iTranslated by AI
re:Invent 2025: A 4-Layer Framework for Generative AI Adoption in the Public Sector and AWS Partner Solutions
Introduction
By transcribing various overseas lectures into Japanese articles, we aim to make valuable hidden information more accessible. The presentation we're featuring in this project, based on that concept, is here!
For re:Invent 2025 transcription articles, information is compiled in this Spreadsheet. Please check it as well
📖re:Invent 2025: AWS re:Invent 2025 - Reimagining Public Sector with AWS Partner generative AI solutions (WPS203)
In this video, Mehmet Bakkaloglu, Principal Solutions Architect at AWS, explains organizational challenges in Generative AI adoption and how to leverage partner solutions. Based on Gartner's research, he points out that highly mature organizations struggle with security risks and integrating AI into existing applications, while less mature organizations struggle with finding appropriate use cases. To address these challenges, he presents a four-layer framework. The bottom layer focuses on secure implementation of Generative AI (Datadog, Dynatrace, Coalfire), followed by data loss prevention (Forcepoint, Zscaler), then embedded Generative AI (Elastic, Trellix, Qlik), and finally, business problem solving (C3 AI, Salesforce, Snowflake) at the top layer. A survey of public sector leaders shows that 40% prioritize secure implementation and 30% prioritize data loss prevention, and a microsite featuring solution briefs from each partner is also introduced.
This article was automatically generated to maximize retention of the original lecture content. Please note that there may be typos or incorrect information.
Main Content
Challenges in Generative AI Adoption and Overview of the AWS Partner Framework
Thank you for joining me. My name is Mehmet Bakkaloglu, and I am a Principal Solutions Architect at AWS, working with our global software partners. Today, I'm going to talk about how these partners can help you scale Generative AI across your organization.
I'll start with an interesting Gartner study on the challenges organizations face in Generative AI adoption. Then, I'll introduce a framework that can help you navigate these challenges and partner solutions.
According to this study, we can categorize customers into two types: highly mature customers and less mature customers. Highly mature customers are those who have already implemented Generative AI and have deployed it into production. However, what they struggle with is scaling Generative AI across their organization. One of the main reasons for this is the security risks associated with widespread use of Generative AI throughout the organization. For example, imagine if you have a chatbot and you open it up to many users. That could lead to confidential data loss.
Another big challenge they face is integrating AI into existing applications. As you can imagine, every organization has various software systems and SaaS applications. To integrate AI into them, typically the software provider needs to do it. We are now at an inflection point, where many SaaS partners you might meet today are transforming their SaaS platforms from SaaS to Agentic SaaS. Essentially, by adopting the capabilities of these SaaS platforms, customers are adopting their Generative AI features.
When we look at less mature customers, these are customers who have dabbled a bit with Generative AI but are really struggling to find the right use cases. They might not have the data quality or the necessary technical skills, and sometimes they approach Generative AI as a technology project rather than working backwards from a business problem. Overall, the bottom-layer challenge is about security, the middle layer is about integrating AI into existing applications, and the top layer is about finding the right use cases and solving business problems with Generative AI.
So you might ask, why AWS partners? As you may know, AWS was born, or started, about 20 years ago. And we have always been very good at providing building blocks. Of course, more recently we've been moving into the solutions space. Specifically, if we look at the building blocks in the Generative AI space, we provide the most cost-effective infrastructure. We also provide various foundation models, both from AWS and third parties.
Regarding foundation models, from the beginning, we have said that there isn't a single foundation model that will rule them all. You need to choose the right foundation model based on your use case, performance, and cost requirements. And over time, we're seeing more and more evidence of this. We also provide tools for Agentic AI, along with the accompanying guardrails, to enable you to build safe and secure Generative AI applications.
At the solution level, we have offerings like Amazon QuickSight, which are more for business users. When we look at partners, some partners provide building blocks like foundation models, but where partners really shine is in the solutions they provide. This is because these partners have a lot of vertical and horizontal expertise, especially in areas like the public sector. They are also very well-versed in how to achieve compliance and how to accelerate those compliances.
However, the challenge remains: how do you navigate and select the right partner solutions for your mission? What we've done is work backwards from these challenges identified by Gartner to actually create a framework and collaborate with a selected group of partners. And we've created a microsite that hosts solution guides for those partner solutions.
At the end of this presentation, I'll provide a QR code for that site so you can check out those solution guides. Regarding the framework, at the very bottom layer, we have partners who can help you securely implement Generative AI. One layer above that, we have partners who can help prevent data loss. You might think these two are related; that's true. But the reason we separated them is that DLP is a very specialized area, and we have partners who specialize in DLP. And one layer above that, we have partners who have embedded Generative AI into their own applications. I think this is a very exciting area, especially when it comes to SaaS partners. And at the top layer, we have partners who can help you solve business problems using Generative AI.
Security and Data Protection: From Secure Implementation to DLP and Embedded AI
So let's dive into the first category. In this first category, secure implementation of Generative AI, we have partners like Datadog, Dynatrace, and Coalfire. When you think about the AI stack from a security perspective, each layer of the AI stack has its own inherent risks. For example, if somebody gets access to a foundational model, that could have a significant impact on public sector functions and, of course, generally across all industries. And, of course, at AWS, security is our number one priority. We provide comprehensive security, access controls, encryption, and monitoring capabilities.
