a deeper understanding of their target audience. By listening to conversations, voice AI can record how customers interact with a specific product or service. This information lends insight into customer priorities and preferences, which is valuable for future product development. 

Voice data can also provide deeper insights into the customer journey. Artificial intelligence can help businesses uncover customer preferences and behaviors. Machines can drill deeper into multiple customer conversations, picking up cues that may otherwise be overlooked. With voice AI, businesses can make better decisions that are in line with their customers’ needs, enabling them to provide more personalized and effective experiences.

Voice AI can also uncover trends in customer feedback, flagging a rising pattern in product or service issues and prompt a more proactive response. Businesses can use this data to avoid potential customer dissatisfaction by pushing out service messaging even before issues arise. Because voice AI can also detect customer sentiment, companies can use extracted data to optimize their customer experience. Leaders can quickly identify areas of improvement for customer experience management.

Data-driven decisions for personalized experiences

Voice AI analytics is a powerful tool for businesses to gain insights about their customers and make more informed decisions. By utilizing analytics on voice data, businesses can leverage customer data to better understand their customers, anticipate their needs, and personalize the customer experience. This technology can also be used to identify potential opportunities and areas for improvement, providing invaluable insights that can be used to create more relevant products and services.

Voice AI will grow to play an ever-important role in customer service, marketing, and other data-driven activities. Companies are already beginning to leverage voice AI to create more efficient processes, personalize customer experiences, and develop innovative solutions. Voice AI can be used to automate and monitor first-level customer interactions, understand customer sentiment, automatically classify data, and to generate automated reports. But ultimately, the access to data which companies already have is what makes voice AI extremely powerful. “I foresee a future where enterprises will have a rich data pool for each customer,” shares Jianfeng. “This will help businesses better serve their customers’ needs.”

Contact one of our specialists today to take that first step in harnessing the untapped insights in your voice data.
Contact us

05
Jan
Talkbot Basics  ·  Voice AI Technology
CX trends and how to make the most of them
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

The customer experience space is changing rapidly. As new technologies emerge, customer behaviours shift – but sometimes not in the way businesses expect. Here are the top CX trends identified by Gartner research, and how brands can leverage these changes.

Proactive service prompts more engagement, not less

Outbound customer engagement is on the rise, and rightly so. In a world that competes for audience attention, proactive service helps keep brands top of mind. Gartner predicts that proactive customer engagement interactions will surpass reactive customer engagement interactions by as early as 2025.

Interestingly, customers continued to engage with a brand after proactive interactions. The nature of the additional interactions are also interesting to note – instead of opting to use self-service channels, customers gravitated towards assisted digital or voice channels after the initial outreach.

In general, proactive service improved customer satisfaction (CSAT) ratings. Brands can ride on this trend and build greater customer loyalty by designing smart proactive service.

How to capitalize on this trend

Now is a good time to revisit customer journey maps and update them based on gathered data. Try creating new prompts that push a customer further down the funnel or along the journey. Prompts to finish incomplete applications – say, for loans or credit cards – is a simple example. WIZ Talkbots have been deployed for this type of use case, helping customers complete application forms through assisted channels. The proactive service can start as a follow-up call regarding the open application, then continue with tips or guidance on how to complete more complex fields.

Customers fall back on familiar channels, even when resolution is slower

Humans are creatures of habit; customers tend to rely on previously used channels to resolve issues or raise concerns. Even when new, more efficient channels are presented, customers go back to the old methods. This is because customers underestimate the time it takes to resolve their case. Many times, this results in abandoned self-service operations.

32% of customers actively engage in a customer service phone call for more than 10 minutes

75% of customers who use the phone report the interaction taking longer than expected.

How to capitalize on this trend

To avoid abandoned calls or self-service tasks, brands can use AI to guide customers through self-service. This starts from the first point of interaction, which can be the company landing page. AI-powered interactions can then move customers down the funnel based on the steps taken, be it answers to FAQs or guided scripts for common transactions. Brands can make the most of previously collected interaction data to design scripts that address the most common queries. Over time, as more data is collected, businesses can refine scripts and update customer journeys.

Customers don’t mind switching channels to resolve issues if the experience is seamless


Omni-channel engagement is becoming more and more important in customer experience. All customer touchpoints now need to be interconnected and integrated so that handoffs become seamless and easy for customers. Success in omni-channel engagement hinges on customers expending the least amount of effort when moving from one channel to another. Then there is the issue of consistency – customers expect that each channel will have access to their data and history from the previous channel. Maintaining consistent experiences across channels will become the test of great customer engagement.

How to capitalize on this trend

Brands must ensure that data from one channel is completely transferrable to another, and that customer interactions are carefully interlinked. Customer histories and profiles should be kept updated, no matter which touchpoint was last used. Language and tone must also be kept consistent across all channels, to make customers feel that they’re only talking to one person or entity. Having a centralized omni-channel platform will help keep things sychronized and seamless.

