Improve your productivity with Amazon Q and Bedrock for SAP use cases

Introduction to generative AI

Generative artificial intelligence (generative AI) is a type of AI that can create new content and ideas, including conversations, stories, images, videos, and music. AI technologies attempt to mimic human intelligence in non-traditional computing tasks like image recognition, natural language processing (NLP), and translation.

It is supported with a very large Machine Learning (ML) models that are pre-trained with a vast amount of data. There are two types of models that we need to be aware of in generative AI, they are :

Foundation Models (FMs) are ML Models trained on a generalized and unlabelled data that learn patterns and relationships to predict the next item in a sequence.
Large Language Models (LLMs) are one of the sub-class of FMs that are trained to learn and predict the next words that will come next in the sequence.

McKinsey’s research estimated that generative AI could add the equivalent of 2.6 to 4.4 trillion USD annually across 63 use-cases. This is quite significant comparing to a GDP of the entire United Kingdom of 3.1 trillion USD.

Many SAP customers are asking how they can benefit from this new technology advancement to improve their business outcomes. Some of the questions being asked :

What are the generative AI use cases available for SAP customers?
What business values can generative AI provide for SAP customers?
What are the offerings available to modernize their SAP landscape with generative AI?
Where should they start on the generative AI journey?
Who can help them in their generative AI journey?
How can they be successful in the process of adopting generative AI in their SAP business processes?

Recently AWS and SAP expanded their strategic collaboration to transform cloud enterprise resource planning (ERP) experiences and help enterprises leverage generative AI. The initiatives includes :

The generative AI hub in SAP AI Core will integrate generative AI models from Amazon Bedrock, allowing SAP customers to access high-performing large language models and foundation models to build customized AI applications.
SAP plans to use AWS Trainium and Inferentia chips for training and deploying future SAP Business AI offerings, which can accelerate the AI model development process. In a proof of concept, SAP engineers trained and fine-tuned generative AI LLMs in 2 days versus 23 days with comparable Amazon EC2 instances.

This blog is written to address SAP customer’s questions and help them start their generative AI journey on AWS and get more value out of their SAP investments.

How generative AI can help SAP Customers

From our interactions across many customers, teams in AWS have identified common generative AI use cases that SAP customers are experimenting with and implementing. The list below is not exhaustive, and we encourage you to contact us with your business pain points to explore any other use cases.

No	Persona	Use Case	Business Benefits	Relevant AWS Services
Business use-cases
1	Sales Inventory Analyst	Gather real-time insights in the warehouse inventory to avoid out-of-stock and over-stock situation	Improve reliability of product availability and price stability for customers	Amazon Bedrock
2	Order-to-Cash Analyst	Create self-service reporting using generative Business Intelligence (BI) capability	Improve decision making to optimize Order-to-Cash process reducing processing time and process bottleneck	Amazon Q in QuickSight
3	Sales Account Manager	Gain customer insights by summarising and analysing sales documents	Enable 360 degree view of the customer’s orders and immediate needs, creating opportunities for upselling	Amazon Q Business or Amazon Bedrock
4	Procurement Manager	Review multiple responses and quotation against requests posted as RFPs (Request for Proposals)	Boost productivity during analysis of RFP responses thus facilitating procurement decision faster	Amazon Q Business
5	Master Data Manager	Generate product descriptions in multiple languages and images with generative AI	Improve productivity and accuracy of master data in SAP therefore avoid inefficiencies in global operation	Amazon Q Business or Amazon Bedrock
6	Executives (including VP, SVP, etc.)	Generate email with real-time insights from SAP reports for C-Suite	Accelerate decision making at the CxO and board level	Amazon Bedrock
7	Finance Audit manager	Gather real-time audit insight and anomalies based on company policy on various financial transactions gathered in SAP	Process audit compliance faster and avoid unnecessary escalation for occurring anomaly	Amazon Bedrock
8	Accounts Payable Manager	Reconcile invoices, detect anomalies from real-time insights in Procure-To-Pay process	Process payment faster, improve vendor relationship and address anomalies at speed	Amazon Bedrock
9	Legal or Policy Manager	Summarise insurance policy documents to understand insurance coverage	Ensure completeness and speed up claim processes improving customer and vendor relationship in the long run	Amazon Q Business or Amazon Bedrock
10	Employees	Understand company policy when it comes to procurement processes	Ensure compliance and accelerate fulfilment of procuring services and equipment	Amazon Q Business
11	Employees	Understand company policy when it comes to Employee benefits, coverage and claim processes	Improve productivity and employee’s satisfaction on benefits	Amazon Q Business
Technical use-cases
12	SAP ABAP Developer	Accelerate SAP projects by generating ABAP programs in Eclipse IDE	Improve developer’s productivity by up to 70%	Amazon Bedrock
13	SAP ABAP Developer	Render multilingual print forms for global deployment	Build print form once, and deploy globally to handle fixed text and SAP long text with context translation that is better with LLM	Amazon Bedrock
14	SAP ABAP Developer	Create technical specification and test scripts from the program without proper documentation	Improve sap consultants and developers on the accuracy of their specification and test scripts	Amazon Q Business
15	SAP System Administrator	Complete development tasks faster for system administrations	Accelerate productivity improvement and reduce mistakes on maintaining SAP Systems	Amazon Q Developer

