Sponsored by: SAP | SAP and Databricks Turn SAP Data Into AI-Driven Business Outcomes

Summary

SAP Business Data Cloud and the Databricks Data and AI platform deliver a unified governed data fabric architecture that combines SAP's business context across finance, HR, and procurement with lakehouse analytics, reducing data movement, delays, and silos.
The integration uses Delta Sharing for bidirectional data exchange between SAP Business Data Cloud and the Databricks Data and AI platform, providing intelligent compute on a shared lakehouse storage layer without requiring ETL copies.
Real-world use cases demonstrated include cash flow forecasting at 95–97% accuracy for supplier payment timing, tariff impact and supply chain risk modeling, AI-generated financial narratives for FP&A teams, and accounts payable automation for invoice processing.

Chapters

FAQs

What is SAP Business Data Cloud and how does it connect to Databricks?

SAP Business Data Cloud is SAP's data fabric architecture that brings together both SAP and non-SAP data in a shared lakehouse storage layer with intelligent compute on top. It connects to the Databricks Data and AI platform through Delta Sharing, enabling live bidirectional data exchange without copying or moving data between systems.

What is SAP's autonomous enterprise vision?

SAP's autonomous enterprise is a vision for a new operating model where humans and AI interact together to run core business processes across three layers: an engagement layer with AI assistants and agents, an autonomous suite handling core enterprise operations, and a Business AI Platform providing the data and governance foundation. SAP Business Data Cloud is the data fabric component at the heart of the Business AI Platform.

What AI use cases are shown with the SAP and Databricks integration?

This video covers cash flow forecasting at 95–97% accuracy for predicting supplier payment timing, tariff impact and supply chain risk modeling, AI-generated financial narratives that explain profit and loss movements for FP&A teams, accounts payable automation for invoice processing, energy trading revenue optimization, and trade promotion optimization using combined SAP and Databricks AI.

How does the combined platform reduce data delays and silos?

By using Delta Sharing for bidirectional data exchange, SAP Business Data Cloud and the Databricks Data and AI platform allow analytics and AI workloads to run directly against a shared data foundation without requiring data copies or separate ETL pipelines. This eliminates the latency and inconsistency that occur when SAP business data must be extracted and transformed before analytics and AI can run.