“Value enhancing” service experiences drive retention and growth

Great service experiences only help to avoid attrition; good service has become the minimum expectation. To improve customer loyalty, service experiences should help customers get more value out of their purchase or subscription. A product or service is the reason why customers patronize a brand in the first place. Value enhancing services could come in the form of tips or guides, or even short testimonials on how other users are maximizing the product or service.
When customers experience a value-enhancing service experience, they have:

82% probability of being retained 
86% probability of spending more money 
97% probability of sharing positive word of mouth

How to capitalize on this trend

Understand how customers use your product or service, and help them get more out of it. This could be another area where proactive service comes into play. Quick prompts on how things have been so far can lead to a series of how-to articles or videos that guide users into a richer product or service experience. Value enhancing service can also bring about opportunities to upsell, once the desire for greater value has been established.

How does AI come into the picture?

These CX trends are closely linked to the possibilities presented by artificial intelligence and AI-powered automation in customer service. Forward-looking enterprises harness the power of AI technology to stay ahead of the curve and deliver delightful customer experiences – services that are seamless, personalized, and value-enhancing
To find out how AI can help power your customer experiences, contact one of our specialists today.
Contact us

29
Dec
Talkbot Basics  ·  Voice AI Technology
How AI can boost results in debt collection
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

For many companies that need to conduct payment collections, resolving accounts that are past due is one of the trickiest parts of the business. Many of these customers are hard to contact, and once contacted, getting commitment to settle dues is another hurdle. Fortunately, conversational voice AI can help.

Harnessing the power of AI technology for debt collection

Artificial intelligence can be used in the debt collection process in a number of ways. First, AI can be used to automate manual processes like transcribing calls and categorizing customers based on their willingness to pay. These two tasks alone provide companies greater control over their debt collection processes because of the wealth of data that gets collected. Automatic transcription of calls transforms unstructured voice data into structured data. With it, enterprises can now do more detailed data analytics of their customer interactions. Those tasked with debt collection can gain a better understanding of risk behavior and improve collection tactics, moving from a reactive position into a more proactive one.

Loan provider gets more payment commitments with voice AI calls

This example features a finance company that provides loans and similar products to consumers. One of its biggest challenges is reaching accounts that are past due and making sure that commitment to payments are made. This often requires persistent outreach and multiple calls, which take a lot of time and effort for human agents. By switching to an AI-powered agent, the company was able to ramp up its call intensity proportionally to the account’s days past due.

call intensity 1-day, 2-day, and 3-day past due customers

With artificial intelligence, the company was able to call customers with overdue payments more consistently and with greater frequency. This resulted in obtaining “promise to pay” commitments from more customers – up to 60% for accounts that are three days past due.

Traditional bank gets clearer picture of consumer credit portfolio

In this second example, a large regional bank decided to deploy voice Talkbots for their credit collection operations. With each call, the AI engine records outcomes and tags customers accordingly.

The first layer of tags involve the call status – whether the customer was contacted or if the call failed to connect. Once the call connects and the Talkbot is able to speak with the customer, the AI engine then identifies whether or not the customer is willing to pay. If agreement is obtained, the AI agent then secures a payment date within a three-day window and records the response.

Having collections information structured in this manner enables the bank to see just how many of their delinquent accounts were willing to pay, and how soon. With this information, the bank is able to forecast its cashflows more accurately and update its collections strategy more proactively.

Fintech company gets record results in three-day collections campaign

This final example involves a non-traditional financial services provider that undertook a three-day call campaign for uncollected payments. Utilizing voice AI Talkbots, the fintech company was able to reach almost half of its customer database. Over 300,000 calls were made across a three-day period, obtaining payment commitments from two-thirds of customers that had overdue accounts. 

Of those that had committed to pay, over half promised to settle their dues within the day. Almost a fifth committed to pay the next day, bringing committed payments to 71% within a 48-hour window.

Getting ahead of back payments with artificial intelligence

AI powered much of the success of the above examples – artificial intelligence is what enabled companies to undertake the huge volume of calls to customers. And thanks to well-designed scripts, these companies were able to obtain payment commitments that previously eluded them. Results can be seen even in a short three-day campaign.
Are you looking to improve the outcomes of your debt collection activities? See how our voice AI Talkbots can help by booking a demo with us today.
Contact us

22
Dec
Talkbot Basics  ·  Voice AI Technology
Spot and tag credit risks with voice AI technology
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

For many financial institutions, consumer credit risk is the biggest risk on their balance sheets. This is underscored by the reality that credit risk is the biggest cause of bank failures, according to Van Greuning and Bratanovik of the World Bank.

To effectively manage credit risk – specifically consumer credit risk – companies in banking and finance need to accurately assess an individual’s ability to pay back a loan on time and in full. This means they need rich and accurate data to make the right judgements. But even with advances in technology and digital transformation, financial institutions still face challenges with data for credit risk management.

 

How AI technology is transforming risk management

Thankfully, artificial intelligence is changing how companies manage consumer credit risk. Voice AI technology is particularly helpful in unlocking data that has already been collected but just sits in storage. One prime example is customer call recordings, which are stored as voice data. While these recordings may be useful as reference, much of them sit as unstructured data that take time to sort and analyze. Recordings are usually not used or reviewed except for quality checks and investigations into specific incidents.

