C3 Transform 2023

C3 AI CEO Tom Siebel believes Generative AI will change everything 

The 2023 C3 Transform conference was held in Boca Raton, FL, March 7-9, with 450 attendees representing 175 organizations from 24 countries. The primary theme of this year’s event and title of CEO Tom Siebel’s opening keynote: “Generative AI Changes Everything.” Where other software companies typically save their major product announcements for a big reveal at their annual user’s conference, C3 AI issued a press release in January announcing the launch of their new C3 Generative AI Product Suite and its first product – C3 Generative AI for Enterprise Search. Six weeks later, the Transform conference would mark the new products’ general availability and the opportunity to go in-depth with customers on just how big a paradigm shift C3 believes Generative AI will be. 

Adoption of AI is accelerating  

“The world has changed…the market is getting increasingly attenuated to AI,” said Siebel, noting the significant shift that has taken place since he founded C3 AI fourteen years ago - when the idea that corporate data would ever be in a public cloud was strongly rejected. Now, “clients are asking how we will use AI to digitally transform our organization, leveraging the cloud.” 

Industrial companies are indeed recognizing the value and the many use cases of artificial intelligence and machine learning and are implementing these capabilities at scale. In our latest LNS Research State of IX Tech survey, 46% of IX Leaders reported AI/ML capabilities are currently implemented, with another 32% running a pilot. That’s an impressive 78% of IX Leaders in our survey that have deployed some level of AI/ML in their enterprise. Followers are also starting to roll out AI/ML, with 24% reporting currently implemented and 22% in a pilot.

The next wave of AI innovation 

With the adoption and deployments of AI continuing to grow across such a broad spectrum of business and industrial processes and use cases, and the early “black box voodoo” stigma of AI going away, we are seeing a new wave of innovation aided by Large Language Models (LLM) like ChatGPT and GPT-3. “For a few years,” C3 AI had, according to Siebel, already been looking at the possibilities these new AI tools would enable. Like many product advancements, a push from an existing customer sparked action.

The development process for C3 Generative AI for Enterprise Search moved quickly after a US Department of Defense (DoD) leader sent Tom Siebel an email simply stating, “I want Google for DoD. A user asks a question and gets an answer. How do I make that happen?”

The C3 team quickly identified a key to adoption for this new approach to complex computing applications would be to leverage a user interface that everybody in the world already knows how to use: the Google search page. “Under the hood,” Siebel describes “an orchestration layer” where they have incorporated natural language processing (NLP), large language models, generative AI models, retrieval models, predictive models, and reinforcement learning. 

I won’t attempt to explain all the capabilities or tech underlying C3 Generative AI in a blog post, but the summary of what C3 AI has built is to utilize large language models and the rest of the C3 model-driven architecture to aggregate all the information sources that a company has. Sources can include text, structured data, unstructured data, or image data to deliver a comprehensive enterprise search capability and use the power of AI to generate insights.

While Siebel said it’s hard to describe how Generative AI will forever transform human-computer interaction, he did put a number on it: “This promises to be a $600 billion software market...” He went on to say that we can expect to see billions of dollars invested in Generative AI every year by IBM, Google, and Microsoft.  
Type in a query and the system will locate, retrieve, and present relevant data from the entire breadth of an enterprise’s information systems. It can, with proper access controls, include data from external systems. The software also consists of a permissions/role-based authentication structure so that users can only access the information and generated insights they are authorized to see. Users can drill deeper and interactively chat with the AI system from the results page. 

Use Cases: 

Prior to and while at the Transform conference, LNS Research analysts were briefed on a few of the use cases and capabilities of the C3 Generative AI Product Suite: 

• • • C3 Generative AI for Enterprise Search: The baseline use case for the US Dept of Defense to which Lt. Gen. Ed Cardon (Ret.), former commanding general of the U.S. Army Cyber Command, said, “This technology breakthrough can help dissolve the biggest barrier that we have to effective action, which is access to timely, accurate information and insight at all levels of the organization.”  
      Thinking about the size, complexity of data sets, and global footprint of the US DoD (a $753 Billion per year operation), we can begin to wrap our heads around how powerful this “Google for the Enterprise” capability will be for industrials and manufacturers who have siloed information in disparate databases, manufacturing and asset management systems, and other critical business applications, all while preserving security and other access controls

• • • C3 Generative AI with the C3 AI ESG application: In the world of ESG reporting, there is no unified global standard of exactly what corporations are required to report. In the ESG example, the user is enabled to identify and navigate data, trends, and topics for materiality and in the context of stated company ESG initiatives. Even without standardization of reporting requirements, C3 AI ESG allows companies to generate reports (with or without Generative AI support) that map to a number of voluntary standards such as SASB, GRI, TCFD, and CDP. 

    • Most relevant to LNS Research and our industrial/manufacturing clients is C3 Generative AI with the C3 AI Reliability application. In this use case example, the system can deliver added context and higher confidence predictions on the likelihood of equipment failure with better-informed recommendations by bringing together relevant data from once siloed systems such as design and engineering specs, alerts summaries, work order summaries, case summaries, and maintenance technician notes previously buried in free text fields of the EAM or CMMS system. 

