Amazon’s Jose Kunnackel on How AI Agent Interoperability Boosts Productivity and Reduces Toil

Hello, and welcome to the latest edition of the Techstrong AI Leadership Series. We're here with Joes Knuckle, who is, uh, lead for the Amazon Cube business, which is their whole effort around AI agents, and we're talking about the latest updates to the platform, including integrations with Zoom. Hey, welcome to the show.

Hey, thank you, Mike. Very happy to be here. Bring us up to speed, if you would.

I think it's been a little bit more than a year since the initial formal launch, and, um, there's been a lot of integrations including this latest offering with Zoom. But, um, what else is going on here? Well, we've got, uh, you know, we've got a fair bit going on.

I think maybe I'll touch on a few different integrations and updates that have happened. Um, one of the key things is that Q Business has continued to be adopted by a number of organizations, and I think we, we had a very large, um, uh, uh, uh, rollout with NFL, which was part of their draft iq. So, um, that, that was interesting.

And then we've got a couple of other, um, integrations like you mentioned. So there's Zoom, which we launched in April, and we also just launched PagerDuty Advance, so that just launched in the last week. So maybe we can, we, we can talk about a few of these just so you get a flavor for the different ways in which people are using Q Business.

Well, let's start with Zoom, since I think more people are using that in general, but, um, how is this changing the way we think about working ourselves and with each other? Because it, it seems like on the face of it, a lot more data's gonna be literally available at our fingertips. Oh, that, that's very true.

So I think one of the biggest challenges that, uh, users have is looking at data, which resides in multiple applications, and how do you get the context of what you want to do when you need it without having to browse, you know, like having that application overload. So, um, well, if you, if you just think about how Q Business works, Q Business provides a way for an enterprise to bring together multiple sources of information while making sure that you have the permissions in place so that each user sees the data that they are supposed to see, and then they are able to search and get those insights as and when they need them. So what, what we've done with the, uh, with partners like Zoom is that Zoom now integrates with the Q Index, which is the index that analyze Q Business.

And as you are using Zoom's custom AI companion, so that's the, the AI offering inside Zoom, it is able to pull together insights that are relevant to the, the activity that you're doing. So, for example, if you're in a meeting and you're asking a question of custom AI companion, that can now be kind of enriched with data from the Q index, which where the data can be from, you know, your SharePoint or from your Outlook, it can be from a number of sources, but you have that sort of fingertip and Zoom's able to present that. So the biggest thing is it is taking away that notion of information overload where you have to, you know, flip to the next step and the next step to get that.

But you're getting it all together, uh, in the, in the places where you work. Is this kind of an example of how multiple agents from different providers are gonna interact, operate with each other, and to create some sort of greater good, but this is just the beginning. Oh, definitely.

I think, I think you, you should expect that multiple agents from different providers are going to work together to make sure that the user is able to ultimately achieve what they want. So like you said, this is a good beginning where you're seeing that Zoom and Q business, right? So Zoom's kind of tapping into the data from, uh, Q Business, uh, in order to get this out to the user, but we are also looking at how this can act in many different circumstances.

So I mentioned the PagerDuty example right now, which was, which is something we just launched in the, uh, last week. Uh, now with PagerDuty Advance, um, Uh, Users, technical users are able to look at incidents and then understand what is needed to resolve them. Now with the integration with the Q index, PagerDuty Advance now has the intelligence of other information that might be relevant to them from the enterprise.

And the, the Q index can be used to share that information securely with PagerDuty users. And so definitely, you know, you've got these multiple, uh, agents and agent tech systems working together now so that they can, uh, interoperate to get you the best results. In fact, one of the things about the PagerDuty, uh, launch was according to PagerDuty, is, um, can internal studies, they've seen that incident response rates have improved by over 30% as they have this additional context so that, you know, in the middle of an incident you don't have to run to, uh, go check on another system, get that information.

It's right there. The same thing with Zoom, right? So you are, you're in a meeting, we are in this meeting, and then you are, you are trying to find something.

You can get that right from a custom AI companion without having to run to that other system and ask for those pieces of information. Do you think that this is changing the way we work in an interesting way? Because, uh, I don't think we always appreciate how much toil we go through to actually go get a piece of data, usually after the fact, after a meeting, or in the case of PagerDuty and some incident that takes forever to com to investigate and complete.

