The Future of AI-Augmented Support Operations

play Watch

In the latest session of our Future-gazing Series, we spoke with Mimi Erlick, Program Management Manager at Taskrabbit. She shares how it's the people and the processes behind automation that will ensure a bright future for CX automation. 

Welcome to the latest installment of our Future-gazing AI series, which features video Q&A's with experts on the future of AI.

As long as Taskrabbit has been automating their support with Ultimate, Mimi Erlick has been right at the cutting edge of this industry. In this exclusive interview, she shares what she’s learned and offers advice to those looking to up their automation game. 

Mimi also offers predictions for what’s to come, and how you can ensure CX success in this increasingly tech-forward field. And her answers may surprise you. Read the full interview or check out the video recording below.  


Bex Shapiro: Hello, and welcome to Ultimate’s Future gazing AI, which is a series of video interviews where we talk to the best and brightest minds in AI and hear all about their opinions of it – both right now and the future. 

Today, we are very excited to be joined by Mimi from Taskrabbit, which is one of our very valued customers, and we're just going to dive straight in. So Mimi, congratulations are in order because I hear you just got promoted to AI Ops Manager at Taskrabbit. Can you tell us a little bit about what that entails?

Mimi Erlick: Sure, thank you. I lead the Automation and Integration Team, AI Ops for short. My team owns and maintains our AI support chatbot. We partner with Ultimate as well as building other technical support integrations, automations, and solutions – always aiming for the efficiency for our support teams.

BS: Amazing. Let's go back in time and rewind. So you started off at Taskrabbit in self-service, and now you're taking on this really technical, managerial, and strategic role. How did that journey occur in the first place?

ME: My career has always had a focus on customer support efficiency, and when I joined Taskrabbit as a Senior Project Manager, I had a broader mandate to manage operational technology projects. One of the first projects that I spearheaded was bringing a chatbot to Taskrabbit support for the first time. 

I have a background in the more technical end of support, so I took on the chatbot as the lead manager and also as the bot builder. That project has evolved and grown and expanded, as did my scope within it. That led directly to making the transition from leading programs as an individual to managing a team that can take on the responsibility of these support efficiency programs like the chatbot.

BS: What a journey; we love to see it. So in that journey so far, what would you say the biggest challenge has been, and conversely, what about the biggest reward in working with AI at Taskrabbit?

ME: Well, I'd say the reward came quickly and has continued to grow: offering self-service to our customers for the first time. It was evident with how quickly the chatbot had a really measurable impact on contact volume that self-service was something our customers needed. 

Having that exist in an AI platform is really the mechanism that has allowed us to refine and improve our self-service continually and efficiently over the last few years.

"A challenge that my team and I face daily is how truly endless the potential of the technology is. Prioritization of our work is essential when our own time is a limiting factor on all of the solutions possible. A lot of the time, opportunities we find – from shadowing our agents’ feedback to reporting analytics – can lead to really big ideas."

It's a good problem to have but one that leads us to a clear need for organization and communication. As we vet scope all of the ideas, that leads to a backlog of all of these amazing improvements that we're itching to implement. So the challenge really becomes staying on course meeting our goals and making these impactful improvements month over month.

BS: Too little time and a lot of great ideas. I get it. 

Getting tangible here, what are some of the results that you and your team have been able to take credit for in the last two plus years that you've been at Taskrabbit?


ME: Specifically with the implementation of the Ultimate support chatbot, we've seen enormous gains. The support chatbot launched with 100% of chat contacts in 10 language regions on all of our channels, apps, and web in 2023. Our deflection rate averaged 30% globally, and our goal was 20%. The estimated support cost savings from those reduced contacts had a 10x ROI. 

In addition to the chatbot program, I also managed Taskrabbit’s move to asynchronous messaging from live chat. This also led to a contact volume efficiency that had doubled the expected reduction of disconnected and duplicate tickets. I'm also very proud of my team. Our Automation Operations Specialist now manages the chatbot’s maintenance, and that's a first-of-its-kind role for Taskrabbit.

BS: That's really heartwarming, genuinely. We could just end the interview here with those incredible mic-drop moments. Congrats to you and the team.

Tell me a bit more about your perspective on AI and how it's shifted. Since you're obviously at the cutting edge of support automation, how does that make you think about the use cases and the future of AI more generally for leading support teams?

ME: I think being a support chatbot builder brings a really unique perspective to all of these discussions on AI and its future. Support automation has been – and will continue to be – a key pillar for my team. For Taskrabbit, even simplistic support automation can make a difference – by minutes – for how fast we're able to resolve issues for our customers. 

"Making those automations is much more than an engineering or a development project. It takes a keen understanding of the realities of a support process, and that has to come from an agent's perspective in order for the solution to actually stick. So I like to operate with a people-and-process-first approach to creating support automations."

I find it's really the only way to ensure that you won't have to remake the automation a few months later. And you may still need to remake the automation a few months later – it happens for any number of reasons – but I found that it's much easier to do that when the first time around was well built and well utilized. 

As for the future of AI, the biggest thing that I've learned is that it really comes down to training. AI is only as smart as the training and content that you give it, so investing in quality training of the bot and training that includes testing and improving over time is really the only way to get an AI-powered tool that is actually effective. As we continue to implement more new AI solutions, I always look at how well trained the AI is and if that training can be improved with testing over time. 

