TOPIC: SHIONOGI
What SAS Innovate 2025 revealed about the future of enterprise analytics
21st June 2025SAS Innovate 2025 comprised a global event in Orlando (6th-9th May) followed by regional editions on tour. This document provides observations from both the global event and the London stop (3rd-4th June), covering technical content, platform developments and thematic emphasis across the two occasions. The global event featured extensive recorded content covering platform capabilities, migration approaches and practical applications, whilst the London event incorporated these themes with additional local perspectives and a particular focus on governance and life sciences applications.
Global Event
Platform Expansion and New Capabilities
The global SAS Innovate 2025 event included content on SAS Clinical Acceleration, positioned as a SAS Viya equivalent to SAS LSAF. Whilst much appeared familiar from the predecessor platform, performance improvements and additional capabilities represented meaningful enhancements.
Two presentations, likely restricted to in-person attendees based on their absence from certain schedules, covered AI-powered SAS code generation. Shionogi presented on using AI for clinical studies and real-world evidence generation, with the significant detail that the AI capability existed within SAS Viya rather than depending on external large language models. Another session addressed interrogating and generating study protocols using SAS Viya, including functionality intended to support study planning in ways that could improve success probability.
These sessions collectively indicated a directional shift. The scope extends beyond conventional expectations of "SAS in clinical" contexts, moving into upstream and adjacent activities, including protocol development and more integrated automation.
Architectural Approaches and Data Movement
A significant theme across multiple sessions addressed fundamental shifts in data architecture. The traditional approach of moving massive datasets from various sources into a single centralised analytics engine is being challenged by a new paradigm: bringing analytics to the data. The integration of SAS Viya with SingleStore exemplifies this approach, where analytics processing occurs directly within the source database rather than requiring data extraction and loading. This architectural change can reduce infrastructure requirements for specific workloads by as much as 50 per cent, whilst eliminating the complexity and cost associated with constant data movement and duplication.
Trustworthy AI and Organisational Reflection
Keynote presentations addressed the relationship between AI systems and organisational practices. SAS Vice President of Data Ethics Practice Reggie Townsend articulated a perspective that reframes common concerns about AI bias. When AI produces biased results, the issue is not primarily technical failure, but rather a reflection of biases already embedded within cultural and organisational practices. This view positions AI as a diagnostic tool that surfaces systemic issues requiring organisational attention rather than merely technical remediation.
The focus on trustworthy AI extended beyond bias to encompass governance frameworks, transparency requirements and the persistent challenge that poor data quality leads to ineffective AI regardless of model sophistication. These considerations hold particular significance in probabilistic AI contexts, especially where SAS aims to incorporate deterministic elements into aspects of its AI offering.
Natural Language Interfaces and Accessibility
Content addressing SAS Viya Copilot demonstrated the platform's natural language capabilities, enabling users to interact with analytics through conversational queries rather than requiring technical syntax. This approach aims to democratise data access by allowing users with limited technical knowledge to directly engage with complex datasets. The Copilot functionality, built on Microsoft Azure OpenAI Service, supports code generation, model development assistance and natural language explanations of analytical outputs.
Cloud Migration and Infrastructure Considerations
A presentation on transforming clinical programming using SAS Clinical Acceleration was scheduled but not accessible at the global event. The closing session featured the CIO of Parexel discussing their transition to SAS managed cloud services. Characterised as a modernisation initiative, reported outcomes included reduced outage frequency. This aligns with observations from other multi-tenant systems, where maintaining stability and availability represents a fundamental requirement that often proves more complex than external perspectives might suggest.
Content addressing cloud-native strategies emphasised a fundamental psychological shift in resource management. Rather than the traditional capital expenditure mindset where physical servers run continuously, cloud environments enable strategic use of the capability to create and destroy computing resources on demand. Approaches include spinning up analytics environments at the start of the working day and shutting them down at the end, with more sophisticated implementations that automatically save and shut down environments after periods of inactivity. This dynamic approach ensures organisations pay only for actively used resources.
Presentations on organisational change management accompanying technical migrations emphasised that successful technology projects require attention to human factors alongside technical implementation. Strategies discussed included formal launch events to mark transitions, structured support mechanisms such as office hours for technical questions and community-building activities designed to foster relationships and maintain engagement during periods of change.
Platform Integration and Practical Applications
Content on SAS Viya Workbench covered availability through Azure and AWS, Python integration, R compatibility and interfacing with SAS Enterprise Guide, with demonstrations of several features. As SAS expands support for open-source languages, the presentation illustrated how these capabilities can provide a unified platform for different technical communities.
A presentation on retrieval augmented generation with unstructured data (such as system manuals), combined with agentic AI for diagnosing manufacturing system problems, offered a concrete use case. Given the tendency for these subjects to become abstract, the connected example provided practical insight into how components can function together in operational settings.
Digital Twins and Immersive Simulation
A notable announcement at the global event involved the partnership between SAS and Epic Games to create enhanced digital twins using Unreal Engine. This collaboration applies the same photorealistic 3D rendering technology used in Fortnite to industrial applications. Georgia-Pacific piloted this technology at its Savannah River Mill, which manufactures napkins, paper towels and toilet tissue. The facility was captured using RealityScan, Epic's mobile application, to create photorealistic renderings imported into Unreal Engine.