So, so will we kinda wake up one morning, it's kind of like that backache you had for 10 years and suddenly it goes away and you, you learn to live with it, but you now you appreciate the fact that you had this pain. I, I think, I think what you just called out learn to live with it is probably the key here because the way we've all been working is across, you know, many, many applications and we've all learned to, you know, it's routine. Like you said, after an incident, you go and, you know, do the things that you're supposed to do.

Whereas what you're gonna find is that as these systems start to work together, it becomes simpler. And that means that you are able to do these things faster and often, you know, spend more time, you know, thinking about the, the things you really want to achieve. And, uh, you just taking away a lot of the toil, I think you use the right words there, you just a able to eliminate a lot of that in order to do things faster.

There's a lot of obsession these days with all things related to productivity, and I get that AI will improve productivity, but I'm not quite sure it's going to do this in, uh, leaps and bounds as much as it's just gonna make the, the average work life more tolerable for people and we will improve. But I, I'm not so sure robots are taking over the world, but to, you know, how do you view the future work? Uh, well, I, I look at it similar to what you said, which is I think it's going to improve the way we work.

It's gonna help us do things better and easier, take away a lot of the toil, but you, you still want, the way, you know, you still want to achieve something the way you want. You don't just want output that is just produced and given to you. So in one of the previous discussions, I think we talked about one of the features in QuickSight, an agentic feature called scenarios that we, um, launched earlier this year.

Now that again, is doing data analysis using agentic approaches on the backend, but the, the user is able to actually define the plan, tell the system what it, what needs to be done, and to kind of pull it all together. So when you know somebody is sitting there and then saying that, Hey, I want to investigate what happens if, uh, I have 20% more sales, right? Will I need more inventory?

Will I need more, more, you know, uh, uh, distribution locations, if you have the data, you're able to create a plan and then have the system execute on it. And so you're making it simpler for the user to do some of these things, but you, you are giving people the control of how these systems work. So in a, in a similar fashion, I think you're gonna find that when, when you are able to, you know, in whether in Zoom or whether in pager duty, you're able to get the information right there, you are taking away a lot of the toil and it becomes easier for people to kind of operate with those systems.

Well, multiple agents start working together across people and teams because I can envision a world where I may have, I don't know, I randomly pick a number, but 10 agents that I regularly use, and you might have 10 agents that you regularly use, and we'll assign them tasks that require them to collaborate with each other, and that's kinda also gonna be the future of work in terms of a team setting. Definitely. I think those are some of the, the things that we are thinking through as we are thinking about the, the future of, you know, the various pieces that we have.

So if, if, if you think about, you know, our own offerings, we've got Q Business, which is allowing you to bring together data from, uh, unstructured sources. So essentially like, you know, files, SharePoint, all these different sources where you have all of those written notes and communications, and then you've got QuickSight, which has, uh, information about structured data, so more databases, metrics and so on. And then you've got these different pieces all coming together.

So we are thinking about, uh, how do these agents, or how do these systems interoperate? And that ex extends definitely to the, the, you might be using 10 agents and then how do my agents interoperate? There's questions about, you know, security, identity, all of those pieces that we are, um, kind of thinking through.

But that is definitely a, a, a, a major area for us to all think about In a lot of ways. We're also looking at essentially a business process, re-engineering at a different level of scale. Are there things that you're seeing organizations doing right upfront that you kind of wish other organizations would, uh, crib a little bit As we enter this age of agentic ai, I, I think, um, one of the foundational pieces, and I think many people in the, in the industry talk about it, is making sure you have the right data.

So what we are seeing is that when we work with, um, you know, customers like Nasdaq or Principal Financial Group who are adopting, you know, Q Business as an example and putting it out there to tens of thousands of their users, they also have to make sure that the data that goes in is the relevant data. So, uh, making sure that you have the right data that is being, uh, input into these systems and making sure that is cleaned and managed in the right ways. That's definitely something that is top of mind as you start to think about what are the systems and what are you trying to get them to do, because otherwise, you know, you end up with a situation of just having a lot of data and then not knowing, you know, what the, the system should pick on.

So that's definitely something I've seen, uh, companies think about, be very deliberate about it as they're rolling this out. And the other part of this would be, you know, how do you roll this out? What sort of users, what's the use case?