With AI technology, rich customer information can be extracted from this previously untapped resource – data that may be useful for credit risk assessments. Artificial intelligence enables call recordings to be quickly transcribed and converted into structured data, which can then be used for analytics. Within all that new data would be signals or keywords related to risk management.

 

Detecting intent with natural language understanding

Transcribing and structuring voice data is only the first step in making use of conversational AI capabilities. One of voice AI’s more powerful capabilities is intention detection – figuring out what a customer plans or wants to do, based on verbal cues. Questions like “How much do I owe again?” or “What’s the minimum amount due?” can signal an intention to settle the account soon. When the AI neural engine detects these words, it can automatically tag the caller as “willing to pay” and proceed to obtain a commitment date on payment.

Intention detection is useful for less common circumstances, too. Keywords like “family emergency” or “job loss” can signal personal distress, and the AI engine can then tag the call as “special case”. The call can be routed to a live human agent, who can then talk the customer through workable options to refinance their debt. This helps banks and financial institutions take a more humane approach towards their customers who might be going through a rough patch. It’s a win-win situation: customers feel cared for, and the finance company gets an early opportunity to rebalance its risk portfolio.

 

Using data analytics to inform strategy

Another key benefit of having calls tagged automatically by AI is the ability to notice patterns as they arise. For instance, if a significant number of calls have been tagged with “don’t know how to pay”, the bank can deploy an information campaign on payment modes. This reduces the risk of non-payment for future customers who may struggle with how to pay their bills online or offline.

What’s more, AI labels can be customized based on business patterns. Companies can choose any number of tags to attach to their voice data, using the most relevant information that can shape their decision-making. Tags can vary from committed payment dates – which help companies follow-up on payments – to payment modes, which inform the bank which channels are more effective for collections.

With more data in their hands, banks can also take a wider view of their risk portfolio and recalibrate their strategy in a more proactive manner. Tags such as “invalid number” or “uncontactable” can signal a bad debt, and keep companies from wasting time on chasing down a non-payer.

The use of smart tags and automatic labeling have already helped a number of fintech companies and traditional banks optimize their debt collections. And these are only some of the ways artificial intelligence can help streamline financial operations.

To find out other ways AI can improve your processes and build smarter customer engagements, speak to one of our specialists today.
Contact us

14
Dec
Customer Stories  ·  Voice AI Technology
JAGADIRI achieves 5.3x higher cost efficiency with WIZ Talkbots
illustrate better return-on-investment

JAGADIRI works to protect the aspirations of its many Indonesian customers under the umbrella of PT Central Asia Financial (CAF). JAGADIRI provides life insurance, accident, and health insurance through digital marketing, telemarketing, and direct marketing channels.

“The thing with insurance is that it’s not like banking,” began Ibu Fitriah Betan, Head of Customer Experience at JAGADIRI. “As an insurance provider, JAGADIRI needs to make premium reminder calls to existing customers and continue providing service even up to 90 days past due.” Because the actual need for insurance happens in urgent situations, and the company does all it can to keep customers protected. On their part, customers need to pay their premiums promptly, in order to ensure that their policy remains active and can cater to unexpected incidents.

Creating a consistent customer experience with conversational voice AI

The need to ensure continuous coverage is why JAGADIRI puts a lot of effort into premium payment reminders. But the challenge is timing. Contacting customers for premium reminders can be tricky, as customers may be busy with their activities during the day. Another factor lies in the personal nature of the interaction – the mood of the agent or the way he or she communicates varies from one call to another. JAGADIRI wanted a more consistent way of engaging with its customer base, and ensure that these customers remain with the brand. 

In June 2022, JAGADIRI decided to transition its premium reminder calls to the WIZ.AI Talkbot. The implementation was done in phases; by September 2022, all premium reminder calls were automated. The results spoke for themselves. “Before Talkbot, we get less than 10 percent of connected calls for premium reminders. Now we get, on average, around 50 percent,” explains Ibu Fitriah. Talkbots are also 5.3x more cost-efficient than using human agents; the cost ratio for Talkbots is 6 percent, compared to 32 percent for agents. 

Using WIZ.AI’s solutions brought very tangible results in our premium reminder deployment. Before Talkbot, our agents get less than 10 percent of connected calls. Now we get, on average, around 50 percent.

– Ibu Fitriah Betan, Head of Customer Experience at JAGADIRI Tweet

An evolving partnership based on two-way communication

Beyond the results, JAGADIRI loves that WIZ.AI regularly checks on how the Talkbot is working, and provides recommendations on how to optimize the AI solution. From sales to project management to customer success and even customer experience, the WIZ.AI team was ready to help JAGADIRI make the most of its Talkbot. “WIZ.AI CX Designers help us refine our strategy by improving the scripts based on the data we get,” shares Ibu Fitriah. “There’s excellent two-way communication between us and the WIZ team. WIZ continually suggests how to update the Talkbots and implementing the suggestions improves our results.”

With the success of automated premium reminders, JAGADIRI is now looking to deploy the WIZ.AI Talkbot for more use cases. As of this writing, the team is developing debt collection scripts for due and overdue accounts. JAGADIRI is also planning to deploy Talkbots for Welcome Calls to new customers. “Talkbot has made contacting our customers much easier for the CX team,” concludes Ibu Fitriah.

