Generative AI Products Must Optimize For More Than Productivity. Introducing SAP’s Joule.
Vin Vashishta
NVIDIA, Salesforce, & SAP AI Thought Leader | Wiley Best Selling Author | Gartner Ambassador | CEO V Squared – Over $3.7B In Value Delivered.
Over the next few months, I’ll show you what was an oxymoron until recently: real-world Generative AI. After 8 months of vaporware, Generative AI products are finally being delivered. Generative Interfaces are one of the most mundane-sounding applications and, simultaneously, one of the most transformational. SAP will help me showcase working AI products with real-world business impacts.
Why SAP? It seems like a weird choice. I’ve had opportunities to partner with Cohere and Microsoft, but SAP will tell the Generative AI story from a new perspective. Companies leading the hype cycle draw a large crowd, but their story is only relevant to about 5% of businesses. SAP is talking to the other 95%.
They are as pragmatic about Generative AI as I am. It’s a 0-hype company that doesn’t deliver vaporware. When they demo a product, it works in the real world. The results they forecast materialize as expected. This week, SAP announced Joule, the Generative AI interface that will orchestrate SAP’s user experience for the next 5 years or more.
With Generative AI Hype And Reality Are Neighbors
You wouldn't believe me if I told you what Generative AI would do to the workplace. You must see it first because the transformation is dramatic. Annual productivity has improved by 1.4% annually in the US and .8% in the EU for almost 15 years. Generative AI tools like GitHub Copilot improve developer productivity by 55% overnight. We lack a frame of reference to compare what’s about to happen with.
As a result, we tend to focus on the wrong metrics because there’s a hidden assumption in the flashy productivity numbers. GitHub’s study showed that developers delivered more code, but does that mean they will deliver more business value? Most Generative AI tools are optimized to deliver a 55% improvement in productivity. SAP’s Joule is optimized for business value metrics.
For example, based on SAP's data, Joule is forecasted to reduce days sales outstanding (DSO) by up to 10%. The Generative AI interface will reduce overdue items by up to 50%. The difference in language reflects a focus on business outcomes instead of flashy productivity metrics. Joule is the first Generative AI interface that targets KPIs that CxOs and shareholders track.
If I just come out and say where Generative AI takes the modern firm, I will sound like a fraud. I am as anti-hype as it gets in the AI field, and I fully believe that Generative AI has been oversold as the panacea. I also know that Generative AI will deliver productivity gains that sound ridiculous today. Incredible claims require tangible proof, and that’s the only way to separate hype from reality when both can be hard to tell apart.
Generative AI Products Don’t Look Like AI
Products that overpromise and underdeliver business impacts are already being exposed. They are shells that hide API calls to ChatGPT or Claude. VCs have invested tens of millions of dollars into applications quickly replicated and improved upon by high school students. There is a lot of vaporware because Generative AI is mistaken for a killer app, and it isn’t.
At Meta, AI has been a big part of the turnaround story. The platform has leveraged AI for more personalized content recommendations. It pulls content in from across the platform and web. Without AI, personalization at scale would be impossible, and users would drown in irrelevant content.
Meta’s engagement metrics have risen, which increased advertising revenue. However, ask users if their Facebook timeline is AI, and most will say, “No.” Their engagement is driven by the experience and functionality that AI enables, not AI itself.
In contrast, Generative AI-first businesses like Runway, Lensa, and Remini have growing retention challenges. One-month retention rates for those three are between 35% and 45%. Retention is a problem for companies that put Generative AI at the center of their products and platforms. Joule is positioned as a front-end UI that provides access to SAP’s suite of enterprise apps and the business’s data.
领英推荐
Businesses that succeed with Generative AI products have high-value functionality behind the technology. At Meta, Generative AI improves the user experience by providing access to content users enjoy but wouldn’t have found independently. Joule improves business outcomes by orchestrating the complex SAP application ecosystem and simplifying workflows.
Joule leverages Generative AI as a killer interface that delivers access to data and application orchestration that wasn’t possible before. There must be substantive functionality on the other side of the interface for it to deliver business value. That’s where ChatGPT and many other productivity apps struggle.
