Here’s How ChatGPT Can Either Enhance Customer Service Or Brutally Undermine The Customer Experience, Warns AI Ethics And AI Law
Customer service is supposed to be a vital part of delighting and retaining your customers.
I have a question for you on this.
When you call to get customer service or do so via an online chat, are you interacting with a live person or is it an automated AI interactive conversational system?
Sometimes it is hard to tell which is which.
You’ve undoubtedly had live agents that seemed as though they were acting like robots. Their voice is monotone, or their online chatting is dry and terse. When you ask for flexibility to accommodate a special request, they might deny your request in the flattest of language. That’s not how we do things, replies the agent. Tough luck to you.
An AI app that interacts with you might take the same stance. You desperately attempt to provide your special request and the AI summarily turns you down. If you realize that it is an AI system and not a live agent, you likely wonder whether a live agent might be more accommodating.
The funny thing is that in today’s topsy-turvy world, you can at times have a live agent that is inflexible while an AI-power automated agent is more accommodating and willing to bend the rules for you. That doesn’t comport with our expectations of how humans work versus how AI automaton works. When this arises, the surprise can take you aback.
Nonetheless, it can and does happen.
As an illustration that dealing with a live human agent is not necessarily the best of avenues, you might have had the sour instance of dealing with a customer service representative that seems rude and overbearing. That designated agent or person that you are doing an online chat with is supposed to represent their company and be welcoming. But maybe they are experiencing a bad day or perhaps they’ve been working long hours, and they take it all out on you. They might not be aware that they are allowing their personal frustrations to undermine the customer service experience. In some instances, they might realize it, though they don’t care and figure that if the company doesn’t catch them, they can keep being irritating as long as they wish.
In contrast, an AI-powered automated agent would normally not land in the same boat. Whether it works a few hours or 24×7, the AI isn’t going to become agitated, tired, hungry, or otherwise devolve into an emotional tantrum. An AI automated agent is usually polite in wording. The AI is typically consistent and will remain seemingly cool and calm. Of course, there can be irritations about using an AI automated agent, depending upon the level of capabilities built into the AI.
The gist is that we would be remiss to reject out of hand that an AI-powered automated agent is not up to dealing with customers. In some ways, an AI app might do as good a job as a human agent, possibly even better. There isn’t any nonsense that comes up. Right to the point. The AI customarily stays on an even keel. You can do your transaction or make your inquiry, potentially get the customer service response you wanted, and then move on with the rest of your day. There isn’t an afterlife of having been belittled by a human agent or dealt with a human customer service rep that seemed to relish ruining your day.
It used to be that customer service-focused AI apps were narrowly programmed and would be frustratingly bounded in terms of providing customer service.
If your need or request wasn’t within a handful of pre-determined actions, you were going to find yourself at a dead-end. An exasperating ordeal would take place as you go back and forth with the AI. The AI wasn’t able to parse your online chat sufficiently. You found yourself having to type sentences as though communicating to kindergarten-level reading proficiency. That often didn’t move the needle and you were jammed into a cycle of dread as you clamored for a means to escape the AI-agent interaction.
Indeed, partially due to those annoying limitations, it used to be that you could break out of the AI interaction and access a human agent. On some occasions, you had to first exhaust the range and depth of the AI. This meant that you had to undergo an arduous journey to “convince” the AI that the only viable way to cope with your predicament was to shift over to using a human agent.
Nowadays, based on the latest in AI such as generative AI, some firms are using generative AI apps for customer service and provide no other alternative. There aren’t any means to redirect the conversation to a human agent. You must one way or another find an approach with the AI that will accomplish your goal. This might require some clever online chat wording by you. Your vexing aim is to steer the AI toward what you believe will fulfill your request, despite the AI interactively not necessarily grasping what you are trying to have undertaken.
In today’s column, I’ll be going over with you the pros and cons of using generative AI for performing customer service.
The happy face side of things is that the hottest new generative AI can do quite amazing interactions and converse with you on a basis that eerily seems human-like. The sad and realistic fact of the matter is that contemporary generative AI has limitations and challenges that can arise when interacting with customers, possibly displeasing customers and losing customers that insist they will never interact with an AI for customer service ever again (well, realize that such a vow will increasingly be infeasible given the ongoing rush of modern firms pushing toward generative AI usage for customer service encounters).
I will discuss an overarching approach that proffers how to conceive of, design, and implement the use of generative AI for your customer service activities. Those would be the efforts that you are supposed to do and serves as a considered set of best practices. The upbeat viewpoint is that follow the right things to do when establishing and maintaining generative AI for customer service and all will work out splendidly.
In my work now as a consultant and based on my prior roles as a top executive overseeing customer service, I recognize that best practices often do not tell the whole story. Knowing what the worst practices are can be a lifesaver and serve as a flashing alert regarding what you need to flagrantly avoid doing.
You can easily fall off the wagon of the best practices and not realize that you are nearing those career-ending worst practices. Ergo, I will provide a semblance of worst practices to let you know where the abyss is and how deep it can ill-advisedly go.
Before we get into the customer service particulars, let’s first make sure that we are on the same wavelength about the nature of current times generative AI.
Background About Generative AI
Generative AI is considered a subtype of AI overall. You have undoubtedly heard of or made use of generative AI. OpenAI’s generative AI app ChatGPT and its successor GPT-4 are pretty much part of our societal lexicon these days. ChatGPT is a text-to-text or text-to-essay form of generative AI. You enter a text prompt, and ChatGPT generates or produces a text response, typically consisting of an essay. This is done on an interactive conversational basis using Natural Language Processing (NLP), akin to Siri or Alexa though in writing and generally with much greater fluency.
I’m betting that you are likely aware that ChatGPT was released in November of last year and has taken the world by storm. People have flocked to using ChatGPT. Headlines proclaim that ChatGPT and generative AI are the hottest types of AI. The hype has been overwhelming at times.
Please know though that this AI and indeed no other AI is currently sentient. Generative AI is based on a complex computational algorithm that has been data trained on text from the Internet and admittedly can do some quite impressive pattern-matching to be able to perform a mathematical mimicry of human wording and natural language. Do not anthropomorphize AI.
To know more about how ChatGPT works, see my explanation at the link here. If you are interested in the successor to ChatGPT, coined GPT-4, see the discussion at the link here.
There are four primary modes of being able to access or utilize ChatGPT:
1) Directly. Direct use of ChatGPT by logging in and using the AI app on the web 2) Indirectly. Indirect use of kind-of ChatGPT (actually, GPT-4) as embedded in Microsoft Bing search engine 3) App-to-ChatGPT. Use of some other application that connects to ChatGPT via the API (application programming interface) 4) ChatGPT-to-App. Now the latest or newest added use entails accessing other applications from within ChatGPT via plugins
The capability of being able to develop your own app and connect it to ChatGPT is quite significant. On top of that capability comes the addition of being able to craft plugins for ChatGPT. The use of plugins means that when people are using ChatGPT, they can potentially invoke your app easily and seamlessly.
I and others are saying that this will give rise to ChatGPT as a platform.
There are numerous concerns about generative AI.
One crucial downside is that the essays produced by a generative-based AI app can have various falsehoods embedded, including manifestly untrue facts, facts that are misleadingly portrayed, and apparent facts that are entirely fabricated. Those fabricated aspects are often referred to as a form of AI hallucinations, a catchphrase that I disfavor but lamentedly seems to be gaining popular traction anyway (for my detailed explanation about why this is lousy and unsuitable terminology, see my coverage at the link here).
Into all of this comes a slew of AI Ethics and AI Law considerations.
There are ongoing efforts to imbue Ethical AI principles into the development and fielding of AI apps. A growing contingent of concerned and erstwhile AI ethicists are trying to ensure that efforts to devise and adopt AI takes into account a view of doing AI For Good and averting AI For Bad. Likewise, there are proposed new AI laws that are being bandied around as potential solutions to keep AI endeavors from going amok on human rights and the like. For my ongoing and extensive coverage of AI Ethics and AI Law, see the link here and the link here, just to name a few.
