From AI Assistants to AI Agents: What Enterprise Adoption Looks Like Now with AWS

3M’s sales representatives are saving more than five hours a week per person across customer prep, meetings, and follow-up analysis after adopting Amazon Quick. A corporate legal VP built a long-horizon scanning agent that tracks patent filings and new regulations worldwide, saving millions in outside counsel spend through daily updates instead of quarterly reports.

At the Six Five Summit: AI Unleashed 2026, Patrick Moorhead speaks with Jigar Thakkar, Amazon Web Services VP of Agentic AI for Business, for a look at how enterprises are moving beyond chat-based assistants toward autonomous agents that can act across systems and complete work on a user’s behalf.

Thakkar walks through why top-down cultural commitment separates companies capturing real AI value from those stuck running small pilots, and why AWS built Quick to connect directly into a company's existing productivity, CRM, and line-of-business systems rather than compete as another point solution. He details how Quick constructs a knowledge graph from that connected data, and lays out the governance layer behind AWS's newly launched long-running autonomous agents, including identity inheritance, admin-level controls, and full audit trails.

Key Insights:

🔹 Scaling AI begins with executive commitment to redesign work. Thakkar contrasts top-down mandates to reimagine end-to-end processes with isolated experiments that simply layer agents onto existing workflows.

🔹 Quick was built as a cross-platform layer, not another closed application ecosystem. Thakkar argues that AWS’s lack of a legacy productivity suite gives it greater freedom to connect across the tools, SaaS applications, and business systems customers already run.

🔹 Quick’s knowledge graph turns fragmented data into usable enterprise context. Structured data, unstructured documents, CRM records, messages, and other connected sources create a working memory that understands relationships across people, projects, and open work.

🔹 Long-running agents make governance foundational. Thakkar says Quick’s autonomous agents can continue working for days, but their permissions never exceed those of the people they represent, and their actions remain fully auditable.

🔹 The strongest results come from reimagining work function by function. The examples from 3M’s sales organization and a corporate legal department demonstrate how connected data and purpose-built agents can produce measurable time and cost savings.

AWS spent two decades abstracting infrastructure. With Quick, it is applying the same idea to enterprise work: how much context can be connected, how much manual coordination can be automated, and where human judgment must remain in control.

Watch the full video at sixfivemedia.com, and subscribe to our YouTube channel so you never miss an episode. For more sessions from this year's event, visit sixfivemedia.com/summit.

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

‍

Patrick Moorhead:
Welcome everybody. We are back at the Six Five Summit AI Unleashed 2026. We're going to be talking about enterprise AI software agents, and we're going to be talking about how the enterprise has gone from basically chat-based systems to autonomous AI agents to get a lot done. Throughout the organization, doing it smarter and faster. Joining us is Jigar Thakkar, ADBS VP of agentic AI for business at Amazon Quick. Jigar, welcome to The Six Five and The Summit. Patrick, thanks for having me here.

Jigar Thakkar: 

It's great to have a chat with you today on this topic.

Patrick Moorhead: 

Yeah, you've got a great job. I mean, essentially, agents are the hottest thing going. Who knows? I still think they're going to be hot two years from now, but it was funny how quickly we went from basically chat-based assistance to AI agents. And this year, our summit is all about unleashing AI, which is really about bringing true enterprise value. Everybody loves cool tech. I love cool tech. You do, our listeners do. But in the end, it's about enterprises capturing value. And in the last 12 months, enterprises have been working very hard to do this. And I'm curious, and this is kind of a broader question here, from where you're sitting in your orange room there, what really separates the enterprises, the government, just organizations at whole that are realizing true value from AI versus those who are still stuck in the pilot phase? And based on our research, there's a lot of them out there. Right, Patrick, that's a great question.

