This wave of Generative AI (GenAI) started from the November 2022 with OpenAI ChatGPT first released into the market and hit one million users in 5 days. Now, with SAP Joule(Joule)'s release in September this year to join the GenAI army, you can simply ask a question or explain problems to Joule in plain language and get answers in the SAP context. Joule generates these answers from drawing a vast amount of business data across the SAP portfolio and external sources. "It’s like tapping your smartest colleague on the shoulder".
However, from last November 2022 to September 2023, it seems SAP is behind the curve, so, why does Joule matter? And with previous attempts (still remember SAP Leonardo?), why it is a different attempt this time?
First of all, let me explain the SAP "Business AI" concept: it's not simply putting a word "Business" in front of the buzz word "AI": it's about how AI can achieve the deep understanding of business models and processes, and keep its behaviour with the boundary of AI Ethics. The magnitude of the problem can be understood through the comparison between "Business Applications" (e.g., SAP offerings) to achieve goals for end-to-end businesses, versus the individual "apps" to fulfil individual goals. The complexity sits in the motion that business applications coordinate interactions between many people and processes, while at the same time maintain the business trust under legal and data privacy regulations by industry or by country. By saying that, the value lies in the years of the industry business practices being built into software components by the software/service vendors, which makes it challenging and time-consuming to select the proper Large Language Models (LLMs) and train the model to acquire that knowledge and experiences.
Historically SAP has used different mechanisms to build business intelligence into applications:
But again, why Joule matters? Probably we should revise the question: why Joule matters NOW? I think this is the tipping point. As you may have noticed, only until GenAI appears, can AI generate meaningful new information. That's the fundamental difference and we will see exponential increase in the business value AI can generate. Think of Apple's early Personal Digital Assistant (PDA) attempt "Newton", advanced in vision and concept as it was, it failed as a product due to the technology limitations and market readiness. Decades later, when the timing was right, Apple's attempt with iPhone (and iPad) picked up the torch from Newton and made the history. Similarly with people's familiarity with consumer grade personal assistant like Amazon Alexa or Apple's Siri, now, SAP Joule's debut becomes natural in the business context with leveraging the experiences from SAP Leonardo and other copilot products which have become part of the technology platform or have evolved with the new AI capabilities.
It's also worthwhile to mention SAP's Core AI Principles: Relevant, Reliable and Responsible.
Relevant: the interactions will be based on customer's specific business data, combining the industry best practices which are embedded in the GenAI model behind Joule. The human/software interaction of manually clicking through many screens to find the relevant number, figure or data can be dramatically simplified through the conversation with Joule. Moreover, new business innovations can possibly be built into BTP with just conversations with Joule to coordinate with the SAP solutions. Imagine by describing the problem in natural language then extension codes can be generated, or an application that understands how to integrate and configure itself in diverse contexts.
Reliable: one of the debates on GenAI is the data quality. SAP systems and data are frequently regarded as the "single source of truth" in business which provides a clean source for the GenAI models. SAP also keeps training the models with huge amount of real data with privacy and consents in place. Moreover, Joule's data access is governed by the counterpart user's authorisations which means there won't be data leaking through the conversation.
Responsible: how to make Business AI trustworthy is a crucial question for all business AI use cases, hence the ethics, data protection and privacy, security and explainability are built in all the AI enabled applications and services, also within Joule. This is not only a differentiator in the market but a manifesto to SAP's core beliefs and values that cannot be compromised.
There's a good saying: don't try to use AI as a new way to solve old problems, use it as a way to solve new problems. Computer systems tend to be specialised in areas where "business things" can be accelerated (efficiency) but lacks the capability of recognising patterns then generate new recommendations and business make decisions ("decision augmentation"). One example can be: existing business applications are mature enough to handle almost all the processes a retail company runs but cannot discuss with a user in natural language about where and when to open another store. Now, Business AI application like Joule just enabled that possibility.
As Jensen Huang (NVidia CEO) said the ChatGPT is the "iPhone moment" for AI, I hope this time Joule will be the "iPhone Moment" for SAP Business AI and open a new era for AI-powered business solutions.
However, from last November 2022 to September 2023, it seems SAP is behind the curve, so, why does Joule matter? And with previous attempts (still remember SAP Leonardo?), why it is a different attempt this time?
