Technology doesn’t go backwards

Historical events aren’t realized when you’re living them. People who were in Star Wars probably didn’t know how big of an impact it would have a few decades later. Same with the Industrial Revolution. When we look back at it now, it was a pivotal point for all of humanity and things have never been the same since. The AI revolution is what we’re living through right now and perhaps posterity will look back to this era as a turning point for them.

The Generative AI movement has tremendously affected how companies build their products. Every SaaS or B2B company we know of has adopted AI into their product in one way or another and that’s saying something. Given how easily we’ve accepted AI into our lives, it won’t be long before it becomes a trove of convenience; going from being a ‘good-to-have’ to a ‘necessity’. And customer support is becoming one of the earliest areas that will witness a fundamental change with AI at the center of it.

Latest data highlights a significant shift in customer service dynamics, with AI-powered agent support and customer service technologies making significant strides. Research indicates that integrating AI solutions can elevate agent efficiency by as much as 40% and slash customer service expenses by up to 30%. Additionally, AI-driven chatbots and self-service portals are now managing up to 80% of standard customer inquiries, allowing agents to dedicate their time to more intricate issues.

What we’ve learned from our customers

AI in customer support has changed the way support leaders think. In a way, that’s necessary. AI isn’t just about a product change but a mindset change above everything else. If your reason for adopting an AI-first approach to customer service is to do what everyone else is doing then you’re probably on the wrong track. You need to inherently understand how AI can change the way you work and accept that as a way of life. That’s the easiest way to transition and make decisions not clouded by constant skepticism.

“Everyone these days is caught up with the ‘Gen AI’ keyword and that, of course, is here to stay. I believe it’s a force multiplier and Threado AI has been that for us,” said Prasun Choudhury, Senior Director Solutions Engineering at MoEngage, in our conversation about what problems they faced as a team before they moved to an AI solution. Prasun talked to us about how a company of that scale requires an extensive knowledge base which, in their case, is spread across Confluence docs, Slack channels, and internal playbooks. Not being able to track a single source of truth creates certain delays in the system which keeps them from being able to scale support operations quickly. When a new team member is onboarded, it takes time to train and get them up-to-speed on how things work, which further increases resolution time for the team.

We had a similar conversation with Neil Choudhury, who looks after Business Operations and Customer Service at Juno. Juno being a digital banking and cryptocurrency space has customers across tiers which can be complicated to deal with and having that information always on demand is not feasible. The biggest challenge for the team was searching for information manually in their databases. “After we started maintaining a central resource hub or a glossary for all our information, the primary problem that agents faced was having to go through the entire glossary and figure out the issues, the resolution, and then craft answers for customers,” he said.

One of the most prominent concerns that has come forward through what we’ve learned from our customers is how to scale support operations as the company scales. You have immediate options such as expanding your team or defining more structural hierarchies but that’s not a long-term solution for indefinite business growth and it’s definitely not the most resource-effective option. Team expansion is tempting, especially if you can hire resources for a lower salary. But what happens when you have thousands of tickets being raised every day? You can’t possibly think of having a team of 500 agents.

You can assign more tickets per agent but the pressure of unrealistic targets is a surefire way of losing your team completely. Customer Support is a department infamous for higher attrition and employee burnout. Most departments across industries reportedly have an average employee turnover of about 15% whereas customer service departments report an average turnover of 30-45% which is atrociously high. This is why it’s important to prioritize your agents and not force them to the edge of burnout.

In our conversation with Savian Boroancă, who is the Head of Community at Sessions, we learned that Sessions’ user base has almost tripled in the last 6 months which has also led to an increase in tickets. And then there are unprecedented scenarios where it becomes extremely difficult to account for an increase in customer concerns in a shorter period of time. “There are times when we have very high traffic like when we got listed on AppSumo the first time, there were about 200-300 tickets for each person in a day,” he said.

The million-dollar question: How do you scale support as the company scales?

Some of the biggest concerns that came forward through our conversations with customers are -

• Support agents have to spend valuable time searching, understanding, and curating answers from a scattered knowledge base which is counter-productive.
• How do you sustainably scale support as the company grows?
• A lot of time goes into hiring and training new agents who need constant help early on in the team.
• For larger enterprises, only members who’ve spent 4-5 years in the company have very specific knowledge but most agents have to rely on knowledge bases and team communication to resolve queries.
• Larger ticket volume (and repeat ticket volume) leads to longer wait times.
• Constant context switching for agents, increased vulnerability to the accuracy of the answers, and reliability concerns leading to unsatisfied customers.

Juno’s support team is about 12-16 members at the moment. Per month they get around 6,000 - 9,000 tickets through emails and chats which get equally divided between all team members. Each member closes about 35 tickets per day. Considering that each team member needs up to 5 minutes to search queries in the knowledge base, that’s almost 3 hours per agent spent every day simply searching for answers.

