2024 MIT Digital Technology and Strategy Conference: Lightning Talk - Maven AGI

SAMI SHALABI: Hi, everyone. I'm Sami Shalabi, MIT alum and founder and CTO of Maven AGI, where we're reimagining customer experience through AI agents. Customer experience is really, really broken. If any of you remember a good customer experience, raise your hand, that you've ever had with a vendor. It's a universally very challenging environment to be able to create great customer experience.

First off, it's insanely expensive. We've canvassed hundreds and hundreds of enterprise organizations and asked them, how much does it cost you to support your customers? Universally, the number that comes back-- it's an average of $40 a ticket. $40 to handle a support ticket where, in general, most people are frustrated with the overall experience. You take that to other parts of customer experience, like customer success. Every interaction is over $100.

This is not because the organizations are not trying to create a great experience. It's because it's a complicated experience. It's got numerous handoffs, the systems are fragmented, there's a lot of lack of automation, and there's just a ton of really bad self-service tools. On average, a support agent needs to use six systems to be able to answer any question that typically comes in.

What we're building at Maven are AI agents. Imagine if you contact support and the person talking to you knows everything about you, everything about the company, and is a domain expert. And not only is telling you how to do stuff, is personalizing the experience and is able to take actions on your behalf to solve your problems. We are building AI that can do that.

So how can these AI agents be used in support? And there are three high level use cases. For users, for our customers' customers, you can use AI agents to help drive a self-service experience where users are able to solve their own problems. That drives an increase in ticket deflection, which is less human agents, less cost, and more self-service and faster time to service. For agents, the folks on the other side, AI agents can be used to assist agents to actually solve the tickets. And in this experience, it helps for drive agent efficiency, and it drives better consistency for all users submitting these tickets.

And then lastly, for support leaders using AI agents, you're able to understand the state of your support environment when you have hundreds and hundreds-- in some cases, some of our customers-- millions of tickets, it's very hard to figure out what's what and where the issues are. And using AI, we're able to categorize and organize and present a unified view of where the issues are, ultimately driving better, issue visibility and ultimately fitting it back to the solution and really solving these customer issues.

So let's look at a real example. In this case, this is a personalized chat agent that was embedded into a SaaS product. And in this case, the Maven agent is loaded. It's able to retrieve the user's profile and figures out that the user needs to verify their profile. So this is turning support into not just a self-service, but a proactive experience. And then the customer asks the question, why do I need to verify? And it's able to answer that question in a fairly unified way, ultimately reducing the need to deflect the ticket.

This is another example where the user is trying to perform an action and hasn't been able to figure it out. And this is using our full page chat experience similar to ChatGPT where the user says, I just want to change permissions. It's able to disambiguate, well, there are two people with the first name Sami, and the user is able to select. And then the action is-- and removing all the complexity necessary to navigate complicated products.

Let's look at a different example. In the case of helping agents, this is an example of the Maven co-pilot coexisting inside of Zendesk. We support every desk there is. Salesforce, Freshdesk, Slack, Twilio-- every kind of support system out there. And the copilot coexists in this expereince. In this particular example, the ticket came in Spanish. The agent only speaks English. An answer is generated in Spanish and translated to English. And the agent is able to edit the answer in English, and it's automatically translated back to Spanish.

And when we've talked to many of our customers, the multi-language use case is where the costs are astronomically high. English to English is usually about $40. Anything in any foreign language typically runs around $120 per ticket at least.

And finally, AI-driven analytics to help understand overall efficiency. And we have dashboards. Not only is it just about solving the problem, is Maven working for you? but giving you insight. We automatically categorize tickets. Compute NPS's. Slice and dice the data so you have true visibility into your overall support experience.

And these are some of the results we've had with our enterprise customers. We've processed millions of tickets now. On average, our AI agents are able to handle 93% of support questions without humans. And the typical savings is 80%, and we've had significant increases in resolution times. And we've pretty much integrated with everything out there, and we're happy to integrate with more. And we have quite a bit of large enterprise customers that are really happy with our solution.

OpenAI saw the work we were doing with Tripadvisor, and if you go to openai.com, there is a detailed case study around the work that we've done with Tripadvisor where, in their case, they were able to answer 90% of their incoming queries without humans. If you are looking to improve your overall support experience, cut costs, we're open for business. We're excited to work across industries, and we can see you around the corner. Thank you.