The SAP Intelligence event for you (TAIEFY) 🔗 was hosted from 9 to 10 March in Amsterdam at Booking.com headquarters. The first day of the event focused on the general use of AI in SAP scenarios — meaning how AI can be used by SAP customers and developers. How it can enable them to perform better and achieve their goals. The sessions were also streamed live and are available on the ABAPConf YouTube channel 🔗.
Sessions
The overall idea of day one was to give the attendees an overview of AI use cases in SAP and how SAP customers and partners and their developers can benefit from AI. Several partners with a strong AI foundation were present: AWS, Google, Microsoft. SAP was also given the chance to present how AI can add value to SAP centric scenarios. The live streom of day 1 🔗 contains all talks and a round table discussion.1
Amazon
The first talk of the day was from Amazon / AWS 🔗. The several tools available to bring AI to SAP were shown. How you can connect them to an AI/LLM, use MCP servers, as well as an ABAP plugin for ADT for Eclipse. Using Amazon Q developer you can easily bring enhanced AI coding best practices to SAP. I confess I am not familar with Amazons Q developer for SAP 🔗. But that’s what these kind of events are for: to learn something new.
The features presented show that it offers mature and useful features for SAP developers. For instance, the ABAP Accelerator 🔗 helps developers dealing with real world challenges like clean core by analyzing relevant code and modernizing existing coding. The example for the code analysis demonstrated this perfectly. The tool finds issues and recommends actions like bug fixing or performance improvements. It helps in transforming a report to clean core over to RAP and to a Fiori app. Interesting would be to see if the tool can first create tests for the productive report and use those to validate the converted app.
The Kiro CLI part of the talk was focused on modern development 🔗. This still might be too much of science fiction for most SAP developers. Yet be reminded that this kind of workflow is common in the non-SAP world. Agents connect to the code base, analyze and run task on it 🔗. You can execute this as part of a pipeline and the developers get the results delivered when done. The tools can run locally and no need to add BTP. A cool feature is how the different tools work together. The cli runs the agents. The agents connect to SAP and gather the needed information. Several features of MCP servers combined and working nicely together.
Microsoft
Next up was the session from Microsoft 🔗. One phrase summed up the whole session: AI in the context of SAP from a customer perspective.
The session started with how to make Joule and CoPilot work together. The session focued on the integration and how to bring Joule to the Microsoft world: CoPilot, Teams, Office. In my opinion, this cannot be stressed enough: end users will not use Joule, but CoPilot. The question for customers, and for SAP is very simple: how to bring Joule to CoPilot. The leader at customers is Microsoft. My office Laptop’s keyboard comes with a CoPilot button. Pressing the button opens CoPilot and not Joule. The race to be as close as possible to the end user: Microsoft won it already. And it was never a competition.
Important is to note that the user is working with Microsoft. Joule is “just” an agent that connects to SAP systems. Of course, this setup is a success at customers. Users work with data and information in CoPilot, not with transactions and SAP. Excel is still the final boss for every SAP transaction, and the ease of accessing corporate information stored in SAP via CoPilot: this is the new Excel.
Microsoft’s main SAP system is still on ECC. Here, the talk shifted from partner to customer—and to the pain customers face when migrating large SAP systems with millions of lines of code that run their core ERP processes. The talk also covered integrations, which are at the heart of Microsoft’s ERP processes, and the challenge of connecting on-premises systems to cloud solutions in a way that’s easy for end users. And for developers, the question was how to help them be more productive. Microsoft’s answer: ADO.
All ABAP code is in Azure DevOps 🔗. ABAP code lives outside an SAP sytem and this enables Microsoft to do modern development: git, continuous integration, pipelines. This allows Microsoft to work on code, not depending on an SAP system. Microsoft is living modern development with SAP/ABAP: code assistance, documentation, code reviews, AI/LLM, automatically fix issues directly on the code. Remind: this works on ABAP code. Not on an SAP system. Microsoft can do this on ECC, or S/4HANA system.
Users can interact directly with their SAP system using one of Microsoft CoPilot or low code tools. Seeing an end user “talk” directly to SAP and get answers is a nice demo. Sure, not as efficient as seeing all needed information in a table. But be reminded: not every user is a power user. Some just want to get one single piece of information. OData is the glue that makes the magic happen. It is the layer that allows you to connect to any SAP system the same way. S/4HANA or the cloud is purely optional. This kind of magic works nicely with ECC and on-premises systems too.
MCP was also part of the session. Just like OData, the MCP protocoal is an enabler 🔗. The famous OData MCP bridge was also shown. And again: this is something that enables all SAP customers to use AI, not just the one that a already in the cloud and RISE customers. Of couse, this works also smoothly together with CoPilot. Combining MCP servers is an offering end users want: easy access to SAP and make use of the data. Microsoft is even going a little bit further and is using MCP servers to enable their users already today to benefit from AI usage, even when their SAP systems are not ready yet 🔗. At the same time, their SAP data access is preprared for the new S/4HANA world. When their SAP systems are ready, they can switch to Joule.
An important part of the Microsoft presentation is that it always gets back to the end user. End users are using AI to interact with SAP. As soon as the problematic SAP layer is abstracted away via e.g. MCP, users start using SAP as a source for information.
Google itself is an SAP customer and put the focus of the talk also on customer scenarios and end users. As you can imagine, the volume of business transactions processed by SAP at Google is off the scale. Google strives to automate as much as possible. Trust into the systems is crucial, of course. Google opted to make use of … IT. AI and Agents, to be more specific. Agents start doing the job of people, and people transform to AI (agents) managers.
