AWS re:Invent 2025 - Reimagining Public Sector with AWS Partner generative AI solutions (WPS203)

🦄 Making great presentations more accessible.
This project aims to enhances multilingual accessibility and discoverability while maintaining the integrity of original content. Detailed transcriptions and keyframes preserve the nuances and technical insights that make each session compelling.

Overview

📖 AWS re:Invent 2025 - Reimagining Public Sector with AWS Partner generative AI solutions (WPS203)

In this video, Mehmet Bakkaloglu, Principal Solutions Architect at AWS, presents a framework for scaling generative AI through partner solutions, addressing challenges identified by Gartner. He categorizes solutions into four layers: implementing generative AI securely (Datadog, Dynatrace, Coalfire), preventing data loss (Forcepoint, Zscaler), embedded generative AI in applications (Elastic, Trellix, Qlik), and solving business problems (C3 AI, Salesforce, Snowflake). A survey of public sector leaders revealed 40% prioritize secure implementation and 30% focus on data loss prevention. The session includes a microsite with solution guides and partner briefs for each category.

; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.

Main Part

Understanding Generative AI Adoption Challenges and the AWS Partner Framework

Alright, thank you all for joining. My name is Mehmet Bakkaloglu, and I'm a Principal Solutions Architect at AWS working across our global software partners. Today I'm going to talk about how these partners can help you scale generative AI across your organization.

First of all, we are going to start with an interesting study by Gartner on challenges in generative AI adoption that organizations face, and then I'll be presenting a framework that will help you navigate partner solutions and address those challenges.

According to this study, we can put customers into two categories: high maturity customers and low maturity customers. The high maturity ones are those that have already implemented generative AI and put it into production. However, what they are struggling with is scaling generative AI across their organization. One of the main reasons for that is the security risks that come with using generative AI broadly across your organization. If you were to imagine having even a chatbot and if you open that up to a lot of users, that can lead to sensitive data loss.

Another big challenge they have is around integrating AI into existing applications. As you can imagine, in any organization you'll have a range of software systems and SaaS applications. In order to integrate AI into those, it usually is the software provider who needs to do that. Right now I would say we are at an inflection point where a lot of our SaaS partners that you might be meeting today are transforming their SaaS platforms from SaaS to Agentic SaaS. Essentially, by adopting those features of those SaaS platforms, customers will be adopting those generative AI features.

If we look at low maturity customers, these are customers that might have dabbled in generative AI, but they are really struggling to find the right use case. Maybe they don't have the data quality or the technical skills required, and sometimes it's because they have approached generative AI as a technical project as opposed to working back from a business problem. Overall, what we can say is that the challenges at the bottom layer are around security, in the middle layer it's around integrating AI into existing applications, and at the top layer it's around finding the right use cases and solving the business problems with generative AI.

So you might be saying, why AWS partners? Well, as you may know, AWS originated or started around 20 years ago, and we have always been very good at providing building blocks. More recently, of course, we are in the solutions space as well. Specifically, if we look at building blocks in the generative AI space, we offer the most price-performant infrastructure. We offer a range of foundation models from AWS as well as third parties.

If you look at foundation models from the very beginning, we have said that there is not going to be a single foundation model that will rule all of them. Depending on your use case, performance, and cost requirements, you need to select the right foundation model, and we are seeing more proof of this as time goes on. We also have tools for Agentic AI and guardrails that come with that to ensure that you are building safe and secure generative AI applications.

At the solution level, we have offerings like Amazon QuickSight, which are more directed towards the business users. If we look at our partners, we also have partners who provide building blocks like foundation models, but where our partners really shine is in the solutions that they provide. That's because these partners have a lot of vertical and horizontal expertise, especially when it comes to sectors like public sector. They are also very knowledgeable about how to get the compliances and accelerate those compliances.

But the challenge remains in how do you navigate and choose the right partner solutions for your mission. What we have done is working back from those challenges which Gartner identified, we actually created a framework and we worked with a select group of partners. We've created a microsite where we host solution guides to those partner solutions.

At the end of this presentation, I'm going to provide you the QR code for that site so you can go and review those solution guides. In terms of the framework, what we have is at the bottom layer we have partners that can help you implement generative AI securely. One level above that, we have partners that can help you prevent data loss. You might say these two are kind of related, and that's right, but the reason we separated these out is because DLP is a very specialized area and we have some partners that have specific solutions for DLP. And then one level above that, we have partners that have embedded generative AI into their applications. This is, I would say, a very exciting area when it comes to SaaS partners. And then at the top level, we have partners that can help you solve business problems using generative AI.

Four-Layer Partner Solution Framework: From Security to Business Problem Solving

So now let's look at the first category. In this first category, implementing generative AI securely, we have partners like Datadog, Dynatrace, and Coalfire. If you think about an AI stack from a security perspective, every layer of the AI stack presents unique risks. If someone, for example, gets access to a foundation model, that can create a lot of impact on your public sector functions, but in general as well, in every industry. And of course, at AWS, security is top priority for us. We provide a comprehensive set of security, access control, encryption, as well as monitoring features.

But where these partners really shine on top of that is in areas such as unified AI observability. Because within your organization, when you implement generative AI, you might have some applications powered by Bedrock, some by SageMaker, but also others by third parties. So these platforms like Datadog and Dynatrace are very good for unified AI observability. Other areas are like real-time monitoring and security posture management. One of the things to bear in mind with generative AI is that when it comes to observability, it's not just about your traditional metrics like latency, but also you need to look at accuracy as well as hallucinations and how that is impacting your business. And then the final area is compliance, where we have partners like Coalfire that are very experienced in helping you navigate those compliance requirements and get your certifications for your solutions.