AWS Generative AI Offerings

AWS provides a full spectrum of generative AI Services for SAP customers from pre-packaged generative AI Applications, building your own generative AI Applications, and developing your own Large Language Models (LLMs).

Generative AI Stack	AWS Services	Other Services	SAP context
Develop your own Large Language Models	AWS Trainium is the second generation of machine learning accelerator that AWS purpose built for deep learning training of 100B+ parameter models. While AWS Inferentia accelerators are designed by AWS to deliver high performance at the lowest cost in Amazon EC2 for your deep learning and generative AI inference Applications.	GPUs, Sagemaker, Ultra Clusters, EFA, EC2 Capacity Blocks, Neuron	In this most advanced scenario, SAP customers have built significant investment on in-house AI/ML capability, and would like to build their own LLMs as the pre-trained LLMs do not need their business domain.
Build your own generative AI Applications	Amazon Bedrock provides environment to build and scale generative AI applications with FMs. It is a fully managed service that offers a choice of high-performing FMs from leading AI companies. It also provide a broad set of capabilities around security, privacy and responsible AI. It also supports fine-tuning, Retrieval Augmented Generation (RAG) and build agents that execute tasks.	Guardrails, Agents, Fine-Tuning, Retrieval Augmented generation	In this context, SAP customers would like to leverage the pre-trained LLMs that are available in Bedrock such as Anthropic Claude, Amazon Titans, Stability AI and others. The LLMs offered are sufficiently meeting their use cases using common technique such as Prompt Engineering, Retrieval Augmented Generation and/or Fine Tuning.
Pre-packaged generative AI Applications	Amazon Q is a generative AI-powered assistant designed for work that can be tailored to your business by connecting it to company data, information and systems. Amazon Q in QuickSight is generative BI assistant that makes it easy for users to build and consume insight out of their business data using natural language prompts. This will accelerate building dashboard and stories, thus accelerating decision-making in the business processes. Amazon Q in AWS Supply Chain is generative AI Assistant that will provide aid in analysis and decision making to optimize the Supply Chain processes. Amazon Q in Connect provides Agents suggested responses and actions to address customer questions providing fast issue resolution and improved customer satisfaction. Agents will have access to knowledge articles, wikis and FAQs from various repositories in real-time conversation.	Amazon Q Developer	Majority of SAP customers probably will fall under this category, where they rely on pre-packaged generative AI to jumpstart their innovation journey. By performing several configuration tasks, they can leverage the existing Amazon Q to build their own generative AI Assistant to improve their employee’s productivity, or build their own generative BI capability to enable self-service reporting capabilities for their business users to accelerate decision making within the company. Amazon Q Developer can help to accelerate Software Development Lifecycle