Jigar Thakkar: 

We're actually at an inflection point where, you know, Generative AI has been around for three years, but the agentic AI side of things, in the last few years, there has been questionable ROI in some pockets. And what we are seeing now with QUIC is extremely encouraging. And when we see our customers going from the basic questions and answers world to actually trying to take get letting AI take actions for you and then actually starting to focus on outcomes or hurdles they have. They're getting into a lot more sophisticated areas of how to use AI and now it's very clear to them what is the value of AI. And to give an example, and to do a question about what is the difference, right? So before that, I want to share a couple of examples, right? If you look at 3M, their sales agents, their sales representatives who started using Quik, after a few weeks, they reported, on average, they're saving five hours per person in all their customer prep, customer meetings, customer follow-ups, analysis of what was discussed. That's enormous amount of time savings they're reporting themselves. Another example I would share is from a legal VP who created a long-horizon scanning agent, which is looking at all the patent infringements, all the new laws around the world coming up, and they're saying they're saving millions of dollars in outside counsel because these things get done, and that too, not like once in a quarter you get a report. But pretty much every day, every week, they are getting information back from the agents. So that's very encouraging to see in the enterprise. And to your question, what is the difference maker here? I think the companies which has top-down commitment of a cultural change to say from CEO down, say, you know, we want to reimagine how the company is working, how work gets done. I think that's where we've seen the biggest delta. The areas where we are asking IT teams to do small pilots here and there, that can go on for months on analyzing ROI. Because what happens is if you're applying AI to a small area, versus reimagining how long operational processes in different functions and across different functions get done, you need to take a holistic view of reimagining how a company operates. And that's when we are seeing really big impact from QUIC on our customers.

Patrick Moorhead: 

Yeah, it makes a lot of sense. And whether it's top-down support from the board all the way down to not just putting agentic workflows on top of a busted process, but re-looking at the process. And then the people element of, you can have great value even with a headless agentic workflow, but in the end, people are going to be at the end of this, they have to trust it, they have to want to use it as well. So, I wanted to follow through and I'll put this under the AWS permission to play in this space. I mean, let's face it, AWS built the cloud, I think you're on your 16th, 17th year, You invented the concept and you certainly made it popular. And you had a lot of SaaS companies who trusted you and still trust you today to operate on your infrastructure platform. But this is an application, right? QUIC is an application, and it competes with a lot of the, not only the biggest frontier models out there, but also the other hyperscalers that are offering this sometimes. bundling it in as part of an enterprise license or something like that. So strategically, why do you win here and why now?

Jigar Thakkar: 

This is the right time. It's as I said earlier, we're at an inflection point. You're either seeing walled gardens or wild gardens. And they both have their challenges. If you think about the last 20 years, how the phones evolved from jamming windows into a small form factor to iPhone coming out with a completely new approach. Or you look at the electric vehicles, like either putting a battery inside an internal combustion car that you've been building for the last 30 years, or reimagining what an EV should look like. And that's where we are. I see the last 25 years or so of thousands and thousands of line of business applications, SaaS applications, productivity applications have been created. And all of them are extremely valuable in various different ways. And they will all be adding AI in pockets in all of these areas. But this is the time. this is the time to reimagine how work gets done. Do you really want a sidecar on every single application in the service? Do you want a remote, five remotes in every room in your house, or do you want the universal remote? And the approach with QUIC was very simple. And actually, to your question, why Amazon is doing it, when you don't have a productivity suite to protect or a SaaS application or line of business application like a CRM system to protect, you can really think from a clean slate. So the way work gets done in the future from Quix's perspective is it connects to everything you have, whether it's a productivity app or its SaaS app, line of business app, and so on and so forth. And that unique perspective adds a ton of value to the customers. And if you think about the evolution of the cloud for the last 20 years at AWS, it's always moving further up the stack. It was just storage and computer networking, then came this database services, then came analytics on top of that. We're adding more and more value further up the stack to get businesses done, and our customers trust us for that. They also want model choice. They don't get locked in with one thing. They also want platform choice. They don't want to get only locked into the office ecosystem or Gmail or Salesforce or whatnot. They want to be able to have their AI work across everything. So this gives us the perfect permission to execute. And honestly, I don't believe we are competing with these legacy apps or the older ways of doing things. We are competing with the inefficiency that is in enterprises. We are competing with a lot of manual labor that people are doing that they don't need to be doing. We want them to be focusing on creative things they can do, focus on their customers, focus on building something new, focusing on growing their top line, focusing on increasing their margins. They will do all of that when we can take away a lot of groundwork. And that has happened in the cloud for the last 20 years. This is, to me, just a natural progression that happens to be an app. But underneath the covers, there is a tremendous amount of AI and agents that are doing all the heavy lifting, the governance behind it. So I think this is very much up to the CEO of what AWS has been doing for the last 20 years.