ChatGPT provides personalized access to content or answers like social media sites do. Users can get that from Meta or Google’s Bard, so many Generative AI apps struggle to differentiate themselves from social media and search apps. They are labeled productivity tools, but no productivity-supporting applications are on the other side of the interface.
Joule will facilitate easier access to data across the enterprise, but what’s the business impact? Business leaders who say their data is accessible are twice as likely to say they are very prepared for:
For each point, access to data is just one part of a more extensive solution. Data is actionable, so the Generative Interface must provide access to apps that enable business leaders to act. Joule’s orchestration capabilities simplify complex workflows and reduce the delay between insights and action.
Businesses Struggle With AI Before They Get To AI
AI adoption isn’t a last-mile problem. If it were, all the talent and infrastructure businesses have spent the last decade investing in would propel us into an AI-driven utopia. AI adoption is a first-mile problem, and enterprises will struggle to adopt Generative AI because the upfront challenges haven’t been resolved.
The first-mile problem is enterprise-wide and spans multiple technologies. That’s another reason I’m partnering with SAP to explain the Generative AI product paradigm. It’s one of the few businesses with applications that support enterprise use cases with horizontal breadth (across business units and use cases) and vertical depth (comprehensive capability to serve a business unit and use case). Forward-looking firms like Sequoia have noted that Generative AI products that gain traction typically have vertical depth, not simply horizontal breadth.
SAP also supports a multi-technology business (digital, cloud, data, devices, AI, etc.). Generative AI now has a dedicated budget in most businesses, but the supporting technologies have yet to catch up. Generative AI workflows and finetuning workloads will primarily live in the cloud. Data must be formatted for analytics and AI applications, not just BI and reporting. Generative AI will rely heavily on existing digital software rather than replacing it. Devices and sensors will be rich data sources for prompt engineering and finetuning.
Companies that see Generative AI as a last-mile problem will adopt tools they cannot fully use. The challenge with Generative AI is that the tools function right out of the box, but they don’t work reliably enough to deliver value. Most tools have horizontal breadth that supports multiple use cases. However, they lack the vertical depth to deliver comprehensive capability to support those use cases in a way that delivers business outcome improvements.
SAP leverages the business’s data and its own proprietary data to deliver vertical depth behind Joule. In the next few months, I’ll work with SAP to showcase vertical depth, first for finance and supply chain use cases. I have planned four live-stream webinars to explain how Generative AI gets beyond the productivity vanity metrics and delivers business impacts.
I talk extensively about product-first AI vs. AI-first products. SAP is helping me deliver real-world examples of what happens when a company optimizes for business impact and value creation. There’s a lot more to come.
Technical Project Manager | Technical Product Manager | PMP? | CSM? | CSPO?| MBA | BE | Specialized in SaaS & data-driven projects | Data science enthusiast | Engineer @ ?? | ? blogger @ Monkidea
1 年I'm looking forward to your upcoming webinars and learning more about the second wave of Generative Interfaces.
Thanks Vin for sharing your insight on the SAP #joule announcement. I look forward to seeing where AI takes us!
Fractional CFO????I help SMBs grow with Financial Clarity & Confidence | Founder & CEO @freshfpa | FP&A, Finance & CFO Thought Leader | Finance Influencer | Former Amateur Boxer ??
1 年Vin Vashishta this is super exciting and super interesting on the application across the entire business! Also, being able to access enterprise wise data to make better data driven decisions is incredible and will help level up the entire businesses value regardless of functional area. This is a game changer particularly for CFOs and Finance teams to break down enterprise data silos. This is ??
Data Scientist | Learning about GenAI | Banking/Fintech | ex-Citi | IIMA, IISc
1 年This promises to be fascinating Vin Vashishta
Host and Industry Analyst @ CXOTalk
1 年Love the idea of exploring developer efficiency and the impact on writing better code. However, does this work for experienced developers or is the benefit primarily to more junior folks?