Amazon Q vs Amazon Bedrock for SAP use cases

Let’s deep dive into Amazon Q and Bedrock offering that will cover majority of SAP use-cases in generative AI to accelerate your generative AI journey with pre-packaged offerings and full suite generative AI capabilities.
Amazon Q Business is a fully managed, generative-AI powered assistant that you can configure to answer questions, provide summaries, generate content, and complete tasks based on your enterprise data . It is embedded in various offering within AWS cloud services. We will discuss two of them here :

Amazon Q Business is great option for you to create your own digital generative AI assistant by pointing your business data sources such as company portals, SharePoint, Amazon S3 bucket, and others. With its ability to leverage these knowledge bases to answer any queries, this will be the lowest fruit hanging for your business to benefit from.
Amazon Q in QuickSight can jumpstart your generative BI journey by enabling your users to have a self-service reporting platform. Users can build their own visual dashboard and stories by using natural language queries without the need to understand the underlying SQL statements, Join mechanisms, and other technical jargons.
For AWS Supply Chain and Amazon Connect, they are fully managed service that is purpose built for certain business scenarios. When you subscribed to those services, Amazon Q is embedded as default enhancing productivity of your users in their execution of critical business processes.

Amazon Bedrock on the other hand, will offer your developer a fully managed machine learning service to build, deploy, and scale custom AI models and application. It is equipped with capability to evaluate, select and fine tune various Large Language Models. It provides a managed environment for model development and training, model hosting and deployment, monitoring and observability, security and compliance, scalability and cost-efficiency, as well as integration with the broader AWS ecosystem. Bedrock simplifies the process of creating AI-powered applications by handling the underlying infrastructure and operations, enabling developers to focus on model development and application logic.

What security standard does Amazon Q and Bedrock support ?

Amazon Q Business supports access control for your data so that users have access to the right content based on their permissions. You can integrate your Amazon Q Business web experience with your external SAML 2.0–supported identity provider (such as Okta, Azure AD, and Ping Identity) to manage user authentication and authorization.
Amazon Bedrock doesn’t use your prompts and continuations to train any AWS models or distribute them to third parties. Each model provider has an escrow account that they upload their models to. The Amazon Bedrock inference account has permissions to call these models, but the escrow accounts themselves don’t have outbound permissions to Amazon Bedrock accounts. Additionally, model providers don’t have access to Amazon Bedrock logs or access to customer prompts and continuations.
Amazon Bedrock doesn’t use your training data to train the Foundation Models or distribute to third parties. Other usage data, such as usage timestamps, logged account IDs, and other information logged by the service, is also not used to train the models.

How do I integrate Amazon Q and Bedrock to SAP ?

Taking into account the prior considerations, we will discuss several architecture guidance on how you can utilize Amazon Q and Bedrock to enable generative AI capabilities on your SAP Landscape. Please note that these are not the exhausted list of generative AI use-cases for SAP. The architecture guidance may vary based on your landscape and requirements.

First, consider Generative AI Hub in SAP BTP for SAP use-cases

In the blog Power your business with secure and scalable generative AI services from AWS and SAP, SAP AI Core service exposes AI assets such as large language models to customers and provides unified interfaces for SAP applications that run on the SAP BTP ecosystem. In this Joint Reference Architecture (JRA), we use the Generative AI hub in SAP AI Core as an access and lifecycle management layer to manage access to Amazon Bedrock and present an endpoint for our application to consume the foundational models. Through the Generative AI Hub, SAP centrally enforces numerous content filtering, SAP-specific risk mitigation, and safety guardrails, providing a compliant approach to safeguard against potential business and legal risks at scale across the SAP ecosystem.