Patrick Moorhead: 

Yeah, I think this makes sense. And I actually, I do appreciate the way you answered that. One thing we've learned through this journey, and it's funny, I took my first programming class in 1984, and my teacher said, garbage in, garbage out. And it even goes back to the mainframe days. Essentially, and I think this is, even the truest or most true in this world of agents, outcomes are directly related to the quality of the data that it's touching. And listen, every vendor tells me their agent is smart. But in the end, a lot of it comes down to what knowledge do they have access to? How did they get there? Is it good data? And that's, you know, when I talk to the CIO and implementation teams, they tell me, you know, in the end it was about, you know, okay, we got the culture, at least for this application, but it was the hardest part was getting the data. Walk me through how Quick gets access to the right data. and how that delivers value. I mean, the model is important, but context, it seems, I believe in the future is gonna mean even more.

Jigar Thakkar: 

That's a great point, Patrick. And I think the best companies have realized after three years of looking at this challenge that they already have what is the most important thing in making them successful in the world of AI, their data and their context. Protecting it, growing it, and making the best use of it is what's going to make them helpful in the world of agentic AI. Models are obviously important, but there are dozens of them out there. And what is critical for them is the data, as you said. So how does QUIC handle this? I think that is the core challenge. That is the core area of our focus. We connect with everything you have. your, and then we create a memory and knowledge graph out of it. So it could be structured data, unstructured data sitting in documents. It could be data lakes. It could be your CLM system. It could be Outlook or the Office Graph. It could be Slack or Teams. So when you connect with all your line of business data, all your productivity data, all your transactional data that is operational in nature for various different functions, finance or marketing and so on and so forth. What we do behind the scenes is we create this knowledge graph. So QUIC knows who you are, who are your people, what are your projects, your files, what are your outcomes, what are your hurdles, what are you looking at doing this quarter? It really starts to understand the more you use, it starts understanding more and more of that stuff. So when you start asking a question or you want to understand why this thing stuck for the last three months, and it could be a big project, it could be a budget proposal, it could be a presentation we need for the board meeting and whatnot, it has the full understanding. And it actually comes and tells you, here are the artifacts related to the question you just asked. Here are the three people you should reach out to. Here's where I think things are stuck. And that is the core thing, that the more data we connect to and we have probably the largest source of enterprise data on AWS. And then you can connect with all the productivity data through either Gmail or Outlook and Teams and Slack and so on and so forth. and then you can reach into the line of business applications. Then you have a full 360 view of what your work looks like. And that's when you start solving the problem and you realize that, look, you already had that data. The answer exists in your company somewhere. The context exists somewhere, but you did not know where to find it. quick solves that for you. And of course, we use the best in class models to then compile it all together to give you the actions you can take directly or do it on your behalf as well.

Patrick Moorhead: 

So, you know, it used to be the days where we were talking about guardrails, when we were talking about essentially a chatbot. But you recently announced autonomous agents that literally run for days, right? And if I look at the theme of this year's summit, it's AI unleashed. And that is exactly AI unleashed, as long as the output is what you're looking for. And it does make CISOs and boards a little bit nervous, right? Headless AI agents is starting to do actions on employees' behalf. that they may or may not be the human in the loop on. How are you helping organizations ensure that there's accountability and limiting the impact that something goes grievously wrong?

Jigar Thakkar: 