The best thing about our experience is that WIZ.AI CX Designers are always helping us refine our strategy. There’s excellent two-way communication between us and the WIZ team. WIZ continually suggests how to improve the performance of the Talkbots and implementing the suggestions improves our results.

– Ibu Fitriah Betan, Head of Customer Experience at JAGADIRI Tweet
To find out how you can leverage voice AI for your customer engagements, speak to one of our specialists today.
Contact us

30
Nov
Talkbot Basics  ·  Voice AI Technology
Deal with surges in customer queries using voice AI
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

The year-end holiday season is upon us, and for many businesses, this is typically when things get really busy. Industries like retail and hospitality often see a surge in sales and inquiries – and as a by-product, an increased need for customer support.
 
If the increased volume were not enough, the holiday season is typically when everyone gets put under pressure. Shoppers are pressed to make their purchases in time for parties and events. Callers have to navigate competing demands for their attention. And call center agents tend to bear the brunt of customers’ short fuses and higher demands.
 
This is true for non-seasonal surges, too. Unforeseen changes in external regulation or policy can affect businesses in a critical way. New travel restrictions are a prime example; back in 2020, companies in the travel industry got inundated with calls and emails practically overnight. When scenarios like this happen, agents are typically pushed into high-stakes, high-stress situations. This comes on top of the mounting pressure to answer a growing number of calls.
 
In some ways, managing surges can be like a game of Russian roulette – you’re not quite sure when you’ll get shot. Fortunately, these risks are easy enough to understand and manage, with the right tools.
 

Inelastic resources can result in insufficient live support

For most call centers or customer support departments, resources and headcounts are fairly fixed. This lack of elasticity means sudden surges in support requests are hard to meet. Resources – specifically call center agents – can get really stretched. Say an agent typically handles 200 calls a day; that number can easily double to 400 calls on a peak season. Unless extra resources were brought in to handle the surge in volume, there will be potentially 200 unaddressed calls, or at the very least, poorly addressed calls. Businesses never want to be in a position where they have 200 unhappy callers – customer loyalty will take a hit, and eventually, so will sales.
 
Hiring additional agents would be the natural solution to this resource crunch. And while this might be workable, training and onboarding agents take time. Companies often spend at least two weeks to a month before deploying a new agent into the live environment. If you’re looking to simply fill a temporary gap, training new people may not make sense, especially if the cycle needs to be repeated after six months.
 
And that’s if talent is readily available. Recruitment is a time consuming process in and of itself. Finding the right people to serve customers is a tricky business; candidates must possess not just the skills, but the right attitude and in some cases, a love for the brand they represent. What’s more, contractual hires for short-term engagements can be even harder to find in a tight labor market. Sometimes this talent problem is solved through outsourcing – but quality can become an issue when it comes to service delivery.
 

Using AI customer service for busy seasons

Addressing elasticity and scaling issues without compromising on quality can be done by deploying artificial intelligence (AI) agents for customer service. Unlike human agents, machine-powered bots can take on more queries as and when they come. A human agent can probably take 20 calls in an hour; in contrast, an AI-powered bot can handle one million calls in the same timeframe. This makes AI call centers more flexible than a traditional call center manned by only humans.
 
Training bots also take less time than training humans. Once a proper script is in place, a Talkbot can be deployed and take customer calls within days. Machine learning capabilities enable the bot to adapt to situations intuitively, collecting more data points and refining its process as it goes along. For queries it can’t address, the bot redirects calls to a human agent that’s identified as the best person to address the issue.
 
Because they run on scripts and programmed dialogues, Talkbots can easily be redeployed to address rapidly developing situations. A call center or customer service team can choose to quickly shift AI bot resources to another function – from products to payments, for instance – in as short as a day’s time. This makes bots a lot more flexible than traditional hires.
 
Finally, AI call center bots can be decommissioned as quickly as they are deployed. This means a company doesn’t need to keep paying for a service it no longer needs. Services can be scaled down once the peak has passed, and more bots can be brought back on board once the business sees another surge coming.
 

Using data from artificial intelligence to stay on top of change

One often overlooked advantage of using AI bots is the wealth of data businesses get from customer interactions. This is particularly true for voice data, which comes as unstructured, messy, and hard-to-extract fragments of information. Unstructured data makes analytics a lot harder and time-consuming. With AI, voice data can be quickly transcribed, tagged, categorized, and subsequently analyzed for insights.
 
As an example, suppose a retailer has put one of its popular toys on a special weekend sale. Several parents buy the product, and a week later, a number of them call the store asking if they can have the item changed. In most cases, the agent will ask for the reason for a return or exchange, make notes, and file the information for reference. The data may not be looked at until after the surge has passed.
 
With AI customer service, the call will be transcribed in almost real-time, the data structured and analyzed, and patterns reported on dashboards almost immediately. The retailer may learn that 80% of buyers changed their minds because buyers realized that their child is no longer interested in the toy. This alerts the retailer to changing buyer preferences and may lead them to reassess their product inventory.
 
Moving it even further, if the AI agent is integrated with the business’ CRM system, the bot can make intelligent suggestions to the shopper based on purchase history. The machine can then intelligently offer products based on known preferences. If the database shows that the caller is a loyal customer, the bot can initiate a promo that gives an extra 10% off on purchases made that day. The result is a delighted customer, and a new closed sale.
 