But where these partners really excel on top of that is in areas like integrated AI observability. Because when you implement Generative AI within your organization, you might have applications running on Bedrock, some on SageMaker, and even some from third parties. So, platforms like Datadog and Dynatrace are excellent for integrated AI observability. Other areas include real-time monitoring and security posture management. One thing to keep in mind when thinking about observability for Generative AI is that you need to look not just at traditional metrics like latency, but also at accuracy, hallucinations, and how that is impacting your business. And the final area is compliance, where we have partners like Coalfire, who are highly experienced in navigating compliance requirements and helping you certify your solutions.
Moving up one level, for data loss prevention, we have partners like Forcepoint and Zscaler. What do we mean by data loss prevention? For example, if you're using a foundational model and you ask a question, that prompt might contain sensitive data, and you might also be providing business-related files that contain sensitive data. That's one example. Another example is that the responses you get from that foundational model can also pose risks to your business, especially in the public sector, which can lead to risks related to national security, citizen privacy, and ultimately, public trust in AI.
The benefits of these partner solutions are that when you implement Generative AI across your organization, these DLP solutions apply consistent rules across all your data channels. They can also address the very real risk from shadow AI, as users within your organization might unknowingly use other AI tools. Another area is that some of these partners also provide industry and country-specific rules. The way PII data is handled differs from country to country. In some countries, providing PII data might be perfectly fine as long as you have customer consent, but in other countries, it might not be. And industry rules can also differ by country. For example, in some countries, manufacturing or semiconductor data might be considered nationally critical data, while in others, oil and gas might be.
And the final area is that when users use Generative AI models, they typically start using them and over time might become bolder, asking more questions, and that can also unknowingly lead to data loss. So these tools can also dynamically adapt rules based on user risk.
Now, moving up one level, this is what I would call 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 completed the heavy lifting of selecting the right foundation models, integrating them into their platforms, building guardrails, and thinking about which workflows they want to automate. So, as a user, you basically just use that functionality, and you're automatically adopting Generative AI.
According to Gartner, embedded AI is the largest and fastest-growing segment within AI capabilities. When we look specifically at these partners, for example, Trellix, one of the studies they conducted showed that users can only review about 10% of the security alerts generated by the platform due to lack of time. What they did was automate the review and necessary actions for those security alerts using Generative AI, while putting humans in the loop for higher-risk security alerts.
Elastic, for example, is enhancing its search capabilities using Generative AI. So if you have some search capability within your organization that uses old keyword search, Elastic can actually help you enhance that using Generative AI. And then Qlik, as you know, they're one of our data analytics partners. If you look at business intelligence, traditionally you had very skilled dashboard developers who built dashboards. Now, with Qlik, this is actually streamlined, allowing even business users who don't need deep technical expertise to create these dashboards.
Solving Business Challenges and Leveraging Partner Solutions
And finally, moving to the top level, we have partners who can help solve business problems. In this category, we have partners like C3 AI, Salesforce, and Snowflake. The difference between these partners and the previous level is that here we have more industry-specific or cross-cutting business use cases that might require a bit more customization based on customer data. But one key thing to note here is that to solve business problems, you need to have a very good data foundation in place, and these partners have the tools to quickly build that data foundation within their platforms and then apply the inference capabilities of Generative AI models to solve real business problems.
If we look at the public sector, there's actually still a lot of Excel-based analysis being done. And even if an organization has built a data platform, they might have a data lake or dashboards being generated, but typically it stops at the dashboard level. So, humans still need to analyze what that dashboard is showing and take action based on it. Now, with these partners, even that action can be performed by Generative AI models.
For example, in the case of Salesforce, we have a healthcare use case where you can review patient notes, determine what actions should be taken based on them, and perform low-risk actions with an agentic AI solution. Fraud detection is also a big area, especially in the public sector, with tax-related fraud, benefits-related fraud, and pension-related fraud. This is another good area that these partners are addressing.
To summarize, a good way to navigate these partner solutions is to start from the bottom. Look at partners who can help you securely implement Generative AI, partners who can help prevent data loss, and partners who have already embedded Generative AI into their platforms, and at the top level, partners who can help you solve business challenges. Now, as I mentioned earlier, we've launched a microsite featuring these partner solutions. And while promoting traffic to this microsite, we also conducted a survey and asked public sector leaders, "How can Generative AI transform your organization?"
It's very interesting that 40% of them responded that securely implementing Generative AI is their top priority, followed by 30% who said data loss prevention. Of course, after implementing these, you can move into the more exciting areas of leveraging embedded Generative AI and solving business problems.
As promised, here is the QR code for the microsite we launched. When you access this QR microsite, you'll find the four tracks I mentioned earlier. Each track has a guide that explains more about that track. For example, you can gain a better understanding of what embedded Generative AI is and examples of its use cases.
And for each partner, we worked with a third-party agency, interviewed the leaders of those partners, learned about their solutions, and based on that, created solution briefs. So you can go to that microsite, download the solution briefs, and if you wish, further engage with those partners to understand their solutions and adopt them.
So that's all. Thank you very much. And one more thing, some of our partners actually have booths here, so you can check out the Datadog and Dynatrace booths, Zscaler for data loss prevention, Elastic, Qlik, and Trellix for embedded Generative AI, and Salesforce and Snowflake booths for solving business problems. And actually, before I forget to mention it, our partners are also here, and we have partners from Elastic and Qlik, so please feel free to talk to them about their solutions and how they can help your organization.
That's all. Thank you very much. And we would be very grateful if you could fill out the session survey. Enjoy the rest of your time. Thank you.
This article was automatically generated using Amazon Bedrock, striving to retain as much information from the original video as possible.
Discussion