And then moving up one level in data loss prevention, we have partners like Forcepoint and Zscaler. So what we mean by data loss prevention is, let's say you are using a foundation model and you ask a question. In that prompt, there might be some sensitive data, but also you might be supplying a file that is related to your business that has sensitive data. That's one example. Another example is the response you get from that foundation model may also pose a risk to your business, and especially in the public sector, that can lead to risks related to national security, citizen privacy, and ultimately affect public trust in AI.

So some of the benefits of these partner solutions are, when you implement generative AI across your organization, these DLP solutions will apply rules consistently across all of those data channels. And also, users in your organization might also inadvertently be using other AI tools as well, so the risk from shadow AI is also a very real risk, and these tools can help you with that too. The other area is some of these partners also offer industry and country-specific rules. So the way you deal with PII data differs from country to country. In some countries, it might be perfectly fine to supply PII data as long as you have customer consent, but in others it may not be. And then also, industry rules can differ from country to country too. In some countries, for example, manufacturing and semiconductor data can be considered nationally critical data, whereas in other countries it could be oil and gas.

And then the final area here is, when a user is using a generative AI model, typically you start using it and over time you might become more brave and ask more questions, and that could also inadvertently lead to data loss. So these tools can also dynamically adapt rules based on user risk.

Now moving up one level, this is where we get into what I would say are the more exciting generative AI use cases. In this category, we have partners like Elastic, Trellix, and Qlik. The great thing about these partners is that they have already done the hard work of selecting the right foundation model, integrated it into the platform, built guardrails around it, and looked at which workflows they want to automate. So as a user, by basically using that feature, you are automatically adopting generative AI.

According to Gartner, embedded AI is the largest and fastest growing segment of AI capabilities. If we look at these partners specifically, for example, Trellix, one of the studies which they did showed that the security alerts which the platform generates, users can only review about 10% of those because they don't have time. So what they did is they automated the review and the actions you need to take on the back of those security alerts with generative AI, and then the more high risk security alerts is where they have a human in the loop to deal with.

Elastic, for example, has powered their search using generative AI. So in your organization, if you have some kind of search, maybe it is using outdated keyword search, using Elastic you could actually now power that using generative AI. And then with Qlik, as you may know, they are one of our data analytics partners. If you look at business intelligence, traditionally you had dashboard developers who were very skilled in building dashboards. Now with Qlik, actually this is streamlined such that a business user who doesn't need to have deep technical expertise can create these dashboards.

And then finally, moving up to the top level, we have partners that can help you solve business problems. In this category, we have partners like C3 AI, Salesforce, and Snowflake. For these partners, the difference from the previous one is that here it's more either vertical or horizontal business use cases, and you might need to do a bit more customization based on the customer's data. But one of the key things to note here is that in order to solve business problems, you need to have a very good data foundation in place, and these partners have tools within their platforms that allow you to build that data foundation very rapidly and then apply the reasoning ability of a generative AI model to solve the actual business problems.

If you look at the public sector, there is still actually a lot of Excel-based analysis that is happening. And even if, let's say, an organization has built a data platform, they might have a data lake, they might have dashboards that are being generated, but typically it stops at the level of a dashboard. So a human still needs to go in and analyze what that dashboard is saying and on the back of it take actions. Now with these partners, those actions can also be taken with a generative AI model.

An example of this, for instance, is Salesforce, where they have a healthcare use case where they are able to review patient notes, on the back of that determine which actions to take, and take the low risk actions with the agentic AI solution. Fraud detection is another big area, especially when it comes to public sector, tax-related fraud, benefits-related fraud, pension-related fraud. This is also another good area in which these partners work.

Survey Insights and Accessing Partner Solution Resources

So to summarize, a good way for you to navigate these partner solutions is by starting from the bottom, looking at partners who can help you implement generative AI securely, partners who can help you prevent data loss, and then partners who have already embedded generative AI into their platforms, and at the top level, partners who can help you solve business problems. Now as I mentioned, we launched a microsite with these partner solutions, and as we were driving traffic to that microsite, we also did a survey where we asked public sector leaders, how can generative AI transform your organization?

Very interestingly, 40% of them said that implementing generative AI securely was top priority for them, followed by 30% which said data loss prevention. Now of course, once you've done those, that's where you get into the more exciting area of leveraging embedded generative AI and solving business problems.

As promised, here is the QR code for the microsite that we launched. When you go to that QR microsite, you'll find those four tracks that I mentioned. For each of those tracks, you'll find a guide which will explain in more detail what that track is about. So for example, you can understand a little bit more what embedded generative AI is and what are the example use cases.

And then for each of the partners, what we did is we worked with a third-party agency where we interviewed a leader from that partner to learn about their solution, and on the back of that we created a solution brief. So you can go to that microsite, download the solution brief, and if you choose so, you can engage further with those partners to understand their solution and adopt those solutions.

So with that, I thank you. And then there's one more thing, which is some of our partners actually have booths here, so you'll be able to find Datadog and Dynatrace at their booths, Zscaler for data loss prevention, and then Elastic, Qlik, and Trellix for embedded generative AI, and then Salesforce and Snowflake for solving business problems. And in fact, before I forget, we also have some of our partners here, so we have our partners Elastic and Qlik who are here, so you might want to also talk to them about their solutions and how they can help your organization.