Providing great customer care at reasonable costs

For many businesses, a customer satisfaction (CSAT) score of above 95% is the holy grail of customer experience. But to provide that level of service often means investing a lot into live support. This tricky balance can be thrown off when seasonal surges happen. After all, nobody can precisely determine how many additional agents they’ll need. Hire too many and you end up spending more than you should, hire too little and you get irritable customers waiting for long periods of time.
 
The good news is that AI call center bots can help manage that balancing act of delighting customers while managing costs. With AI-powered agents, customers will experience shorter wait times and human agents get more manageable workloads. Because a huge chunk of easily addressable queries are managed by machines, humans have more bandwidth to handle the most important tickets – giving the right attention to the right issues. Businesses end up with happier customers and fewer unresolved queries.
 
Artificial intelligence can also empower companies to be more proactive rather than reactive to sudden market shifts. Companies can prepare for upcoming surges by training bots once they see changes on the horizon. Furthermore, data collected by AI bots can be quickly analyzed and used to inform strategy to help mitigate new risks.
 

Customer service that’s personalized and deeply human

Today’s technology makes it possible to deploy bots that speak and text like the humans they serve, providing a personal touch that customers crave. That’s something that outsourcing your customer care operations will find difficult to achieve, especially if services are offshored. In customer service, shared context matters a lot in creating enjoyable experiences.
 
By using AI to automate low-value tasks, businesses are in a better position to deliver high-value services that delight, creating experiences that breed customer loyalty.
Wondering if AI customer service is the right solution for your business? Speak with a representative today.
Contact us

17
Nov
Talkbot Basics  ·  Voice AI Technology
Understanding Asia – Natural Language Processing in AI
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

NATURAL PROCESSING LANGUAGE

 Natural processing language is an aspect of artificial intelligence and computer science that handle the interface between human languages and computers. It involves the computational modelling of different characteristics of language and the deployment of variety of systems. These systems include spoken language systems incorporate natural language with speech. NLP works with linguistic computational features, it employs computer in comprehending, handling speech and text of natural language to achieve useful feat. There are several fields NLP can be applied to; speech recognition, expert systems, artificial intelligence, cross language information retrieval (CLIR), text processing, language translation, speech recognition, and user interfaces. This innovative technology is saddled with getting computers to communicate and process human languages, and perform closer to human level of language thoughtfulness. Computers are yet to reach same instinctive comprehension of natural language like humans do. There is clear difference in the method in which human communicates with one another and the way they do with computers. During program development phase, the structure and syntax are carefully selected to suit the intended task, unlike conversing with other people whereby a lot of freedoms are considered. Ranging from sentence length, sarcasm and jokes, to several ways of expressing same thing.

Recent advancement in innovative technologies has enabled computers to perform range of things with human or natural language. Deep learning supports the implementation of programs to perform task like text summary, language translation, and semantics understanding. The rise in the implementation and application of artificial intelligence to our daily activities has made it ubiquitous. It is imperative for human to be able to communicate more with computers in the language we are familiar and comfortable with, speaking to computers in their natural language. Natural Language Processing (NLP) is seen as the canopy term that binds other natural language technologies which include Natural Language Understanding (NLU), Natural Language Generation (NLG), and Natural Language Interaction (NLI).

COMPLEXITIES OF UNDERSTANDING DIFFERENT LANGUAGES USING NATUAL LANGUAGE PROCESSING

Recently, significant feat has been recorded in enabling computers to comprehend human language using Natural Language Processing (NLP). Nevertheless, the multifaceted multiplicity and dimensionality features of data sets, make the execution a problem in some cases. Concerning implementation of NLP in Asia, with main focus on south East Asia, voice and text-based data and their practical applications will vary. In other to capture the whole process, NLP needs to include several diverse procedures for interpreting Asia local language. It could involve machine learning, statistical, algorithmic, or rules-based approaches. Ambiguity is an aspect of cognitive sciences without a definite resolution, range of language ambiguity differs greatly based on the speaker. Technically, any language sentence with plenty grammar can generate another meaning, for human to find it challenging in dealing with conversation vagueness sometimes, then it is inevitable for natural language understanding systems.

  1. TYPES OF AMBIGUITY

Outlining ambiguity can sometimes seems vague. There are different forms of ambiguity regarding natural language processing (NLP), and artificial intelligence (AI) systems.

  1. Lexical Ambiguity: This is a single word ambiguity. A word can be ambiguous with respect to its syntactic category. Lexical ambiguity can be decided by Lexical type clarification like parts-of-speech labeling. It also stores word and complementary knowledge.
  2. Syntax: This is a part of grammar that define how words are assembled and linked with one another to make a sentence. Syntax involves the transformation of a linear order of tokens (a key to each word or punctuation mark in natural language) into a classified syntax tree. The main issue with syntax level are: sentence assembling, speech tagging, and identifying syntactic categories.
  • Semantics: This type of ambiguity is characteristically associated with sentence interpretation. It includes task like interpreting one natural language to another, synonyms searching, creating question-answering systems, and clarification of word sense.
  1. Morphology Ambiguity: This ambiguity came into being due to advance processing carried out on the root words to make use of them in a specific sentence. It involves processing of word forms.
  2. Discourse: Discourse level processing needs a pooled knowledge and the interpretation is carried out using this context. Anaphoric ambiguity comes under discourse level. One of the exhausting task in Natural Language Processing (NLP), some of the problem are belief, sentiment, and user intention processing. It also process connected text.
  3. Pragmatic Ambiguity: This is refer to the situation whereby whereby the circumstance of a phrase gives it multiple meaning. It involves user modelling, and intention processing.
  • Referential Ambiguity: When a phrase or a word in a particular sentence could refer to two or more properties or things, it is referential ambiguity. It is always clear from the circumstance which meaning is intended but not always.
  • Phonology: It is described as words that sound the same way but have different meaning. This type of ambiguity forces the NLP model to interpret the context of the sentence and place it in the right context. It can be referred to processing of sound.
  •  

STAGES IN NATURAL LANGUAGE PROCESSING (NLP)

Basic steps necessary to be followed to build Natural Language Processing (NLP) model are as follows:

Stage 1: Segmentation of Sentence

The first stage required to build NLP model is breaking of prearranged paragraph into single sentences. This is done to process the meaning line by line.

Stage 2: Word Tokenization

After sentence segmentation, it is followed by word extraction from each sentence one after the other. The tokenization algorithm can be programmed to identify a word whenever a ‘space’ is observed. All these would be achieved following Asian natural language.

  • Stage 3: Prediction of Parts of Speech

It involves classifying words into their respective part of speech as duly represented in Asian language. Parts of speech classification will help the machine learning model to comprehend its role in sentence. Machine learning might not actually know the meaning of each word in sentence setting the way human being do. A lot of data has to be fed into the model along with precise label of each word’s meaning and part of speech.

  1. Stage 4: Text Lemmatization

The machine learning model learns to identify the most basic form of words in a sentence. By differentiating between closely related words.

Stage 5: Pinpointing Stop Words

This stage is saddled with identifying the importance of each word in a sentence. There are a lot of filter words in that appear frequently in English language, and it is definite that Asia Language will also have some commonly used filter words that introduces a lot of noise into a sentence. It is necessary for machine learning to identify them and flag them as stop words i.e. words that can be filtered out before undertaking statistical investigation.

Stage 6: Dependency Parsing

It is the stage where grammatical laws of Asian language would be employed to identify how words are related to one another

Stage 7: Entity Analysis

This is achieved by going through the entire sentence in Asian Language and identify all the important words in the text. And the words in the sentence will be categorized as been programmed to work.

Stage 8: Pronouns Parsing

This is the last stage in building NLP model and it is one of the hardest stage. This stage will employ machine learning to keep track of pronouns with respect to the sentence context. It is very easy for human to comprehend the meaning right from the context of the sentence unlike computers. Therefore, a Machine Learning model is required to be fed with a lot of data alongside correct tags for the model to be able to identify the pronouns effect in a sentence.


17
Nov
Talkbot Basics  ·  Voice AI Technology
The Rise Of Conversational AI In Customer Service
Jennifer Zhang

Jennifer Zhang

CEO & Co-Founder

THE RISE OF ARTIFICIAL INTELLIGENCE IN ASIA AND THE WORLD

First coined by American computer scientist John McCarthy in 1956, artificial intelligence (AI) refers to the cognizant abilities of machines which have been programmed to autonomously perform, think and learn like a human. Today, from unassuming home appliances such as robotic vacuums, to more awe-inspiring projects such as autonomous cars and self-learning weather forecast technologies, AI has permeated into various aspects of industry and society. As its applications become increasingly adapted into various aspects of our lives, its ability to bring about the digital transformation of the world would pose various disruptive implications for the economy, environment and our day-to-day lifestyles.

Arguably, the main attraction or headliner of AI, as it has captured the attention of the commercial world, is its potential to drastically increase productive and cost efficiencies. It’s transformative effects have been echoed by various industry experts, who have projected that such technologies would boost corporate profitability in 16 industries across 12 economies by an average of 38% by the year 2035[1]. By eliminating repetitive, low level tasks which were traditionally performed by humans, one of AI’s most noticeable disruption would definitely be on the labor market, as companies look into using AI to optimise internal operational process and alter the way service and products were originally offered to their customers. The nature of long-established roles are also likely to evolve in the meantime.

Global investment in AI is growing rapidly, with an estimated investment of $26 to $39B investment in AI. At present, the two major global hubs of AI development are the United States and China. Funding for artificial intelligence companies in the United States has increased exponentially in recent years, growing from around 300 million U.S. dollars in 2011 to around 16.5 billion in 2019[2]. For China, PricewaterhouseCoopers predicts that $7 trillion of China’s $38 trillion GDP by 2030 would be attributed to AI through new business creation in fields such as autonomous driving and precision medicine, as well as existing business upgrades in terms of improved efficiencies and reduced costs[3].

