Customer support was one of many earliest features to yield tangible purposes for generative AI.
From summarising instances behind the scenes and aiding brokers on the fly to straight interfacing with clients, the use instances for generative AI in customer support are quite a few, and plenty of companies are both experimenting with or have already deployed GenAI on this means.
However which of them are seeing actual outcomes, how are they benefiting, and what classes can we take from their expertise?
Verizon: AI-assisted customer support brokers raise gross sales
Telecoms big Verizon has been progressively ramping up its use of generative AI in a customer support context.
In Might 2024, the corporate introduced a set of latest customer- and employee-facing purposes together with ‘Private Analysis Assistant’, a conversational AI assistant that may recommend context-based solutions to clients’ enquiries, and ‘Private Shopper/Downside Solver’, which analyses a buyer’s profile to recommend who they is perhaps and why they could possibly be calling.
Since then, the CEO of Verizon’s client group, Sampath Sowmyanarayan, has revealed that gross sales at Verizon rose by 40% because of the brand new expertise liberating up customer support brokers to promote to clients.
“We’re doing reskilling in actual time from buyer care brokers to promoting brokers,” he advised Reuters.
Verizon has carried out the GenAI expertise in partnership with Google, and has said that it allows customer support representatives to comprehensively reply 95% of queries.
The corporate skilled a model of Google’s Gemini massive language mannequin on practically 15,000 inside paperwork, which kind the idea of the responses utilized by customer support brokers, saving them effort and time that may in any other case be spent looking for the suitable particulars.
To information its implementation of generative AI, Verizon convened an AI council, Chief Expertise Options Officer Debika Bhattacharya advised Enterprise Insider final yr. It additionally launched a set of AI rules to make sure the accountable use of AI.
Based on Bhattacharya, Verizon’s goal is to allow hyper-personalisation by generative AI, “operat[ing] at scale however view[ing] every caller as a phase of 1”. The whole lot the corporate does with GenAI, she mentioned, is finished with the objective of creating buyer interactions as seamless as attainable.
Takeaway:
Verizon’s option to reskill buyer care brokers as gross sales specialists and have them dedicate extra consideration to promoting is an attention-grabbing one, however displays the priorities that Verizon has for its enterprise whereas illustrating the extent to which generative AI has freed up buyer help brokers to do different issues.
This isn’t a use case that can apply to each enterprise, a lot of whom will probably want to double down on the extent of service offered. Both means, although, it demonstrates that with automation taking good care of extra time-consuming legwork, human know-how might be directed to the areas the place it’s most precious.
Essentially the most intriguing features of Verizon’s use of generative AI are the predictive parts, with AI evaluation capable of determine the character and kind of a buyer inquiry forward of time. Whereas that is probably not foolproof, when it really works, it absolutely helps buyer help brokers to reply proactively and never simply reactively – whereas saving time that may in any other case be spent establishing context.
ING: Conversational AI ups buyer satisfaction and loyalty whereas sustaining belief
“First nail it, then scale it” is the way in which that Ayush Mittal, IT Chapter Lead at Dutch multinational financial institution ING describes the corporate’s strategy to generative AI.
ING used a generative AI chat answer to sort out a few of its key issues with customer support, reminiscent of excessive quantity – ING receives queries from 85,000 clients per week – and the necessity to supply help to clients outdoors of conventional working hours.
The corporate isn’t any stranger to automated customer support, having experimented with using chatbots since 2017; nevertheless, these early chatbot experiences have been inflexible and solely allowed clients to pick out from a variety of pre-defined choices. Utilizing massive language fashions and pure language understanding (NLU), ING might create a extra fluid dialog: as Mittal places it, “Digital brokers have the potential to reply any query and supply a extra pure conversational stream.”
Working with QuantumBlack, the AI arm of McKinsey, ING piloted a brand new GenAI assistant with 10% of consumers within the Netherlands who have been utilizing the cellular app’s help chat perform.
Inside the first seven weeks of use, the chat agent had served 20% extra clients than the financial institution would sometimes serve with buyer help, offering a blueprint that ING might progressively scale throughout ten markets. Mittal said that conversational AI additionally allows personalised, environment friendly interactions and has improved buyer satisfaction and loyalty.