The ability to search systems for relevant data, even using NLP models, is obviously not new but is generally done a system at a time and is thus very time-consuming and highly inefficient. This new approach using Generative AI with a search bar and interactive chat (ask a question – get an answer) has enormous potential and will be another disruptor in the AI space. However, the message that this technology “changes everything” would have been stronger had the use cases been actual customer presentations at the Transform event demonstrating these capabilities. 

Challenges ahead 

Too fast, too soon will hurt trust and adoption of Generative AI

Tom Siebel himself cautioned that certain components (speaking specifically about Open AI’s ChatGPT) are new and will improve over time. “There's lots going on with chat…when it gets fully developed, it will be very interesting for commercial applications.”  

I am not arguing that C3 Generative AI was released prematurely. The ChatGPT component is just a part of the overall solution. However, expectation setting is important when introducing advanced technology such as Generative AI. It won’t be perfect on day one as it will take time for the generative AI and reinforcement learning models to improve results and identify and apply weighting or a trust score to data sources in the enterprise. For example, Bob’s notes in an email venting about why Pump 1440 continues to fail may bring some context. However, it likely has less importance in the overall analysis, predictions, and recommendations than a detailed case summary on Pump 1440 from the APM system, where a comprehensive Root Cause Analysis was performed.  

How long will the Generative AI and reinforcement learning models take to “learn” and output meaningful and trustworthy answers and insights? That is THE question. The C3 Generative AI product brochure says 8-12 weeks, but potential buyers will demand that it be validated by actual customers who have implemented the C3 Generative AI solution and can share their experiences and best practices. 

Competition in Generative AI solutions for industrial is already heating up

C3 AI has established itself early in the Generative AI market, but US venture capital funding of Generative AI was up 27% year over year in 2022, hitting $1.4 billion, according to PitchBook. How much of that $1.4 billion focused on industrial applications isn’t differentiated in the report, but LNS Research is already seeing other technology providers in the manufacturing sector introducing Generative AI into their offerings, including SparkCognition. Over the last few months, we’ve seen the introduction of large language models (LLM), including Microsoft Copilot and Google Bard, following the buzz (and over 1 million users) working with OpenAI’s ChatGPT. Connected Frontline Workforce (CFW) platform provider Augmentir just announced the addition of ChatGPT into their AI toolset. Game on. 

Customer Presentations

A conference highlight was the presentation by Roshan Shah and Steven Bakalar of
Georgia Pacific, who discussed how Georgia Pacific is improving manufacturing operations with AI. In one example from their presentation, the team demonstrated how they connected 90,000 sensors that send data to their models every 10 minutes. “We can tell you the 40 that are likely to fail in the next 45 days…What does that mean? That effectively means for all the folks that are in our facilities, instead of everybody going in and manually taking readings on a route or schedule basis looking for issues, we can rely on AI to tell us where the problems are. So we can guide the work for folks that are actually going and fixing versus looking for problems.” The Georgia Pacific team again demonstrates how advanced they are in their use of AI and their overall Industrial Transformation initiatives.

Another highlight was the presentation by Mark Ratcliffe from Cargill, where Production Scheduling Optimization is a crucial part of their strategy. They are now live at eight plants, six that are fully up and running, and two that just went live in the past few weeks. This shows significant progress in the Digital Transformation journey of Cargill from last year when Mark outlined the project at C3 Transform 2022. 

Both Georgia Pacific (part of Koch Industries) and Cargill are privately held companies, thus, no specific financial information on ROI for these projects was shared. Although Georgia Pacific hinted, “on an annual basis we can see nine figures,” confirming for the audience at Transform that the value generated is significant.

Conclusions 

Overall, C3 Transform 2023 delivered what a user conference needs to for its customers: new product initiatives, current product enhancements, and compelling success stories. In this current period of economic uncertainty, with layoffs sweeping through the tech industry, it was also essential to communicate stability. To address that point, Tom Siebel ensured the audience knew the company has a strong cash position “with just shy of a billion dollars in the bank,” a message not only important for customers and prospective customers attending Transform but for employees as well.

Conference schedules are always a challenge, but an opportunity for C3 AI to improve the Transform conference experience is to allocate more dedicated time for attendees to experience the Solution Hub (demo area). It was a jam-packed two-and-a-half days, but more time with the product managers and navigating through the software's functionality is a win-win.

Takeaways for Manufacturers

1. 1. 1. Start to identify your data silos and information pain points. If you someday deploy a Generative AI solution and have the capability to search the entire breadth of your company’s information systems, where is data trapped?

      2. To build trust, get familiar with and start testing large language models like ChatGPT, Copilot, and Bard against known information. Trust in technology is critical for user adoption in manufacturing.

      3. Have patience. AI-enabled enterprise search is a fairly new capability for industrial users, and it will take time for the models to be trained. These approaches will mature and improve as more use cases are developed and deployed.

More soon from LNS Research on Generative AI

Generative AI as a discipline is already radically changing the content creation process with its ability to take existing images, videos, or sounds and create entirely new content. A core precept of Generative AI is the ability to autonomously generate high-fidelity synthetic data when models are data starved. The Pharma industry has already seen how Generative AI can significantly reduce the drug discovery timeline, as has the semiconductor industry with chip design. LNS Research will be doing more on Generative AI and how it will change manufacturing in our 2023 and 2024 research agenda.