While the bulk of AI investments are presently made by tech giants Alibaba, Amazon, Google, Baidu and Facebook in a race towards AI as a long-term strategy for business sustainability and competitiveness, AI as a concept is still poorly understood and intimidating to the ASEAN region. However, despite the region’s relatively slow advances in AI technologies, its oncoming impact is undeniable. Previously published MGI research estimated that currently demonstrated technologies have the potential to automate roughly half of the work activities performed in ASEAN’s four biggest economies: Indonesia (52 percent of all activities), Malaysia (51 percent), the Philippines (48 percent) and Thailand (55 percent), with these tasks currently generating more than $900 billion in wages[4].

Out of the ASEAN member states, Singapore, as the region’s technology capital, has made the greatest AI advances thus far, and is a natural first choice for AI tech startups to establish their presence in Asia. WIZ.AI, having its research institute based in Nanjing, China, established its first overseas headquarters in Singapore in 2019, from whence it aims to continue developing its pioneering and proprietary conversational AI talkbot technologies and push it out to ASEAN. 

 

CONVENTIONAL CHATBOTS VERSUS AI CONVERSATIONAL TALKBOTS

2.1 THE RISE OF CHATBOTS IN CUSTOMER SERVICE

Chatbots are computer programs built on the concept of artificial and data analytics, and is commonly installed on websites or social media platforms. With a chatbot application,  companies are able to automatically respond to customer messages round the clock, through a virtual assistant which recognises entered keywords and is able to provide instantaneous, standardised text replies and guidance to the customer.

 

This rise of chatbots is projected to be significant; as reported by Global Market Insights, the chatbot market will be worth $1.34 billion by 2024, with 42.52% of that alone from the customer service sector[5]. The rise of chatbots is tightly linked to new technological advancements and evolving customers’ expectation of brand interactions. With the prevalence of social media and mobile messaging applications, the average consumer now expects a company’s to resolve issues and respond to requests with speed.  Based on a 2017 customer survey conducted by Microsoft, 54% of those polled expressed higher expectations for customer service today compared to the previous year, with the percentage increasing to 66% for younger respondents at the 18 – 34 age group[6]. Falling back on traditional forms of communication such as emails or text messages are no longer acceptable; in the same survey, 68% of the respondents have a more positive view of brands which take the initiative in providing proactive customer service notifications.

 

In addition to it being used as a medium to provide basic customer service, chatbots are also suitable for use in marketing & sales of products, which further spurs demand and market growth in this sector. These platforms aid companies to expand their reach by connecting with a larger audience, aiding in decision making by addressing customer queries on the spot and subsequently pushing suitable product recommendations.

 

2.2      WIZ’S AI CONVERSATIONAL TALKBOT – ONE STEP UP FROM CHATBOTS

In the 2017 Global State of Customer Service Report by Microsoft, email and telephone are still the primary communication channels for many customers, while live chat, self-service, social media, and chatbots are relatively lower in terms of raw volume[7].

 

It would seem then that maintaining a hotline available for urgent customer queries 24/7 is key to keeping customers happy and loyal. Despite the increasing number of self-service options made possible by technology, customers still express a preference for live-agent support, with 30% of global respondents frustrated when they are not able to do so. The value of being reachable to customers is further demonstrated when 30% of people polled are of the opinion that being able to speaking with a knowledgeable and friendly agent is a significant factor that makes or break a customer service experience.

 

Working from the same belief that voice conversations are still the most natural way of interaction for humans, and that artificial intelligence is key in the next wave of hyper-personalized customer engagement, WIZ.AI combines the best of both worlds, integrating its intuitive AI conversation capabilities into the simplicity and familiarity of a phone call. Pioneering a new market category, its AI Conversational Talkbots takes centre stage as a cutting-edge, turnkey customer service solution.

 

Capable of understanding human speech in a variety of accents, expressing empathy and engaging in unlimited multi-round conversations with the customer, WIZ’s AI Conversational Talkbots provide a more interactive, human-like and fulfilling customer service experience that is superior to text-based chatbots. Compared to human agents, its Talkbots are also more consistent and reliable. Offered at the fraction of the costs of maintaining a human call centre, WIZ’s AI Conversational Talkbot allows any corporations with heavy customer communication needs to maintain its personal touch with its customers in a cost-effective manner, with endless scaling possibilities for both new inbound and outbound call campaigns. Its backend CRM system captures and sorts through customer intentions in real-time, collecting and digitalising valuable customer intention data.

 

WIZ.AI’s proprietary AI talkbots have been adapted for mass commercial applications in the banking, telecommunications, health care and e-commerce industries. Besides Singapore, WIZ.AI also operates offices in China, Jakarta and the Philippines with a team of scientists, developers, linguists, and dialogue designers.

 

  1. AI – BRINGING CONVERSATIONAL AI INTO THE MAINSTREAM

3.1      COMPONENTS OF WIZ’S AI CONVERSATIONAL TALKBOTS

WIZ.AI’s proprietary AI talkbots can be broadly broken down into three main elements:

  • Natural Language Processing & Understanding: Recognises and processes speech patterns and nuances through the responses of the speaker, allowing the talkbot to understand the intent of the speaker.
  • Automatic Speech Recognition: Recognises and understands local accents and lingo
  • Text to Speech: AI voice tailored to sound human and speak with local accents

 

3.1.1   NATURAL PROCESSING LANGUAGE

Natural language processing (NLP) is a subfield of linguistics, computer science, and artificial intelligence, employing the computer in comprehending and handling speech and text of natural language. There are several fields NLP can be applied to; speech recognition, expert systems, artificial intelligence, cross language information retrieval (CLIR), text processing, language translation, speech recognition, and user interfaces.