Savian from Sessions mentioned that on days when they witness higher traffic, each agent sees up to 300 tickets per day and on most days they have to resolve 80-100 tickets. Having to resolve up to 3x more tickets is not an easy feat to achieve but it was necessary for them nonetheless. Since Sessions’ team started using AI as a solution, things got under control but it wouldn’t have been the same otherwise. Scaling the team, especially as a startup is not optimal. It comes down to accepting technology and conforming to newer ways that are sustainable and progressive.

AI in Customer Support: The way forward

‍

The proposition of AI as a whole in customer support has evolved tremendously in the last year alone. In the wake of the AI revolution, customer support leaders realized that there’s more to AI than just chatbots. This brings forward the idea that agents can benefit more from AI than customers. If the endgame is better customer support, why not enable your agents to offer efficient support that doesn’t just improve customer-related metrics but also boosts team productivity?

This is what we learned from our conversations with support leaders and our customers. For MoEngage, Threado AI became the single source of truth by connecting their entire database. They wanted an authoritative source of answers, especially for the ones who were new to the team. “It’s very difficult for one person to remember everything about the product, so having something that is constantly reading into answers from other conversations or knowledge base is ideal. This is one of the most powerful things Threado AI can do and what we’ve had success with. We’ve also started looking into use cases for engineering, consulting, and pre-sales side of things,” said Prasun.

We saw this recurring pattern with every conversation we had that one of the most convenient things an AI solution helps with is the ability to connect the entire database and answer queries wherever convenient for the team. Engineers or agents only have to ask the question and AI answers it directly, completely eradicating the need for agents to look for answers themselves.

Neil from Juno said, “With Threado AI, we can just ask the question without having to manually go through our entire knowledge base. Instead of having our team scour through each account and their information, all of it is just one question away on Threado AI. Icing on the cake is that the tool also gives us the source of the information provided which makes Threado AI behave like our very own search engine!”

What’s more, is that Threado AI helps will the links of where the information has been sourced from which the agents can use to quickly verify the accuracy of the the answers provided. Speaking of accuracy, in every case we saw that our customers have been extremely satisfied with the answers Threado AI provides. AI for Customer Support is not only a solution but a medium to identify gaps and upgrade your knowledge base. And AI is only as smart at the knowledge base so it functions intricately and interdependently where both of these things improve one another.

“The first thing that we saw with Threado is, if we had proper underlying documentation, it was almost 99% accurate. In fact, as we went deeper into the implementation like our Confluence database, internal documentation, and public KPI documentation, Threado AI was capable of putting very nice answers together for all questions.” - Prasun Choudhury.

Neil mentioned that the bot has an accuracy of 85-90% and in the time that Juno’s support team has used Threado AI, its accuracy has only gotten better which has resulted in the team being more reliant on the bot for query resolutions. The team has asked a little over 1200 queries to the bot since they started using it and about 1050 of which have been answered by the bot, accounting for about 87% accuracy.

We mentioned that Juno’s team loses up to 3 hours per day per agent in simply searching for answers. But Threado AI has been able to resolve queries in under 15 seconds. Up to 5 minutes saved on every query for 35 queries solved per day amounts to 3 hours saved every day from manual redundancies. Considering this for their team of 16 agents, they save almost 48 hours worth of work every day.

One of the biggest concerns of scaling support operations with business growth is also accounted for with an AI solution in your stack. In our conversation with Sessions, Savian mentioned, “Because Threado AI helps resolve 50-100 queries every day, the ticket volume per agent remains tremendously under control, and despite an increase in product users, we haven’t seen an increase in support tickets per agent.” Without Threado AI, the team would’ve had to handle more tickets leading to stress and inefficiency.

AI has evidently affected KPIs as well. Making the team more efficient by helping them save several hours of work directly translates to better team metrics. “Time optimization is an exceptionally important factor for every company and this has also helped us improve our Average Handling Time for tickets,” explained Neil from Juno. “We have also seen improvements in terms of the First Response Time and First Contact Resolution rates that have significantly increased, helping us a lot in terms of uplifting the team’s productivity.”

“Having Threado AI on your side as the authoritative way of getting instant and accurate answers, has shortened the margin with which we solve queries. I can safely say that it has had a tremendous impact on the First Response Time (FRT) and an impact on the overall customer experience because it helps with the Resolution Time,” said Prasun from MoEngage in talking about how Threado AI has shown improvements in their KPIs.

What does this mean for the future of customer support?

It almost feels like an “adapt or be left behind” moment in customer support now. Up to a certain point, this was up for debate - is AI really worth investing in? But now it’s no longer a question of “why” but rather “when”. You can’t outrun it and its ubiquitous presence will exponentially become more undeniable with time. Customer expectations are an added reason to adopt AI as soon as possible because your competitors are already ahead of the curve. Reports have shown that 45% of customer support teams have already adopted AI into their operations. Won’t be long before this number hits the ceiling.

From our conversations, it’s become clear that some of the most pressing concerns that support teams have can be easily solved with AI. They’ve accepted it, have seen immediate changes with it, and are focusing on how can they conform and become AI-first as conveniently as possible. That’s it, that’s the only way forward.