First of all, let me explain the SAP "Business AI" concept: it's not simply putting a word "Business" in front of the buzz word "AI": it's about how AI can achieve the deep understanding of business models and processes, and keep its behaviour with the boundary of AI Ethics. The magnitude of the problem can be understood through the comparison between "Business Applications" (e.g., SAP offerings) to achieve goals for end-to-end businesses, versus the individual "apps" to fulfil individual goals. The complexity sits in the motion that business applications coordinate interactions between many people and processes, while at the same time maintain the business trust under legal and data privacy regulations by industry or by country. By saying that, the value lies in the years of the industry business practices being built into software components by the software/service vendors, which makes it challenging and time-consuming to select the proper Large Language Models (LLMs) and train the model to acquire that knowledge and experiences.
SAP Business AI Approach (Source: SAP Business AI White Paper)
Historically SAP has used different mechanisms to build business intelligence into applications:
R/1 to R/3 until the 1990s: deterministic programming, rule engines.
Business Suite in the 2000s: predictive analytics and forecasting.
HANA in-memory capabilities in the 2010s: enable the specific analytics and intelligent use cases over large datasets.
Intelligent Enterprise in the 2020s: Foundational Models, LLM and Large Process Models (LPM). GenAI capabilities through direct investments into the AI companies like Aleph Alpha, Anthropic and Cohere. There are more than 24,000 SAP cloud customers already actively use 130+ AI capabilities in the SAP applications and Business Technology Platform (BTP).- In the mid-to long-term, SAP is exploring differentiating capabilities to leverage the assets and to address limitations of LLMs.
But again, why Joule matters? Probably we should revise the question: why Joule matters NOW? I think this is the tipping point. As you may have noticed, only until GenAI appears, can AI generate meaningful new information. That's the fundamental difference and we will see exponential increase in the business value AI can generate. Think of Apple's early Personal Digital Assistant (PDA) attempt "Newton", advanced in vision and concept as it was, it failed as a product due to the technology limitations and market readiness. Decades later, when the timing was right, Apple's attempt with iPhone (and iPad) picked up the torch from Newton and made the history. Similarly with people's familiarity with consumer grade personal assistant like Amazon Alexa or Apple's Siri, now, SAP Joule's debut becomes natural in the business context with leveraging the experiences from SAP Leonardo and other copilot products which have become part of the technology platform or have evolved with the new AI capabilities.
AI Impact on Business Value (image credit: SAP)
It's also worthwhile to mention SAP's Core AI Principles: Relevant, Reliable and Responsible.
Relevant: the interactions will be based on customer's specific business data, combining the industry best practices which are embedded in the GenAI model behind Joule. The human/software interaction of manually clicking through many screens to find the relevant number, figure or data can be dramatically simplified through the conversation with Joule. Moreover, new business innovations can possibly be built into BTP with just conversations with Joule to coordinate with the SAP solutions. Imagine by describing the problem in natural language then extension codes can be generated, or an application that understands how to integrate and configure itself in diverse contexts.
Reliable: one of the debates on GenAI is the data quality. SAP systems and data are frequently regarded as the "single source of truth" in business which provides a clean source for the GenAI models. SAP also keeps training the models with huge amount of real data with privacy and consents in place. Moreover, Joule's data access is governed by the counterpart user's authorisations which means there won't be data leaking through the conversation.
Responsible: how to make Business AI trustworthy is a crucial question for all business AI use cases, hence the ethics, data protection and privacy, security and explainability are built in all the AI enabled applications and services, also within Joule. This is not only a differentiator in the market but a manifesto to SAP's core beliefs and values that cannot be compromised.
There's a good saying: don't try to use AI as a new way to solve old problems, use it as a way to solve new problems. Computer systems tend to be specialised in areas where "business things" can be accelerated (efficiency) but lacks the capability of recognising patterns then generate new recommendations and business make decisions ("decision augmentation"). One example can be: existing business applications are mature enough to handle almost all the processes a retail company runs but cannot discuss with a user in natural language about where and when to open another store. Now, Business AI application like Joule just enabled that possibility.
As Jensen Huang (NVidia CEO) said the ChatGPT is the "iPhone moment" for AI, I hope this time Joule will be the "iPhone Moment" for SAP Business AI and open a new era for AI-powered business solutions.
- SAP Managed Tags:
