Navigating the AI Revolution in Customer Support

Do not index

Benjamin Franklin’s quip “Nothing is certain except death and taxes” requires modification. But then again, it withstood 234 years of progress. In just 2 years, Large Language Models (LLMs) have demonstrated that they are capable of performing certain tasks almost as well as humans. And so, the new adage should be “Nothing is certain except death, taxes, and the fact that AI will impact everything.”

Customer Support is going to evolve more rapidly than it ever has. And if your job has anything to do with Customer Support, you will want to wrap your head around how AI will impact the industry.

💡

30-second Summary

Advancements in AI have enabled auto-routing, auto-tagging, and conversational chatbots at near-human levels of performance with significantly faster response times

Humans continue to win on empathy, large context windows, experience, and adaptability, but improvements in AI are extremely fast and may close this gap soon

CX leaders need to augment their team with AI; copilots are the way forward

Customer Support is no longer only about ticket resolution; the scope has expanded to providing data for other functions like product, engineering, growth, etc.

2024 is the year of accelerated adoption; companies that move fast will win

Understanding AI and Its Capabilities

Comparing AI (LLMs) and Humans

This tweet sums up how people felt about chatbots until large language models (LLMs) happened. It was incredibly challenging to build chatbots that solved customer problems while delivering exceptional experiences.

Along came GPT 3.5, and then ChatGPT, to make chatbots relevant again.

Chatbots are just one of the many things that generative AI has helped improve. More on that in a bit.

To decode AI’s impact on customer support, let’s first understand its traits as compared with humans. We can then present what it can and can’t do as of today. Later in the article, we establish where one should use AI to augment human agents.

🏆 = 1 point

☹️ = -1 point

Capability / Skill	LLMs (GPT-4)	Humans	Notes
Availability and Scalability	🏆	ㅤ	ㅤ
Speed	🏆	ㅤ	ㅤ
Vastness of Knowledge Base	🏆	ㅤ	ㅤ
Language Proficiency	🏆	🏆	Humans are more eloquent and generally write more ‘human’ (as odd as that may sound). LLMs are more versatile with writing style and knowledge of languages.
Computational Processing	🏆	ㅤ	ㅤ
Creativity	🏆	🏆	Creativity could be looked at as one type of output of information processing. Humans do well at assimilating a wide variety of information although generative pre-trained models are catching up. On a certain level and for certain tasks, AI models are better than most humans. I draw only slightly better than the average 5th grader. Dall•E 3, Midjourney, and Stable Diffusion can replicate the styles of the world’s finest artists and demonstrate creativity. It can also draw like the average 5th grader.
Consistency	🏆	🏆	Consistency is nuanced. LLMs and Humans exhibit consistency and inconsistency in varying ways - a topic for another day.
Ability to Learn	🏆	🏆	Again, this isn’t an apples and apples comparison.
Accuracy	ㅤ	ㅤ	ㅤ
Emotional Intelligence	ㅤ	🏆	ㅤ
Personalization	🏆	🏆	ㅤ
Adaptability	ㅤ	🏆	ㅤ
Experience and Intuition	ㅤ	🏆	ㅤ
Real-Time Judgement	ㅤ	🏆	ㅤ
Cultural Understanding	🏆	🏆	I don’t know much about culture in, say Argentina or Nigeria. An LLM knows far more than me. But I’m more deeply aware of the cultures that I live in.
Bias	☹️	☹️	ㅤ

AI in Customer Support

We’ve written a comprehensive piece on how AI integrates into customer support, but it’s worth summarizing for context.

Like most computational systems, LLMs are superior to humans at processing lots of data with high accuracy, consistency, and speed. They have high availability and are massively scalable.

However, LLMs and generative AI are vastly superior to previous AI models in their ability to process and output natural language.

This combination makes it ideal for handling certain language tasks such as:

auto-routing of tickets

auto-tagging based on intent, sentiment, etc.

recommending support articles

suggesting responses for queries

synthesizing and summarizing historical conversations

providing insights from the user’s product usage

providing insights from the customer requests

real-time translation

In addition, generative AI is also powerful enough to enable conversations directly with knowledge bases and deliver product insights.

While there has been significant progress, the challenge for all generative AI-powered products is dealing with inaccurate or nonsensical responses that are characteristic of LLMs. Hallucinations, as they are termed, are detrimental to the customer experience. In most cases, wrong information is more harmful than no information.

Hence, defining where AI plays a role in your support workflow requires deep consideration.

Examples of AI in Customer Support

Our customer data indicates that 30% of support requests are knowledge-based, meaning they don’t require fixing an issue or bug, building a new feature, or taking a manual admin action.

While these requests can be resolved relatively quickly, at 30% of your volume there’s a significant drain on time and emotional energy. That’s where the power of LLMs shines. After ingesting your help center or knowledge base, it is capable of suggesting responses to agents. Consider it a copilot for agents, drafting responses that are often good enough to use without many tweaks.

Ingesting a knowledge base also allows generative AI models to recommend relevant articles from your help center. The process through which all of this happens is termed as “Retrieval Augmented Generation” or RAG as has become popular among AI practitioners.

A significant portion of our customer base has adopted our Copilot for Agents and we continue to develop it further.