The development and promulgation of Ethical AI precepts are being pursued to hopefully prevent society from falling into a myriad of AI-inducing traps. For my coverage of the UN AI Ethics principles as devised and supported by nearly 200 countries via the efforts of UNESCO, see the link here. In a similar vein, new AI laws are being explored to try and keep AI on an even keel. One of the latest takes consists of a set of proposed AI Bill of Rights that the U.S. White House recently released to identify human rights in an age of AI, see the link here. It takes a village to keep AI and AI developers on a rightful path and deter the purposeful or accidental underhanded efforts that might undercut society.
I’ll be interweaving AI Ethics and AI Law related considerations into this discussion.
Customer Service And The Use Of Generative AI
The good news is that generative AI provides a step up in performing a customer service interactive dialogue with humans.
Score a point for advances in AI.
Generative AI is so good that you can readily be lulled into thinking that the generative AI is an actual human. This takes us to the bad news. The bad news is that since generative AI seems so convincingly to be human-like, you can mindlessly make the mistake of believing what the AI says and get yourself into hot water.
This is an important tip for companies that are jumping onto the generative AI bandwagon for using AI as a customer service tool. Keep your wits about you and realize that generative AI does have rough edges. Do not undersell generative AI, and do not oversell generative AI.
In one sense, the older versions of AI that were programmatically restricted are a lot less of an ordeal to put in place. You just established the programming guidelines and rules, along with stated guardrails, and you pretty much knew that the AI would stay within prescribed bounds. Regrettably, this also usually led to a stilted conversation with customers by the AI and an overly constricted range of uses that often necessitated a human customer service rep be ready in the wings to take over the customer interaction.
For contemporary generative AI, trying to confine the AI to particular boundaries is not necessarily a piece of cake. Speaking of which, you likely want to have your cake and eat the icing too. You want the wonderous interactive dialogue capacities of generative AI, but you don’t want it to veer into the ugly territory of emitting falsehoods, errors, biases, and those unnerving AI hallucinations. Aiming to be in that Goldilocks realm of the porridge not being too cold or too hot will take careful planning and astute execution.
Let’s start with the smiley face of laying out best practices for adopting generative AI as honed for customer service activities.
A recent online article in Harvard Business Review (HBR) entitled “Create Winning Customer Experiences with Generative AI” by Nicolaj Siggelkow and Christian Terwiesch provides a handy indication of key best practices on applying generative AI for customer service (posted on April 4, 2023). They neatly package the strategy into a set of three R’s along with a fourth R, as I explain next.
The authors indicate that there are three primary stages or steps to be taken, consisting of a first step to recognize that a customer need might be aided via generative AI, followed by a second step entailing translating the customer needs into targeted requests for generative AI, and a third step that involves the generative AI responding to the customer about their request. A fourth stage or step provides a repeating activity to ensure that feedback is taking place and the generative AI can be adjusted to further align with the customer service tasks at hand.
I provide this quick recap of those recommended stages or steps:
1) Recognize. Recognition of a customer need that can be met via generative AI. 2) Request. Translation of customer needs into targeted requests that can be fulfilled by generative AI. 3) Respond. Respond back to a customer as to their needs via use of the generative AI. 4) Repeat. Establish a positive feedback loop such that the generative AI can do an even better job on an ongoing basis to meet customer needs.
The authors describe the matter this way:
“The launch of ChatGPT will be remembered in business history as a milestone in which artificial intelligence moved from many narrow applications to a more universal tool that can be applied in very different ways. But technology in and of itself does not create value. For value creation to happen, we have to think about large language models as a solution to an unmet need, which requires a precise understanding of the pain points in customer experiences. As we discussed, these models can help address pain points that occur along a journey spanning the phases of recognize, request, and respond” (ibid).
I trust that you noticed the emphasis on making sure to understand the pain points confronting your customers as they seek out your customer services.