Full transcript

[00:07] Welcome to the session sponsored by SAP. A quick introduction about myself. My name is Ramaswamy Srinivasan. I'm the global head of strategic solution advisory within SAP. And in this session we're just going to
[00:19] talk about how the partnership between SAP and Databricks delivers a unified governed data fabric architecture which enables multiple AI use cases. Yeah?
[00:34] Quick overview about SAP. I'm sure most of you know about SAP. We are one of the world's largest providers of enterprise applications and business business AI software. We have more than 400,000 customers
[00:46] and in more than 180 countries and then we have experience with over 50 years experience in 26 different industries and of course
[00:58] one of the core functions of SAP is that we automate the core functions, core enterprise functions including finance, HR, procurement, etc. Right?
[01:11] I'm sure some of you would have attended Sapphire if where we launched our autonomous enterprise. You know, that's our vision of our you know, vision for the future
[01:23] where our business runs where you know, it's a new operating model where humans and AI interact together. And I'll just take a few minutes to quickly explain the different levels or components of
[01:36] what what this strategy is and vision is all about. You can break it into three major components. You have Jewel which is our engagement layer which is where you would interact with the business
[01:48] applications and you have a system of agents and assistants at the top which is your interaction or engagement layer. That brings together the data, the workflow, and the agents
[02:01] um uh with from SAP and non-SAP systems, right? And then you have the autonomous um uh suite um and the industry AI layer, which is the um core of our uh core of
[02:13] our enterprise um right. And then um which is where we bring together the AI data and processes. Now, in the last part of the uh autonomous enterprise as we do is the business AI platform, which
[02:26] delivers the business context, which is where you have you can build your own AI applications and agents and also govern them. Right? Uh the The business data cloud is part of the business data uh business AI platform,
[02:39] and that it's a it's at the heart of the um business AI platform. Now, the business uh the business data cloud is uh is our data fabric fabric architecture,
[02:52] uh which is where you can bring both SAP and non- uh SAP data together in the lakehouse storage, right? And then you have the intelligent compute on top, which is where we have our own in-memory database um like the Hana cloud. Um and
[03:05] you could also use data lake um solutions like Databricks or uh Snowflake in the intelligent compute layer. On top of it is our own knowledge core, which is where we had we have our own SAP Data Sphere, which is our um
[03:18] data modeling and analytical modeling solution. And then you also have the analytics cloud, which is our BI and planning offering. Right? And our strategy is that, you know, we share the data once
[03:31] um um at the at the object store level and then use it not only for third-party uh purposes for extracting it out and using it uh by by third-party partners like Databricks and Snowflake, but also we
[03:43] have built our own intelligent content and applications on top of it. Uh for example, you have the core enterprise analytics or the people analytics applications, which is our out-of-the-box uh
[03:55] intelligent applications and business content and analytics on top of this data. Right? >>
[04:08] the different layers of the business data cloud, at the heart of it is the data products, right? Which I mentioned before, which consists of both the SAP managed data products, which is what is different compared to what
[04:21] customers were doing before. It's you know, in the past customers used to extract the data out of SAP systems, you know, replicate the data into third-party data data lakes, for example, and then
[04:35] rebuild all the business context and contact and the relationships, right? Now, with BDC, our core strategy is that we do the extraction of the data from our core enterprise applications or from
[04:49] you know, SuccessFactors or Ariba, etc., and then put it in the SAP managed object store, and then make it available not only for third-party partners like Databricks and Snowflake, but also use it for our own
[05:03] you know, intelligent applications, right? So, the idea is you share the data once, you know, and it's curated data sets. That is the data products, which you share it once and it can be used both by SAP as well as
[05:16] partner solutions. Right? >>
[05:28] session also, which my colleague posted with together with Adi Das, right? We have We have As we said, the data is stored in the object store level, and then we provide we are supporting the
[05:41] Delta Share format, so we do the Delta Sharing once, so that it's a bidirectional Delta Share with Databricks. That is That is the way we work. So, the data is stored at the object store level and then shared
[05:53] bi-directionally to Databricks, and then you can do your all your AI/ML workloads for analytical use cases and bring the data back into SAC BDC. >>
[06:19] great example, you know, where they're using SAC Databricks to um um you know, leverage the SAC data and then power their own
[06:32] use case. In this case, you know, they're able to better leverage the SAC data and then understand their and serve their customers better. So, how do you go about doing that, you know? Typically, what happens is, as I
[06:45] said, we have a set of curated data products, which is then you can Delta Share those data products to Databricks, for example, and then you install these you integrate