Actually, Patrick, that is the main feature. We've had these autonomous agents running and built a long time ago. And we decided to only launch them when we had the confidence on the governance aspects of it. So the sophisticated governance guardrails are the actual main feature here. And I give some examples of how it works, but when we speak with business leaders or the CEOs, I want them to see the full power of what an agent can do. You know, what used to take months for dozens of people, how can an agent do in a very sophisticated way by creating it just by chatting with Craig? So they see the power of what this agent can do. On the other hand, when I speak with folks in security, compliance, governance, IT, we show them the breadth of governance features we have. And that's what drives the confidence to say, you know what, if we have all of these controls, we can actually roll it out. to the entire company and let them go be creative with it, because we don't want the entire company to worry about it. So to go back to your question on how do we do the guardrails, so agents run on behalf of the humans who are accountable at the end of the day. It enhances the identity, so you never increase the permission or access to an agent more than the human who's executing it. You always have a subset of the boundary that the human has. On top of that, there'll be admin controls at the connector level, the tools level, document level, and these things can be set by an admin. So even if I decide I want to create something with all the access I have, the admin may decide to prevent some of these things based on the company's policy. Further, there are user-level controls that I can decide what my agent can do. And a very simple example would be If a certain email comes, I can create an agent to say, monitor my emails and do these three things. At the same time, I may say that only send this email to these three people and cannot send out an email to the whole company. So I can create that kind of restrictions. Also, agents will have their own identity. So you can distinguish what I did personally versus what my agents did. And that way you can further narrow down what's going on. And the last thing I want to point out is the complete auditability. Everything is tracked, audited, and that's where we can see the paper fail of what's going on in the company.

Patrick Moorhead: 

Yeah, and that audit trail, I mean, all companies, but particularly in regulated industries, was really the difference between are these companies able to do AI or aren't they? And the other thing we're seeing too is the number of potential projects that they're looking at has actually decreased, and the ones that they kept were the ones that they could govern. So yeah, governance is a key feature here. And I love the kind of the fine tuning that you're doing with these guardrails. It just makes sense. You're not being too onerous that you can't do any headless work, nothing gets done. You're letting people kind of pick and choose this transaction amount. In this country, these people are the only people I want you to send this to. And I think that to me that is showing that we're at a sense of increased maturity in an industry as delivered by QUIC. So it's been a great conversation. I've got one final question here. I talk to CIOs every week. In fact, I was in London talking with a group of them before Wimbledon. And they were talking about what you would expect, hey, we initiated brought in deployed chatbots, co-pilots. It looked amazing. 12 months later, they can't explain the ROI that the CFO needs to see. So can you talk about the biggest determinant of success that you're seeing when customers are scaling at, well, deploying QUIC at scale across tens, thousands, maybe hundreds of thousands of seats?

Jigar Thakkar: 

So we definitely, like I said, the previous approach of adding the sidecars and chatbots and copilots next to every single app, and they don't talk to each other, they don't have context of what's going on in every app, it becomes very difficult to maintain because you're already doing so much context switching. This is why they did not see the ROI. Humans are doing so much context switching on top of that. Even AI is now, it's not contextually connected to all those surface areas. This is why Quick took this very fresh approach that we connect with everything and it has a common context, memory behind it, knowledge graph behind it. We also measure the ROI using LLM as a judge techniques inside Amazon and we encourage our customers to do the same thing. So what are the things that differentiate? I think, number one, there are three things I'll say. One is the cultural change. You want the entire company to know that we want to have a new way of getting work done and leveraging Quix to really reimagine how these large operational processes can be compressed. The second we talked about was data and context. Any AI is going to be only as smart as what you connect it with. So all the knowledge bases, the data lakes, the CRM systems, finance systems, when you connect all of these things, when the magic happens, people who use our desktop app tell us after a week or two of usage that this feels magical. It's actually not magical. It just connects with everything. And there is just a lot of science behind it. And they tell us, we don't know how we worked before this because it's so proactive. It tells you what are the top three things you should focus on this week. The last thing I'll point out is for each function, we encourage companies to look at unique operational challenges, whether it's sales or finance, HR, and so on and so forth, and across these functions as well, and then make every employee a builder to solve these problems. And the way you build, just chat with Quick.

Patrick Moorhead: 

That's it. No, it sounds great. I was super impressed when I saw the data sources that Quick had out of the box. I thought that looked great. But yeah, culture, data, workflow, and getting the people actually doing the work engaged. So I think those were sage and wise words here. Jigar, I want to thank you for joining us, chatting about enterprise AI software and agents as part of our 6.5 Summit AI Unleashed. Thank you so much. My pleasure, Patrick. Thanks for having me here. And all of you out there, don't forget to hit that subscribe button, follow us on social media, check out all of our Six Five Summit content out there, and I'll see you next time.