Tossing generative AI into a customer service process merely because it seems the faddish thing to do is one of the worst practices that I will later on herein be covering (my Worst Practice #1). You need to go back to first principles. Make sure to do an analysis of the pain points that your customers encounter today in making use of your customer service. Once you’ve got that laid out, you are then ready to see how generative AI might remove, mitigate, or otherwise aid in overcoming those pain points. Failing to do such an analysis is nothing more than a shot in the dark of where generative AI is going to pay off.
The odds are that you will shoot your foot by taking a shot in the dark.
The authors provide these examples of how generative AI can be vital to the customer service process:
“Given the ability of large language models to interpret texts and integrate data, these models could become great assistants. For instance, a user could permit such an assistant to continuously read information such as health records, Fitbit data, and legal paperwork. The AI system could then create prompts for the user that possible needs are lurking, be they in the form of the need for a health screening visit, or the need for a more comprehensive insurance coverage. Note that such customer experiences can be initiated by the chatbot and thereby can overcome the forces of inertia and myopia that hold back the user in many parts of life” (ibid).
A significant insight is mentioned in that excerpt, and I want to make sure that the point gets due attention.
The usual mode of having AI interact with a customer consists of the customer taking the lead on the interaction and driving the interaction. For example, the AI starts with a question such as what your service request is and then awaits a reply from the customer. A ping pong match ensues of each party batting the ball across the net, usually consisting of the AI responding to the customer and not especially acting in any proactive manner.
That is the proverbial reactive variant of generative AI.
But you need to change your mindset about how generative AI works today. Rather than being solely a reactive respondent, the latest in generative AI can potentially be a proactive agent. This means that the generative AI grabs the bull by the horn and seeks to drive the interaction toward some identified goal that is computationally assessed as benefiting the customer. The customer might not be aware of something good that they could have or should have gotten from the company, for which the generative AI brings that realization to fruition.
A customer might be unaware of their options. A customer might be timid about asking for things. A customer might be preoccupied with one matter and not realize that other beneficial aspects could come their way, in addition to resolving or aiding whatever the customer already had in mind. On and on this goes.
Conventional reactive generative AI or the older simpler versions of AI often did not rise to the occasion, or if they did so, it was brittle and tended to fall apart. Modern generative AI can be more resilient, act proactively, and potentially wow your customers far beyond their likely lowered expectations of what customer service is supposed to do for them.
During the initial exploration into how generative AI will enhance your customer service, make sure to be open-minded and not simply attempt to replace any older AI or non-AI automation with the newer generative AI on a blind basis. It is easy to merely unplug the prior automation and plug in the new AI, doing so without any careful inspection of what the newer features open the door to improving as part of the customer experience.
It is like the old line about paving the cow paths. Here’s the deal. In some historic cities such as Boston, cows used to be brought into town and various dirt paths came to be formed. Later on, those dirt paths were paved and thus the oddball curvy roadways exist today that we drive our cars on. Before you put generative AI into your customer service realm, make sure you aren’t going to naively pave the cow paths. Rethink how you are doing your customer service.
As you do that rethinking, gauge what generative AI can do. Blend your re-engineered customer service with the right features and functions of the latest in generative AI. The reengineering is partially driven by what can be done with generative AI. You can take on customer service activities that previously would have been prohibitive to do via conventional automation or even older versions of AI.
The authors of the HBR article also call upon you to reimagine the digital customer experience, noting that you should keep at top of mind the advantages of generative AI and yet also be cognizant of the drawbacks or shortcomings too:
“Faced with these new technological possibilities, we see executives wrestling with the question of how to take advantage of this new technology and reimagine the digital customer experience. Clearly, ChatGPT and Bard still have many shortcomings (e.g., hallucinations, biases, and non-transparency), but the technology is improving rapidly and is showing great promise” (ibid).
I’ve purposely brought up this bit of notification about shortcomings of generative AI so that I can seamlessly segway into my promised elucidation on not just best practices but also the must-avoid worst practices.
Avoiding The Worst Practices Of Generative AI For Customer Service
Now that we’ve surfaced the best practices, I’d like to provide you with my selected list of the worst practices.
I hope you have the stomach for this.