BS: Thanks so much, I love what you said about this being a people-and-process-first approach. I think that's probably one of many reasons why we've been able to have such a good collaboration with Taskrabbit and Ultimate. I think there's very aligned values behind this kind of approach to AI powered automation.

Let's dive in a little bit deeper. When you reflect on the sort of post-chatbot era that we're in, now over a year since the emergence of generative AI, what misconceptions do you think exist in this era, and how can they be challenged?

ME: Whether it's a misconception or a false understanding, I find that people who have only interacted with chatbots as a consumer see a kind of magic in their ability to respond. And while it is amazing how far AI has come, in the support world responses shouldn't be magic. They need to be accurate, clear, and trustworthy. To make a support bot that does that takes really skilled people building it. 

"It's a misstep in my opinion to think that a chatbot doesn't require real people behind the scenes paying very close attention and giving a lot of care. The management and improvement of a support chatbot takes time, creativity, and consideration. The chatbot needs to provide meaningful resolution, not just a quick answer. A good bot is much more than tech. It needs a team, strong content, and careful design in order to make it successful."

BS: I couldn't agree more. You've just listed all the perfect ingredients.

When you reflect a little bit more on your career career journey, and you think about other aspiring leaders out there looking to make a move into a more technical realm, what advice would you give those people that you'd wish you'd known yourself at the beginning of your journey?

ME: The first thing that comes to mind is to ask a lot of questions. You never know what you don't know if you don't have a conversation first. 

Another piece of advice, which is something that I learned from the more technical projects that I've managed in the last few years is to not get lost in the technical requirements and then forget how change affects people. It can be really hard to take that step back when you're weeks from launch and running into technical issues. But for every change, the human element needs to be considered. 

I've found that this is true in all kinds of technical work with the chatbot. Keeping customer experience center focus takes a lot of meaningful focus, and things that affect the CS teams like ensuring that all technical solutions are implemented with clear training, documentation, and logic really helps avoid pitfalls. 

It can be exhausting to get into the weeds of a technical problem, but always remember to come back up for air. It's easier said than done, but I think that the effort is worth it.

BS: Great advice, thank you so much. I'm sure it will be helpful to lots of people. 

Now getting a little bit more future oriented, what is the one trend in CX that every customer-centric and future-facing organization should be looking to implement?

ME: I'd say moving from live to asynchronous messaging conversations. I led Taskrabbit’s implementation of asynchronous messaging last year. The change was spurred by our understanding of what our customers need from support, and for us a unified, undisrupted thread of communication makes for a far simpler interaction. 


Our CS teams engage with our customers over the course of a day usually and having the freedom on both ends to come and go without losing the conversation levels up the customer support experience. One conversation is able to span the full scope of an issue all the way to its resolution and ensuring coverage of support becomes a lot easier with these asynchronous contacts because they almost entirely reduce unnecessary tickets

We're also able to be much more intentional with our chatbot. Collecting necessary context early on allows for much smoother handling by our agents 

BS: Love that, such a thorough and convincing approach for async being the future. 

Do you have any final tips that maybe we haven't covered so far for fellow support leaders and managers on successful implementation of AI-powered automation in support contexts? 

ME: Yeah, so implementation has been a big part of my role so far, and I really find it to be a team effort. Making cross-collaborative communication happen early and often is incredibly important. 

Getting a lot of people involved in testing. When a bot builder tests, they know what they're looking for and what they're seeing so getting fresh eyes on the bot is the best way to find improvements and problems. 

I'd also say that you should give room in the road map for new ideas. Some of the best ideas that I've had, and that I've heard, have come when we least expected it. So having some flexibility for that can be really valuable. 

"I’d also say a great team built on problem solving and customer service excellence can make anything happen. "

BS: I agree. That's so nice. 

So getting away from the kind of realities of implementation in the present tense and really letting your imagination go a bit crazy, let's talk about the next five years. 

Obviously, we've seen such an increase of innovation just in this last year. So when you think about what AI for leading support teams could look like in the next few years, do you see any trends, or do you have any predictions?

ME: I think the piece that always continues to shine is that the more you center the agent and the customer in the building of all of this, the better. 

Even with what we're seeing with support chatbots across the industry, they can miss the mark. And that really happens, in my opinion, when the bot builders and the managers aren't carefully considering that everything the bot says and does needs to be impactful, helpful, easy, and clear. 

"I think as long as we, as an industry, keep the customer experience – and the positive impact that can come from an interaction with an AI chatbot – aa the focus, and are careful not to lose track of that, the future is really bright."

I have a little bit of concern that if that loses the energy behind it, it can become difficult for companies to get back to having really excellent customer support.

BS: Stay human, stay close to your customers, and really use technology as a tool versus something that you can't control. 

ME: Customer support teams still need to be about that human connection between the teammate and the customer. The chatbot needs to be a tool to make that conversation efficient, quick, and helpful – not the be-all-end-all. So as long as we continue to consider customer support to be about building that connection between the customer, the company, and the brand, I think we'll reach a really fun and innovative future. 

BS: The future is bright. And on that fun and innovative note, I think we will call this to a stop and just say a massive thank you for spending the time to talk to Ultimate today. We love the work that we're doing together with you at Taskrabbit and are excited for lots more greatness. 

Thank you so much, bye.

ME: Bye.

Gaze further into the future of AI