Google is also benefiting from offering value to end users. Agents offer features that make (work) life easier. Focus is on benefit. Agents do the boring job and tasks are automated. Next step is then to go for Agent 2 Agent. Let as much as possible be handled by agents, add humans only when needed. Basically, you have your ERP system, then comes the agents layer, the orchestration, and on top: users. Also here, users are working with agents, not necessarily a specific SAP system. Just like with AWS and Microsoft, it is: access to data, to information. Not access to a specific SAP transaction.
Using AI with ABAP is easy. You can take the ABAP SDK from Google and connect it to AI and orchestrate them. The output can be exposed to the Fiori Launchpad. Meaning once again: the user is using an UX to interact with SAP, without needing to know the specific details of SAP. The user works with an App, which interacts with an AI agent, that takes care of the details. To enable all of this, the underlying platform needs to offer some features, and NetWeaver does this. It comes with the data, memory, performance, interfaces.
Google showed that a small team can enable an agentic ERP. Automation is then the new normal and processes are run without humans slowing it down. Trust the AI and the role of humans is transformed and they can deal with the important tasks.
SAP
The session started 🔗 with explaining what tools SAP offers in the cloud for AI. With Business AI customers can get access to a wide range of LLMs and you can orchestrate your agents and develop AI apps with an SDK. RPT-1 was also briefly shown. Another demo was used to get into some details of LLMs and how they work. SAP gave a high level introduction to AI. Some nice information on how an AI works, processes data and why the result you get can be plain wrong.
While the session started promising, SAP decided to not show anything business related, but to go for fun stuff. The demos were not connected at all to a customer scenario. Nothing about how SAP is using AI to run their ERP. Nothing about how SAP employees are using MCP servers. While the other companies - Amazon/AWS, Microsoft, Google - showed how their AI offers value to them, as an SAP customer, as well to their customers as a partner, SAP seems not to have business cases to present. No internal business case, not one from a customer or a partner. Nothing about how SAP wants to deal with agentic ERP. Nothing that is important to customers was discussed.
I was wondering what SAP’s CEO announced in his e-mail regarding a shift in AI for SAP. Seems that SAP is abandoning customers. The first “fun” demo could easily work with a real example of a busines form: enter data, validate it, etc. But there is not SAP customer that uses this? Not even at SAP someone is using this? Or anything AI related?
Panel
Next up was the panel discussion 🔗. I am not a friend of a panel discussion at conferences. This panel won’t convert me into a fan. The questions were good. Definitely something that needs to be answered. Just: you won’t get reliable, data-backed, and official answers from the panelists. This is over their paycheck. A better format might have been a keynote, or a roadmap like sessions where the questions are answered. Because, yes, everyone wants to know: how do you find a business case?2 How do you meassure success? How to get people onboarded? How to enable secure access to an SAP system and try things out? Important questions, no doubt. But the format is not supportive. One panelist answering the question while the other 4 are waiting and looked bored into the distance. After all 5 panelists had time to answer the question, there is a certain chance you forgot the answer and the first two answers. Nevertheless, watch the panel, the questions were superb. And it’s a recoding, you can go back and forth in the video all the time.
Wrap-up
SAP does not care about AI for on-premise customers
The statement of the day for me was from AWS.
SAP tools for AI require cloud or RISE. For the process of transitioning SAP customers to the cloud, these are not the best. You can use the tools from AWS in the old world, to move to the new world. And that’s a market SAP is not interested in. So there is also no competition between partners offering tools for SAP customers that are not (yet) in the cloud. SAP has a weak offering for ECC, so there is no competition and - hold your breath! - SAP is fine with this. Regarding AI: SAP is leaving the market of non-cloud SAP customers to partners.
The message from SAP after the demos can be summed up as: customers should use any AI tool from Google, AWS or Microsoft. Just do not use SAP when you have to work on real world challenges.
Agentic ERP
The same message was transported also in the Microsoft presentation. Together with Google, the new reality that is starting to take over how end users work with SAP is by using an Agentic ERP. Both show that this is already a reality.
The fun part of such Agentic ERP is: the agents do most of the work. The boring, automated work and humans are dealing with special cases. As an ERP like the one from SAP is based on standardization, agents can be easily trained and created to work with standard processes. If one day a company wants to switch to another ERP that is using standard processes: they can. Adjusting the agents will be an easy task and most processes will continue to work. Or be rapidly adjusted. The only part that will need some time and addtional effort to adjust the new ERP will be the highly customized and human interaction heavy processes. But those are only a small % of the overall processes.
MCP servers have the potential to change more how users are working with an ERP system than we might imagine today. They are not just a nice tool for developers. It is still too early to roll this out at every company. But AI is a fast paced innovation. What today might only work for a few, next year might work for all. When the AI understands how SAP or ERP works, how sales orders, documents, business partners, financial data is processed, and can find the needed information with the help of MCP, APIs and other data, it will used. No, it will be demanded by users. And users here means: the one that work with an ERP system, the ones that pay - or be paid - for working with an ERP system. The possibility to switch later on to another ERP system is not only given for between SAP releases like ECC to S/4HANA. This can work for switching from SAP FI to Workday, from Oracle Applications to Infor, from SAP to SalesForce. And the user is continuing to use a PowerApp or Office without having to worry too much about the underlying data source.
SAP reduced to a system of records
In architecture, the worst thing that can happen to an ERP system is to reduce it to a system of records. An ERP system is doing so much more than just storing data. Yet, if SAP is not taking care, they will be reduced to a system of record. An expensive system of record.