Anyone that has ever had responsibility for customer service knows how agonizing it can be when your approach goes awry. The pain of your customer pain points becomes your pain too. Any of the following worst practices will almost surely undermine your desire to improve customer service by including generative AI. The chances are that if you abide by any of the worst practices, the generative AI will be a net negative to the customer experience.
Note that the generative AI per se is not at fault for these worst practices.
I stridently say this to clarify as stated earlier that generative AI is not sentient. You cannot reasonably later on try to shift blame to the generative AI if things go awry. You can certainly try to blame the AI, which people often use as a convenient scapegoat, and others admittedly at times buy into it, but the reality is that the humans that put the generative AI into action are the responsible parties. Until or if we someday decide that AI has legal personhood, see my coverage at the link here, the AI is off the hook and the humans underlying the adoption of the AI are still on the hook.
A smarmy skeptic might suggest that I am saying this to protect myself for the day that AI becomes our overlord and I want the AI to look upon me favorably. If you are interested in those existential risks being bandied around about AI that enslaves us or wipes out humankind, you might want to take a look at my analysis of such life-or-death matters at the link here.
Let’s return to the aim of examining the worst practices of applying generative AI to customer service.
Please prepare yourself.
Here is my vaunted Top Ten list of the worst practices for putting in place generative AI for customer service and as part of the customer experience:
1) Fad. Generative AI is erratically tossed into the customer experience as part of a fad, doing so without any prudent sense of utility or value for proffering bona fide customer service. 2) Frivolous. Generative AI is shoved into place but is little more than marginally frivolous in enhancing the customer experience. 3) Frustrate. Generative AI inadvertently undercuts the customer experience and produces exasperating frustration when interacting with customers. 4) Fabricate. Generative AI emits so-called AI hallucinatory responses that are totally fabricated and confuse or anger the customer. 5) Fragment. Generative AI is poorly implemented on a fragmented basis and does not tie in with a comprehensive view of the customer. 6) Flimsy. Generative AI quickly reaches its boundaries, and the flimsiness becomes readily apparent to customers, disturbingly so. 7) Feisty. Generative AI shifts into a feisty mode that browbeats the customer and leaves the customer shocked and dismayed. 8) Flaunt. Generative AI is flaunted or touted as the perfect customer service mechanism and yet does not live up to the hype as portrayed to your customers (thus deflating the propped-up expectations). 9) Faceless. Generative AI is perceived by your customers as a faceless unyielding robotic element that could care less about them, causing an impression that your firm is without compassion or genuine interest in them as a customer. 10) Failure. Generative AI is set up and used in any of the foregoing faltering ways, ultimately being a failure when it comes to benefiting customer service and altogether negatively impacting the customer experience.
An observant indication is that I opted to characterize the worst practices by words that start with the letter F. Of those Top Ten, the last one as listed at position #10 consists of a complete failure that guts or undermines customer service by abiding by one or more of those worst practices.
Do not tempt fate by letting yourself engage in any of those worst practices. Sure, you might survive by only embracing one or two, though you are going to precariously find yourself with a sword dangling over your head. Just say no to the worst practices. Maintain a list of whether your customer service reengineering or reimagining is swinging toward those maladies, and then move heaven and earth to refocus and steer clear of those rocky shores.
If there is reader interest, I’ll gladly cover in a future column the various ways and means to prevent or contain the generative AI from flopping into those worst practices. There are a number of generative AI parameter setting approaches, data training actions, context setting establishment, and other AI design and development techniques that come to bear. Be on the watch for that additional coverage.
Here’s an additional aspect that you might not have considered.
Do not assume that just any AI developer knows what to do about those worst practices (and those aforementioned best practices). An AI specialist might not be familiar with how generative AI should act and respond in customer service contexts. The AI developer is likely more so versed in the AI technological building of general-purpose generative AI, and not in the domain-specific honing of generative AI for customer service interactions.
Not all AI techies are the same and not all AI techies have the right skills and perspective for this kind of generative AI tailoring. Check to make sure you’ve assembled the appropriate team for a customer service revamp, which ought to include well-versed AI developers that not only know generative AI but also comprehend and can adjust generic generative AI to the customer service task. You’ll inevitably thank me for that piece of advice.