In other to capture the whole process, NLP needs to include several diverse procedures for interpreting Asia’s local languages. It could involve machine learning, statistical, algorithmic, or rules-based approaches. The steps to to build a Natural Language Processing (NLP) model are as follows:

  1. Stage 1: Segmentation of Sentence

The first stage required to build an NLP model is the breaking of prearranged paragraph into single sentences. This is done to process the meaning line by line.

  1. Stage 2: Word Tokenization

After sentence segmentation, it is followed by word extraction from each sentence one after the other. The tokenization algorithm can be programmed to identify a word whenever a ‘space’ is observed. All these would be achieved following Asian natural language.

  • Stage 3: Prediction of Parts of Speech

It involves classifying words into their respective part of speech as duly represented in Asian language. Parts of speech classification will help the machine learning model to comprehend its role in the sentence. Machine learning might not actually know the meaning of each word in sentence setting the way human being do. A lot of data has to be fed into the model along with precise label of each word’s meaning and part of speech.

  1. Stage 4: Text Lemmatization

The machine learning model learns to identify the most basic form of words in a sentence by differentiating between closely related words.

  1. Stage 5: Pinpointing Stop Words

This stage is saddled with identifying the importance of each word in a sentence. Similar to the English language, Asia Languages also contains many commonly used filter words that introduce noise to a sentence. It is necessary for machine learning to identify them and flag them as stop words i.e. words that can be filtered out before undertaking statistical investigation.

  1. Stage 6: Dependency Parsing

It is the stage where grammatical laws of Asian language would be employed to identify how words are related to one another.

  • Stage 7: Entity Analysis

This is achieved by going through the entire sentence in Asian Language and identifying all the important words in the text.

  • Stage 8: Pronouns Parsing

The last and most challenging stage of building a NLP model, this step employs machine learning to keep track of pronouns with respect to the sentence context, allowing the bot to comprehend the meaning right from the context of the sentence. To achieve this, the Machine Learning model has to be fed with a lot of data alongside correct tags for the model to be able to identify the pronouns effect in a sentence.

3.1.2 AUTOMATIC SPEECH RECOGNITION (ASR)

This is where a chatbot is differentiated from a AI conversational talkbot – the latter has the added challenge of first having to understand the different accents and local lingo in order to recognise the customer’s intention. ASR training involves collecting speech samples from a variety of language backgrounds, and through machine learning over time, the talkbot gets more adept at deciphering the local accents and lingo.

3.1.3   TEXT-TO-SPEECH (TTS)

Text-to-Speech (TTS) is the channel whereby the talkbot responds and converse with the customer in real-time. WIZ.AI’s proprietary TTS generation system has been designed to be life-like, encouraging customers to share more, and allowing companies to gain deeper insights into customer intentions and needs.

With customisable voices, the customer call experience is further enhanced with a quality TTS which matches the brand persona. The use of a single distinguishing voice sets your company apart from competitors and also ensures brand consistency across all different customer touchpoints.

  1. LOOKING FORWARD

A disruptive technology, AI is transforming the face of customer service and forcing companies to relook into their customer touchpoints and service strategy in order to retain customer loyalty, business sustainability and competitiveness in the long run. Conversational AI technologies are likely to lead this evolution, and as the talkbot product gets better understood and commercialised, we expect more companies in Asia to adopt this product, changing the landscape of ASEAN customer service and economies permanently. 

[1] Accenture, How AI boosts industry profits and innovation, June 21, 2017

[2] Shanhong Liu, Artificial Intelligence funding United States 2011-2019, June 6, 2020 https://www.statista.com/statistics/672712/ai-funding-united-states/ 

[3] PwC Global, Sizing the Price – PwC’s Global Artificial Intelligence Study – Exploiting the AI Revolution, June 2017. https://www.pwc.com/gx/en/issues/data-andanalytics/publications/artificial-intelligence-study.html

[4] Mckinsey Global Institute, Artificial Intelligence and Southeast Asia’s Future, 2017 https://www.mckinsey.com/~/media/mckinsey/featured%20insights/artificial%20intelligence/ai%20and%20se%20asia%20future/artificial-intelligence-and-southeast-asias-future.ashx

[5] Global Market Insights,  Global Chatbot Market worth over $1.34bn by 2024, August 26, 2019 https://www.gminsights.com/pressrelease/chatbot-market

[6] Microsoft, State of Global Customer Service Report, 2017 https://info.microsoft.com/rs/157-GQE-382/images/EN-CNTNT-Report-DynService-2017-global-state-customer-service.pdf

[7] Microsoft, State of Global Customer Service Report, 2017 https://info.microsoft.com/rs/157-GQE-382/images/EN-CNTNT-Report-DynService-2017-global-state-customer-service.pdf


Page 1 of 2