Figure 0. Amazon Bedrock through SAP AI Core

Architecture guidance for SAP insight (example: use-case 1)

The image below describes capability for a Sales Manager to inquire inventory stock status directly from SAP Fiori Launchpad without creating and running a custom report. This can potentially save around 14 man-days from ABAP development and testing effort on a typical custom development effort.
Flow of use-case:

Query in English language : “Product highest inventory value at Chennai Warehouse”
The product with highest inventory value at Chennai Warehouse is Infrared Camera with total value of $1700”

Figure 1. SAP Insight

We utilize SAP Datasphere and SAP HANA Cloud with its capability to integrate with SAP S/4HANA. In this scenario, we capture the user’s natural language query in SAP Build Apps, which can trigger a REST API provided by API Gateway. The REST API itself will trigger Lambda function to trigger Retrieval Augmented Generation (RAG) that can be served by Amazon Bedrock, and augmented with the data coming from SAP HANA Cloud. This architecture can be enhanced to use SAP Generative AI Hub on AI Core to minimize latency and improve performance.

Figure 2. SAP Insight architecture diagram

Architecture guidance for self-service reporting for business users (example: use-case 2)

The image below describes how Q in QuickSight helps business users to create their own self-service reporting through visualization and stories. This removes the need for expert developers or analytics team involvement thus potentially reducing their effort of generating the various reports from days to just a few hours.
Flow of use-case:

Write the prompt in natural English to create story with target objective
Select the dashboard visualization to be included to complete the story
Click at Build
Title is automatically written by Q
Sections are written by Q creating story flow
Conclusion finally written to close with recommendation

Figure 3. Self-service reporting with Q in QuickSight

We utilize SAP Datasphere and SAP HANA Cloud with its capability to integrate with SAP S/4HANA. Athena can be configured to read from SAP HANA Cloud with its data federation capability. Business users can create visual dashboards and stories using natural language prompt in Q in QuickSight which obtain the data through Athena. The data source wizard for SAP will create the relevant Lambda function to support this activity.

Figure 4. Self-service reporting architecture guidance

Architecture guidance for generative AI Assistant for SAP (example: use-case 2, 3)

The image below describes how Q can help maintenance engineers to chat with their work instructions from SAP Work Order application. It eliminates the need for them to print hundreds of pages and browsing each of them to find certain information. This will improve productivity, accuracy and quality of work in plant maintenance.
Flow of use-case:

In the Maintenance Order, we attached the Work Instruction PDF that describe on how to repair a valve
Once it is uploaded to SAP, the PDF will be crawled and available for Amazon Q Business. User can query the document such as “How to repair the control valve if it has leakages ?”

Figure 5. Generative AI assistant for SAP

SAP ERP or S/4HANA contains both transactional (structured data) and documents (non-structured data).
The transactional data can be extracted using Amazon AppFlow into an Amazon S3 bucket. This can be loaded into Redshift, or you can also read this directly through Athena. Amazon Q in QuickSight will use either Redshift or Athena as data source to enable generative BI capability for self-service reporting using natural language prompt, enabling business users without SQL or programming knowledge to build their own dashboard and stories. See point 1,4,6.
The documents attached to SAP transactions can be extracted using AWS SDK for SAP ABAP or SAP Content Server to S3 buckets, for example: Invoices, Contracts, Quotations, etc. These non-structured data can then be crawled and index with Amazon Q Business, while protected by AWS IAM Identity Center for end-users to access.

Figure 6. Generative AI assistant for SAP architecture diagram

Description of steps as per above diagram:

You extract SAP reports, attachments, archive data to the Amazon S3 Bucket through the AWS SDK for SAP ABAP
Amazon Q will crawl the extracted data from Amazon S3 bucket (i.e. purchase orders, invoices, material master, equipment master, maintenance orders, work instructions, etc.)
Amazon Q will crawl from other data sources which contains project information, work logs, photos of job site, site safety observations, daily surveys, which can be stored in SharePoint, Jira, Zendesk and others.
Structured data and reports from S3 thru Athena can be loaded to Redshift which can then be consumed by Amazon Q in QuickSight to create Storyboard / Report that can be stored in S3
AWS IAM Identity Center can be integrated to authenticate users with Active Directory, Okta, and other Identity Providers
User will access Amazon Q and QuickSight through web browsers, which can be embedded to SAP front-end such as Fiori and others.