And then making sure you are aligned in terms of both the technology and the, the data and the, the user pieces. So in which do you think will be the bigger challenge going forward? Is it the technology or is it really gonna be the culture within the organizations?

Well, I think technology is going to keep evolving at a fast clip and, you know, we, we are all in this AI space and you can see how quickly it's evolving. I think, um, making sure that you are, uh, helping people in the ways that they want and not just, um, you know, presenting quick, uh, experiments or, you know, shiny objects that people can't really use in their day to day. That's going to be the, the trick here, meaning every company has to think about how do they make sure that these technology, um, advancements are being utilized in the best way that users can take advantage of it, rather than being a shiny object in the room that looks good, but doesn't really, you know, work in every, uh, or work in the different scenarios that people do work.

That that's key. So that's why many of the, um, the advancements that we've made in the last, you know, year have been focused very much on where do people do work and how do we help them. So if you think about the, the Zoom and the PagerDuty integrations that we talked about, that is about how Q business can really help you in the moment that you're in a meeting or you're trying to resolve, uh, uh, an issue.

And at the same time, we also have things in Q business, like we have a browser extension so that as you're browsing the internet, you can look at, uh, a page and then quickly ask you to summarize that page. We've got a Slack integration, we've got integrations into M 365, so we are building these integrations so that you can take those technology advantages and make sure that they are really beneficial to users as in when they do work, rather than just be, you know, like a, a great thing, but that's, uh, disconnected from the user. So definitely a big focus area, uh, for us, and I think it should be for every organization.

How do, how do the users come across technology advances and how are they being used? A lot of these AI agents are becoming increasingly autonomous as the reasoning capabilities of the LLMs continue to improve. But if they work together, how will they kind of negotiate humans?

All the, we negotiate all the time, every time we're in work scenarios with somebody, we're basically having a constant series of negotiations. How will a AI agents negotiate amongst themselves on our behalf? I think those are some of the areas of, um, you know, essentially I'd say, uh, not research, but areas that we are working through in terms of how do you, how do you set the, the boundaries of what you're willing to negotiate with.

Like if you or me were to have that conversation, those would be in our heads, right? We would have, you know, I'm not gonna, it's like when you, when you try to make a sale or a purchase, you, you know what your limits are, you know how comfortable you are with certain numbers, but, uh, you do want to make sure that the AI agents that are doing work on your behalf have that ability, or it's really the, the guardrails and the, the knowledge that you impart or you prescribe. And so as a result, many of the things that, uh, we are looking at is also how do you provide that control to business users so that these, um, agents, when they work on your behalf, you are giving them the, you know, essentially your thinking, how do you get them to do work on your behalf rather than just behave like autonomous systems.

So definitely an area that we are looking at, like, you know, how do you make this simpler for business users to manage control and to use many of the, the pieces that are out there. The scenarios work that I talked about, uh, earlier is a, is a great example because you are not asking the system, Hey, tell me what happens if sales go 20%? You're actually asking that of the system, the system's giving you a plan, you are modifying the plan.

You are basically putting in your, you know, prescriptions as a user and then saying, you know, what, um, you know, constrain sales to, you know, the North America region, and then think about what happens so that it's not just doing like an autonomous run about some fictitious, uh, object, but you are giving it the controls. And that's again, something that we are looking at how that expands to other parts of our system. So we, you know, we've got QuickSight Q business, so how do you just expand some of that?

So thinking at your crystal ball a little bit, where are we gonna be a year from there? The rate of which technology is advancing, that's gonna be a very interesting question, but I do hope that, um, just seeing all the, the work that you know is going on in the industry, we are able to provide users which much more AI in the context of their, of their business and their work. A lot of what you see today is in the context of standalone applications, right?

You've got, you've got all these applications where you can go do things on the site, come back to your application, do something, and what, what I, I think in a year from now, if we are in a stage where people are able to use more of the, you know, AI gen technologies in the context of their work as they're doing things, and um, that would be, that would be a great thing. All right, folks, you heard in here, hey, maybe we're all evolving past being just mere knowledge workers to being knowledge managers. And we'll do that with the help of some AI agents and maybe even, hopefully be less stressed out someday.

Hey buddy, thanks for being on the show. Of course. Thank you, Mate.

All right. ai series. Catch this episode, others on our website, we invite you to check them all out.

Until then, we'll see you next time.