[06:59] you know, probably bring in your third-party data, and then you blend them together, and then you can do do your, for example, AI/ML use cases, power the AI/ML use cases, and then
[07:11] you can also train the models based and then deploy the trained models as you know, endpoints and bring it back into SAC and BDC, and then
[07:23] build your own analytics or dashboards, or even power the transactions further. >>
[07:35] LOBs. For example, you know, cash flow and liquidity analysis, you know, variance analysis is another example that is very common in the finance domain. We also have like you know supplier
[07:47] credit risk assessment where you can blend some for example external ESG data you know about suppliers and blend that with your supplier information coming in from Ariba and then power your own use cases. So
[08:01] there are multiple examples. I'm just going to take the next few minutes to go through some of the other use cases which you know across different domains and also different industries.
[08:17] So one of the industry one of the use case that is common across multiple industries is this anomaly detection in financial statements like I'm not sure if any of you attended some of our partners like Accenture also had some sessions around how around the
[08:29] finance area how to power some of these use cases. So this is one common use case you would find you know which is applicable to pretty much every customer across industry domains, right? A lot of times
[08:43] you know um um the auditors go through the accounting data to detect any anomalies in the general ledger entries, right? Um similar to it there is also another
[08:55] relevant use case which is to uh do a variance analysis or you know movement analysis as they call it. If you compare to the previous period if there are you know what is to be able to
[09:08] explain the movement in the balance sheet or profit and loss statements etc. and then go into the details of what is what is the reason for any variance or movement in the different financial statements. That's
[09:21] something which is a common use case, right? Now if you take this use case you get the core data products from SAP S/4 HANA and then like you know general ledger accounting entries
[09:34] you get the data product from there or you could also get it from if you're done in an on-prem world and you have BW, you could get that data from SAP systems either the standard data product or a custom data product and then
[09:46] you could use a database you know to build your own AI model to do this anomaly detection, right? And then what you could also do is once you've done your model building and then training and
[09:58] then you're ready to deploy it, you can push that data back to SAP Data Bricks and then use it for further analysis and planning. That is how you for example you could use
[10:10] this partnership leverage this partnership between SAP and Data Bricks. So the difference is in the past you would have to replicate all this data, extract it out, replicate into Data Bricks and then perform all
[10:22] these EIML scenarios, but in this with the new architecture with BDC what happens is that the data you just share it only once I you know through SAP data products which is stored in the object store and then
[10:34] you can make it available securely to Data Bricks using the Delta Share. Right? Now Yeah, this is another interesting use case which my team was involved in with
[10:47] a huge oil and gas major you know in Asia. Now as you know in the downstream industry in the downstream oil and gas industry you know they have a huge challenge with making
[10:59] you know payables you know payments to their suppliers, right? Now this customer all their they had they were using the IS oil and gas application. They were also using the vendor invoice management
[11:11] module. So any
[11:24] ready for payment, right? Now these runs into hundreds of millions of dollars. So today the finance department and the treasury team, they were actually um you know
[11:36] not able to forecast in a shorter time frame. So, they used to keep the funds ready for whatever of the invoices are likely to be paid in the next 10 days. Now, with with this specific customer, what we did was we just used the
[11:51] information vendor in vendor invoice management information from SAP. And then we got that information along with the payable supplier information etc. into the SAP databricks
[12:05] via delta sharing. And then we built a machine learning model to predict the most likelihood of these invoices going through the process of what are all the various steps an invoice has to go through before it is ready for payment.
[12:18] We started classifying them into buckets of what's most likely due within the next 3 days versus the next 10 days and 10 days and beyond. So, what this facilitated with the customer was
[12:30] to better manage their cash flow so that they wouldn't have to you know, draw overdrafts or borrow money to pay for or budget for these payments. So,
[12:42] they could better manage their working capital by ensuring that you know, whatever is most likely due in the next 3 days is what they would keep money in the bank for. The accuracy of this forecast was
[12:54] approximately 95 to 97% is what they could get. 95 97% accuracy as what is the most likely um amount you know, that's that needs to be
[13:06] kept ready for making these supplier payments. This is another example. Again, all the data in this case is coming from SAP you know, S4 and ISU utilities, but you could also get for
[13:19] example information from third-party sources. Now, these are use cases where SAP databricks makes a lot more sense, right? There are also other cases where you might also use BDC Connect and then,