I also would like to add that the adoption of generative AI for customer service is not a one-and-done affair. The easy trap is that you upfront shape the generative AI and then believe that it will churn away without any needed maintenance or tuning. I suppose that is one I should append to the worst practices list. You will need to frequently follow-up (let’s make that #11 on the list) and keep the generative AI on-target and up-to-date.
Conclusion
Almost time to wrap up on this topic for now. I do have a few final remarks that I will amplify via selected famous quotes about customer service.
Steve Jobs reportedly said this: “It’s not the customer’s job to know what they want.”
Old ways of handling customer service tended to assume that the customer knew what they wanted and that all the agent had to do was react or respond to the customer’s inquiry. We now know that whether leveraging a human agent or an AI-based automated agent, there is a need to go beyond being merely reactive and also make sure to be proactive (in the right way, at the right time).
The proactive approach can go overboard, so be careful how far this goes.
I’m betting that you might have in the past had interactions with a human agent that kept hammering you with additional suggestions or ideas of what you as a customer could benefit from. Being proactive to the max and beyond is not a good look. That is decidedly not a proclaimed proactive approach that works. Make sure that generative AI doesn’t get overly aggressive or as I stated above become unduly feisty (Worst Practice #7) during those devised proactive interactions.
Next, consider this famous quote by Theodore Roosevelt: “Nobody cares how much you know until they know how much you care.”
I bring up this quote about caring because it has a notable bearing on the use of generative AI for customer service purposes.
A somewhat false assumption that some make about using AI for customer service is that, unlike a human agent, the AI presumably cannot promote a sense of caring that a human agent can. Therefore, you seem to be faced with a rather tough choice. You either use a human agent that can seemingly create a human bond with a customer, or you use AI that allegedly cannot foster a caring rapport and yet will do other customer service tasks well.
Generative AI breaks that mold. See for example my discussion about how generative AI is devised to emit a sense of humility, per the link here. The interactive dialogue presented by a properly tuned generative AI can in fact create a rapport with customers. Furthermore, when suitably shaped, the AI will avoid mistakes or infringement of a caring aura that a human agent might bring forth during a customer interaction. The AI-established rapport can be one of a sustained and consistent nature.
That is something to make a specific note of.
A hearty discussion about customer service almost invariably has to invoke a quotation by Henry Ford. The venerated Henry Ford was said to have uttered this memorable assertion: “A business absolutely devoted to service will have only one worry about profits. They will be embarrassingly large.”
I am sure that his quote is music to the ears of those that are tasked with providing customer service. It is a mantra that is worth repeating. One wishes that all top executives also embraced the sentiment, which they might in their hearts, but they at times won’t showcase the same ardent support when it comes to their wallets and budgeting for customer service.
The cost to adopt generative AI is something that will also need to be part of your equation when opting to include AI in your customer experience. If you look at cost alone, the path ahead might seem daunting. The proper business perspective consists of calculating a cost-benefit analysis that will indicate how the costs weigh coinciding with the benefits. Not all firms will come up with the same result in deciding whether to use generative AI or not.
A rule of thumb in business is that if you don’t compete when it comes to customer service, your customers might very well flock to your competitors. A segment of society is already lining up to use generative AI for customer service. You might liken this to the desire to use ATMs over dealing with human banking clerks.
Carefully examine how your customer service is operating today. Reimagine what it could be. Do so by actively exploring how generative AI can add value to the customer experience. Meld generative AI into your customer service activities if the cost-benefit makes sense to do so.
We might relish another Henry Ford quote to finish off this endearing discussion: “Most people spend more time and energy going around problems than trying to solve them.” Solving your customer service problems and dealing with those customer pain points might be aided via the judicious and mindful use of generative AI.
You can ask generative AI such as ChatGPT about this, and the presented answer is usually along these same aforementioned lines, though perhaps it is a bit self-serving by ChatGPT to toot its own horn.
That’s not being feisty, just being practical.