Generative AI is after all not with out its shortcomings, reminiscent of a propensity to confidently present incorrect info – one thing that could possibly be extremely harmful for a monetary establishment and can be disastrous for buyer belief. As such, ING has carried out strict guardrails: ING threat stakeholders have been introduced in from the start of the method, real-time monitoring and auditing are in place, and a low-confidence AI response will set off human intervention.
ING’s conversational agent additionally can’t give recommendation on particular matters, reminiscent of mortgages and funding merchandise. As Bahadir Yilmaz, Chief Analytics Officer at ING, advised Fintech Finance, “Introducing generative AI strategies to a enterprise drawback is simply 5 % of the job.
“Ninety-five % of the job begins after that. It is very important construct techniques round AI instruments and that takes loads of effort.”
Takeaway:
Many an government has lamented the necessity for putting strict compliance guardrails round using generative AI, fearing that it’s going to hamper innovation. If something, nevertheless, ING has leaned into these guardrails, making buyer belief in its GenAI implementation the utmost precedence.
This additionally gives peace of thoughts for firm executives, who know that they’ve already achieved due diligence and don’t have to fret about GenAI introducing threat parts or making a sudden reputational disaster.
As McKinsey senior companion Andrea Del Miglio mentioned, “There’s an enormous problem for the banking trade to enhance the shopper expertise, however there’s additionally an enormous quantity of threat. You can not simply take expertise popping out of a field to do that.
“Leaders should ask themselves: What’s the worth you’re including to the expertise? What’s the edge you may add … to make it extra useful for purchasers and higher meet their wants?”
United Airways: GenAI-enhanced flight tales enhance buyer satisfaction
United Airways’ “Each Flight Has a Story” initiative makes use of SMS messaging and app notifications to supply clients with detailed reasoning for flight delays, resulting in an improved expertise on the a part of the ready clients.
The airline started this initiative a number of years in the past, however has lately launched generative AI into the method of making the messages, enabling higher scale and liberating up employees to unravel more difficult points in lieu of modifying templates.
On account of introducing GenAI into “Each Flight Has a Story”, buyer satisfaction has risen by 6%, as United Airways CIO Jason Birnbaum advised CIO.com.
United Airways’ human “storytellers” nonetheless look over the messages to make sure that they’re proper for the model, however Birnbaum mentioned that United is turning into rather more snug with utilizing generative AI.
“We labored arduous to fine-tune this mannequin to take operational feeds, notes from our operations groups, the crew, and all these completely different sources of information,” he mentioned, “and have AI take all this information and create a story that’s extra clear, empathetic, decisive, and clear as we might be.”
The flight delay notifications would possibly embody particulars of an incoming plane arriving late on account of runway building, or an early heads up about crowded safety on account of an NBA All-Star sport, full with recommendation for travellers to reach early and use the United Airways app to streamline their expertise.
Birnbaum advised CIO.com that United has additionally developed LLMs to be used in procurement and manager-employee communications and is beta testing using GenAI to enhance operations summaries for shift handovers.
The airline’s constructing and coaching of generative AI fashions is facilitated by United Knowledge Hub, a centralised information lake that consolidates United’s information sources and allows real-time entry to information.
Takeaway:
United Airways’ use case right here reveals how generative AI, when carried out in the suitable means, can improve and scale up an initiative that’s already efficient.
“Each Flight Has a Story” wasn’t conceived for generative AI – however it’s being extensively expanded on account of its introduction. In February 2024, Inc. reported that United was solely offering the extra detailed backstory for 15% of flights, however that it hoped to scale this as much as 50% because of the introduction of GenAI.
United Airways started the initiative as a means so as to add humanity and nuance to delayed flight notices, and the addition of GenAI hasn’t decreased that humanity – thanks partly to the continued oversight of human ‘storytellers’.
The case examine from United additionally reveals how work by firms to get their information ‘home’ so as pays dividends with regards to implementing generative AI. The work to create United Knowledge Hub has offered a strong information basis for United’s introduction of each exterior and internal-facing LLMs, which are actually getting used for flight standing updates, procurement, operations, and extra.