[13:33] you know, leverage the enterprise data bricks, too. This is another example which I've got in the oil and gas industry, but this is more on the upstream side of things, right? These are, as you can see, very niche
[13:45] use cases. In this case, this is oil and gas major. You know, they have something called as I don't know how many of you are aware of this crude optionality. You know, when there is a physical loading of a trading of,
[13:58] you know, oil, what happens is that all these contracts have an optionality. There is a plus minus 5% or 10% quantity that they can load
[14:10] ships. Now, what many of these suppliers or, you know, sellers tend to do is to maximize the profit. What they do is they track the movement of the
[14:22] oil prices, right? If they see a potential for the oil prices going up, then they would do that minus 5% or minus 10%. If they see that, you know, the oil prices are going to go down, then they would like to maximize the
[14:35] price and then try to, you know, load as much quantity as possible into the bulk carrier. So, this is an example where we took all the contract data and the tolerance rules, the trade positions, etc. from SAP.
[14:49] And then, they also built a model to predict what would be the price of oil going forward in the next few days. And then, normally it's 3-4 days time horizon. That's what we are talking of when they have to load these
[15:02] oil into into these ships. And this is an example where we worked with Databricks to extract this data from SAP, the contract data, the tolerance rules, etc. Build a model to predict the oil
[15:15] prices and then help the guys take a call on what would be the best strategy to, you know, when it comes to the oil prices and trade positions to maximize their um
[15:28] uh revenue and price which they get per barrel. Right? So, these are some examples. We also have a lot of examples in other industry. In retail is another area where
[15:40] there is you know customers tend to you know blend SAP and non-SAP data also. Some of the use cases are like returns forecasting or assortment planning or trade promotion planning.
[15:53] These are some of the common examples which you would find in the retail industry, right? So, >>
[16:06] be able to in order to better predict the returns, you should also get the actual returns data, get some external data like holiday calendars or weather information or economic metrics, Google Trends, etc.
[16:19] whatever, right? And then you should be able to predict the returns in a much better way. That's another use case which is very relevant in the retail industry. So, that's something which is another good
[16:32] example. We also have as I said assortment planning, you know, what sizes, what you know, colors, etc. of assortments of
[16:44] garments for example you would like to do. That's another use case which can be powered with the joint partnership with Data Bricks. Um
[16:58] another good good example of our use case would be the promotion planning. Trade promotion planning is something which many of the you know, retail and CPG customers tend to do. Now, again a similar this thing
[17:10] you get the promotion information you you know, you get the sales um uh actual information, the retail promotion, the merchandising category, etc. All this information which could come from SAP, SAP CAR, which is our
[17:22] customer activity repository, and also from Ice Retail, get that information, and then you combine it with third-party information like campaign, competitive pricing, demographics,
[17:34] you know, all this information, and then you could do your own trade promotion planning using the SAP Databricks joint partnership and solution. Yeah?
[17:46] I'll just take a pause here. We're just running out of time. So, in summary, uh, what I wanted to say was it's more than a year since we have this partnership with Databricks. We are seeing a huge uptake in, you know, joint
[18:00] customers, you know, activating SAP Databricks, and also using BDC connect to Delta share SAP data products with Databricks, and um,
[18:12] uh, this is our strategy, as I said, you know, the whole objective is not to keep replicating the data. Get the data. We will We have more than 300 plus data products available as of today, but the plan is
[18:24] there are more coming out. You know, today we have for S/4, we have data products out of the box SAP data products for, HR. We have people intelligence and enterprise core analytics, but the road
[18:36] map has a lot more coming down the year including the, you know, spend related data products and intelligence and applications as well as Concur, and all these applications will start providing more
[18:49] and more data products. As the data products economy matures, you know, the whole objective is that we are expecting that customers would not be doing the replicating and having multiple copies of the data, but the
[19:01] whole value proposition is that we push the data, we take the responsibility, we offer it as a SaaS solution, and then you can share it. I also met with a few to in the booth who were asking that they were also having both Snowflake and
[19:14] Databricks for example or have this is a practical scenario which we see at many customers have Databricks as well as you know say Microsoft Azure or AWS or Google, right? And we have announced a
[19:26] partnership with the all the three hyperscalers too with Google, Microsoft and AWS. So the whole idea is that we share the we store the data once in the object store and you could use VDC connectors
[19:39] use provide access to both Databricks and Snowflake to the same set of data. That's that's our strategy. Okay? All right. I'll just take a pause here. Thank you very much. Please um um uh ensure that you fill the surveys.
[19:53] That's the last thing on my list thing. Thank you.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.