Architecture guidance for ABAP Assistant for SAP (example: use-case 12)

The image below describes how we can generate ABAP code and documentation with capability of Bedrock integrated to the Eclipse development tool. This application is estimated to boost developer productivity by 70% by facilitating faster ABAP code writing and aiding in ABAP documentation creation, thus accelerating Clean Core implementation in SAP.
With the SAP ABAP Assistant plugin, SAP developers can generate ABAP code and ABAP documentation from Eclipse IDE. The ABAP Assistant eclipse plugin is installed in eclipse IDE and runs locally in the developer’s personal computer.
Flow of use-case:

Write the prompt to create ABAP program based on the business requirement, then Invoke Amazon Bedrock
The ABAP program will be automatically generated from the prompt

Figure 7. ABAP assistant for SAP

Figure 8. ABAP assistant for SAP architecture diagram

Description of steps as per above diagram:

The ABAP developer uses Command Line Interface (AWS CLI) to authenticate with the AWS IAM Identity Center to get credentials which will be used by the plugin.
The ABAP developer adds the SAP system as an ABAP project in eclipse.
When the ABAP developer invokes the ABAP Assistant Plugin, it calls AWS Security Token Service (STS) to assume the configured AWS Identity and Access Management (IAM) role and generate short term credentials required to call Amazon Bedrock.
The ABAP Assistant plugin calls the Amazon Bedrock service to generate ABAP code and documentation and return the result back to eclipse.

Conclusion

We have discussed various generative AI SAP use cases and their potential business values. We have also discussed the available AWS service offerings that can help you to jumpstart and accelerate your innovation on your SAP landscape.
You can now unlock the power of generative AI for your SAP workloads. Amazon Q and Bedrock make it easier than ever to get started. Head over to the Amazon Q documentation, blogs and learn how to set up your own Q-powered generative AI Assistant. You will find step-by-step guides, sample configurations, and best practices to quickly deploy Q models that can assist with everything from summarising documents attached to SAP to generating recommendation to accelerate your decisions.
Start your self-service reporting for your business users using Amazon Q in QuickSight as your generative Business Intelligence assistant that will push the boundary of decision making speed for your business users by creating visual dashboards and stories easily.
Be sure to check out the Amazon Bedrock documentation and blogs. Bedrock provides a comprehensive foundation for building large language models and deploying them in productive scale. With Bedrock, you can fine-tune pre-trained models, host them securely, and integrate them into your SAP applications.
Try out the AWS Generative AI for SAP Workshop Studio to get your hands on with Bedrock, Q and QuickSight capabilities. This workshop is currently being revisited to include Generative AI Hub in SAP AI Core. Don’t forget to explore PartyRock to build AI-generated apps in a playground powered by Bedrock.
Lastly, start exploring the generative AI hub in SAP AI Core, which integrates generative AI models from Amazon Bedrock, allowing SAP customers to access high-performing large language models and foundation models to build customized AI applications.
Read more on AWS for SAP blogs to get inspiration on how you can get more out of your SAP investment. Get started with Amazon Q and Bedrock today and unlock the power of generative AI for your business!

Join the SAP on AWS Discussion

In addition to your customer account team and AWS Support channels, we have recently launched re:Post – A Reimagined Q&A Experience for the AWS Community. Our AWS for SAP Solution Architecture team regularly monitor the AWS for SAP topic for discussion and questions that could be answered to assist our customers and partners. If your question is not support-related, consider joining the discussion over at re:Post and adding to the community knowledge base.

Credits

I would like to thank the following team members for their contributions to this blog: Gyan Mishra, Adren D Souza, Derek Ewell, Beth Sharp and Spencer Martenson.

AWS for SAP