SAS has been at the forefront of statistical analysis for almost five decades and has brought its rich, deep experience to help solve industry-specific challenges through packaged AI models to facilitate rapid deployment with low overheads.
The principles of mathematics and statistics are unchanging, and SAS has been delivering reliable, repeatable insightful analytics since the 1970s. What has changed in that time is the computing power we have available to perform these analytics along with an explosion in collected data, stored in many ways and forms all across every organisation.
This combination of big compute and big data has led us to the age of artificial intelligence - the stuff of science fiction no more, we have the potential to build smart apps that use existing data along with algorithmic models to make future predictions and assessments. Advice everywhere is that you can infuse AI into your organisation's activities. In fact, not only can you, but you should, you must, to keep ahead of the competition.
That's well and good to say, but there are two big issues, and SAS has determined to solve both. For one, AI like ChatGPT is known for its hallucinations. Yet, enterprise AI can't be making up results that make no sense, or are plain wrong. And secondly, while large tech companies are providing tools to simplify model creation, it's still a challenge to identify what you could do with your data to work smarter, and you need the skills in-house to act on it.
Or - you could take up SAS' innovations in packaged industry-specific AI models. The company has leveraged its strong experience to build models ready to use, giving genuine real-world solutions, and underpinned by guiding principles of transparency, explainability, and fairness.
{loadposition david08}
Last year, SAS committed one billion US dollars to delivering vertical solutions on top of the SAS analytics platform, currently available as SAS version 9.4, and as SAS Viya (for "visual analytics"). The base product remains a powerful computation engine, but SAS has previously enjoyed success with industry-specific solutions that sit on top of this engine, such as its fraud and anti-money-laundering products, and its customer intelligence 360 product. The commitment to fund further vertical solutions was significant, and the new models are the first fruits of this commitment.
What's more, the models reflect a third way of doing business with SAS. Taking the stage at SAS Innovate 2024 recently, SAS CTO Bryan Harris explained some companies wish to build AI - which they can do on the SAS platform, or they wish to buy solutions - which SAS offers in its verticals, or, they wish to subscribe to models - which it now offers.
And, many companies wish to invest in all three. For example, Harris explained, a pharmaceutical manufacturer relies on its ability to develop a strong drug pipeline, which in turn means drug discovery is critical to business.
In this case, those performing drug discovery would opt to work with the raw power of SAS, experimenting, creating, and innovating with it. Meanwhile, others in the business want predictive maintenance to keep plant and equipment functioning with minimal downtime and wasted expense. Here, they have existing SAS solutions to leverage. And, elsewhere in the business, some want to optimise the supply chain and can now subscribe to SAS models that make this happen.
Or, a bank may have people building on top of the SAS platform to assess viable product offerings during loan origination. In contrast, others use SAS vertical solutions for asset and liability management, and others yet again subscribe to models to rank customer risk.
My favourite kind of donut visualisation - actual donuts to represent the SAS ecosystem#SASInnovate pic.twitter.com/Eju7LNN8K3
— davidmwilliams (@davidmwilliams) April 18, 2024
SAS says its introduction of lightweight, industry-specific AI models is a "game-changing approach for organisations to tackle business challenges head-on." The models are available for individual license and means organisations of all sizes and types can leverage readily deployable AI tech to productionise real-world use cases with unparalleled efficiency.
Additionally, while the world is captivated by large language models (LLMs) right now, the SAS models go beyond LLMs and deliver industry-proven deterministic AI models that span use cases such as fraud detection, supply chain optimisation, entity management, document conversation, health care payment integrity, and more.
The models have been engineered for quick integration, and help organisations quickly take advantage of trustworthy AI to accelerate their realisation of tangible benefits.
“Models are the perfect complement to our existing solutions and SAS Viya platform offerings and cater to diverse business needs across various audiences, ensuring that innovation reaches every corner of our ecosystem,” said SAS VP AI and analytics Udo Sglavo. “By tailoring our approach to understanding specific industry needs, our frameworks empower businesses to flourish in their distinctive environments.”
“SAS Models provide organisations with flexible, timely and accessible AI that aligns with industry challenges,” Sglavo said. “Whether you’re embarking on your AI journey or seeking to accelerate the expansion of AI across your enterprise, SAS offers unparalleled depth and breadth in addressing your business’s unique needs.”
Furthermore, these models are explainable, transparent, and fair. These were non-negotiable fundamental principles the company has built its models on, explains SAS VP data ethics practice Reggie Townsend.
By this, the company means even if the AI provides a result that is not intuitive, you will be able to understand and trace through the reasoning that led to it. This is critical for trust, Townsend explained. If an AI is not transparent, not explainable, you cannot trust its decisions.
Further, by fair, the company means its models are stamped with "nutrition labels" that surface any potential bias in the source data, such as underrepresented population groups, and help ensure these are mitigated.
The SAS data ethics practice that Townsend heads is underpinned by its trustworthy AI lifecycle workflow, mapped to the NIST AI risk management framework, and its AI governance advisory services.
The first SAS models will be available generally later in 2024, with more information online.
The importance of these models can't be understated. "Companies cannot scale on the heroic actions of individuals," said SAS EVP and CIO Jay Upchurch. "You have to have tech that allows employees to be more productive than they are and the machine of the enterprise to grow and scale."
"Otherwise you're bound by human capital expense," he said. "You can't grow that way."
Thus, according to Upchurch, the challenge to business and technology leaders worldwide is how to bring data, AI, and automation to every business process you have. With SAS packaged industry-specific models, you get industry expertise and AI capabilities rolled into one.