The Impact of AI on Customer Support Teams

Reshaping the Role of Customer Support

AI will change the way you work and reorganize your team:

Scaling will become a lot easier with AI doing the heavy lifting

Customer support agents will graduate from handling routine inquiries to managing more complex and nuanced tasks

Agents will engage in more consultative and problem-solving roles

The message is clear. Integration of AI in customer support is seen as a competitive advantage, as it allows teams to focus on areas that require human intervention and expertise. Ignore it at your peril.

Balance Between AI and Human Touch

Humans buy from humans and expect to receive support from humans.

Print this out and frame it. It should be your guiding principle that the human touch shouldn’t be replaced. AI’s cold and emotionless efficiency is best deployed as a supplement for human agents to become 10x versions of themselves.

If you recall our comparison table of LLMs and Humans, it’s very clear that AI and computation in general are ideal for speed, scale, and efficiency. At Atlas, we’re deliberate in how we integrate AI into our offerings. With each feature, we evaluate how every agent can become a 10x Agent while enhancing customer experience.

We believe in Augmentation, not Replacement.

Adaptation in the Time of Automation

There is always a fear of becoming redundant or being replaced when labor and productivity impacting innovation takes place. While the onus is on the individual to adapt, the need of the hour is for managers to become leaders. Because, in the end, people matter. Companies are nothing without their people. If it’s true for OpenAI, it’s true for you.

As a leader, your role is to augment your team and not replace it. Some things to keep in mind:

Train people and help them upskill; create 10x agents

Be human; empathy and kindness trumps efficiency in support

Customer satisfaction remains the North Star

Customer Support Is More Relevant Than Ever

AI hasn’t just elevated support’s ability to deliver better experiences. It has helped customer support to integrate more closely with marketing, tech, and product. The modern customer support tool brings together all customer data in one place for agents to act with richer information and better context.

It’s easy to see how much progress has been made:

Personalization and Context have helped solve problems faster

Predictive support and proactive solutions: the best customer experiences come from having problems solved before they are brought up by customers.

Automating out repetitive work has made employees less susceptible to burnout

And so, what does the future hold?

Predicting AI’s Role in Customer Support

If you’ve followed our path of reasoning, we’re not going to take a crystal ball approach to predicting AI’s role in customer support. Rather, let’s look at what improvements are on the horizon for AI and the implications thereof.

More adept at understanding and processing human emotions: Expect AI to take on more front-line work without a drop in the quality of support

Response time reduces, and so do the costs: AI will be deployed everywhere - in internal workflows, within products, and customer-facing roles

Multi-modal LLMs (input-output text, image, audio, and eventually video): Increase in the complexity of tasks AI can handle, making Agent Copilots much more effective

Going from responses to actions; the ability to take action on behalf of agents: Agents transition to becoming orchestrators and supervisors of AI Agents

💡

In 2024, we’ll see AI transform from a fringe player to part of the core team.

If we had to put it together succinctly, AI is going to be everywhere - as a Copilot for support, as well as the first line of contact for end customers.

How To Prepare For An AI-Integrated Future

Technological revolutions have a profound impact on society and economy. The most common reaction is to resist or ignore. History indicates that embracing breakthroughs is the only way to stay relevant.

Several conversations with folks across companies, geographies, and roles indicate different schools of thought:

This is yet another hype cycle, we’re staying away for now

It’s interesting, but we’re going to wait and watch

It’s promising, we’re dipping our toes in the pond

This is the future, we’re going all in

As I started this last and final section, the first few words that popped into my head were ‘ostrich burying its head in the sand’ and ‘deer in headlights’. It was almost instinctive. When revolutions happen, they are swift. Those who ignore or sit around watching and waiting, end up missing the bus. They get blindsided.

But you won’t. You’ve read this piece. Unlike the ostrich or the deer, you’re going to be well-positioned to act with confidence.

Here’s how we think leaders across support and other functions can cope with rapid adoption:

Staying informed about the latest developments in AI and understanding their potential impact; not just on customer support, but on human productivity in general

Investing in the training and development of customer support teams to work effectively with AI. The era of the 10x agent has arrived. Are they on your team?

Establishing ethical guidelines and privacy policies for AI use in customer support is paramount for building transparency and trust

Developing hypotheses, experimenting with, and assessing the effectiveness of AI-augmented agents; doing so requires the right metrics and tooling

Laying the foundation for integrating AI by kickstarting an exercise of documentation and development of knowledge assets, processes and systems, and other moving parts of the customer’s journey through the product

There is no one-size-fits-all when it comes to adopting AI in your organization. Experimentation is the only way to discover what works best for you.

At Atlas, we want to help SaaS companies deliver superhuman support to their customers. Book a call with us to learn how you could integrate AI into your customer support function.

Atlas gives your agents superhuman abilities by combining a suite of support tools with AI and modern tooling. Agents get a comprehensive view of the entire customer footprint on your product from product activity to session replay.

Forward-thinking companies such as Loops, Near, and Padsplit have been investing in customer satisfaction by moving to Atlas as the choice of tool.

Get in touch if you’d like to know how Atlas can help you build a better support team.

We become what we behold. We shape our tools and then our tools shape us.