The application focused on optimising automated guided vehicle deployment and routing strategies. Rather than testing scenarios in the physical environment with associated costs and safety risks, the digital twin enables simulation of complex factory floor operations including AGV navigation, proximity alerts, obstacles and rare adverse events. SAS CTO Bryan Harris emphasised that digital twins should not only function like the real world but also look like it, enabling more accessible decision-making for frontline workers, engineers and machine operators beyond traditional data scientist roles.
The collaboration extends beyond visual fidelity. SAS developed a plugin connecting Unreal Engine to SAS Viya, enabling real-time data from simulated environments to fuel AI models that analyse, optimise and test industrial operations. This approach allows organisations to explore "what-if" scenarios virtually before implementing changes in physical facilities, potentially delivering cost savings whilst improving safety and operational efficiency.
Marketing Intelligence and Customer Respect
Content on SAS Customer Intelligence 360 addressed the platform's marketing decisioning capabilities, including next-best-offer functionality and real-time personalisation across channels. A notable emphasis concerned contact policies and rules that enable marketers to limit communication frequency, reflecting a strategic choice to respect customer attention rather than maximise message volume. This approach recognises that in environments characterised by notification saturation, demonstrating restraint can build trust and ensure greater engagement when communications do occur.
Financial Crime and Integrated Analytics
Presentations on financial crime addressed the value of integrated platforms that connect traditionally siloed functions such as fraud detection, anti-money laundering and sanctions screening. Network analytics capabilities enable identification of patterns and relationships across these domains that might otherwise remain hidden. Examples illustrated how seemingly routine alerts, when analysed within a comprehensive view of connected data, can reveal connections to significant criminal networks, transforming tactical operational issues into sources of strategic intelligence.
Data Lineage and Transformation Planning
Content on data lineage reframed this capability from a purely technical concern to a strategic tool for transformation planning. For large-scale modernisation initiatives, comprehensive mapping of data flows, transformations and dependencies provides the foundation for accurate effort estimation, budgetary planning and risk assessment. This visibility enables organisations to proceed with complex changes whilst maintaining confidence that critical downstream processes will not be inadvertently affected.
Development Practices and Migration Approaches
Sessions included content on using Bitbucket with SAS Viya to support continuous integration and continuous deployment pipelines for SAS code. Git formed the foundation of the approach, with supporting tools such as JQ. Given the current state of manual validation processes, this content addressed a genuine need for more robust validation methods for SAS macros used across clinical portfolios, where these activities can require several weeks and efficiency improvements would represent substantial value.
Another session provided detailed coverage of migrating from SAS 9 to SAS Viya, focusing on assessment methods for determining what requires migration and techniques for locating existing assets. The content reflected the reality that the discovery phase often constitutes the primary work effort rather than a preliminary step.
A presentation on implementing SAS Viya on-premises under restrictive security requirements described a solution requiring sustained collaboration with SAS over multiple years to achieve necessary modifications. This illustrated how certain deployments are defined primarily by governance, controls and assurance requirements rather than by product features.
Technical Fundamentals and Persistent Challenges
A hands-on session on data-driven output programming with SAS macros provided practical content with life sciences examples. Control tables and CALL EXECUTE represented familiar approaches, whilst the data step RESOLVE function offered new functionality worth exploring, particularly given its capability to work with macro expressions rather than being limited to macro variables in the manner of SYMGET.
A recurring theme across multiple contexts emphasised that poor data quality leads to ineffective AI and consequently to flawed decision-making. The technological environment evolves, but fundamental challenges persist. This consideration holds particular significance in probabilistic AI contexts, especially where SAS aims to incorporate deterministic elements into aspects of its AI offering.
London Event
Overview and Core Themes
The London edition of SAS Innovate 2025 on Tour demonstrated the pervasive influence of AI across the programme. The event concluded with Michael Wooldridge from the University of Oxford providing an overview of different categories of AI, offering conceptual grounding for a day when terminology and ambition frequently extended beyond current practical adoption.
The opening session presented SAS' recent offerings, maintaining consistency with content from the global event in Orlando whilst incorporating local perspectives. Trustworthiness, responsibility and governance emerged as prominent themes, particularly relevant given the current industry emphasis on innovation. A panel discussion included a brief exchange regarding the term "digital workforce", reflecting an awareness of the human implications that can be absent from wider industry discussion.
Life Sciences Stream Content
The Life Sciences stream focused heavily on AI, with presentations from AWS, AstraZeneca and IQVIA addressing the subject, followed by a panel discussion continuing this direction. The scale of technological change represents a tangible shift affecting all parts of the ecosystem. A presentation from a healthcare professional provided context regarding the operational environment within which pharmaceutical companies function. SAS CTO Bryan Harris expressed appreciation for pharmaceutical research and development work, an acknowledgement that appeared both substantive and appropriate to the setting.
Observations from selected sessions of SAS Innovate 2024
21st June 2024SAS Innovate 2024 provided insight into evolving approaches to analytics modernisation, platform development and applied data science across multiple industries. This document captures observations from sessions addressing strategic platform migrations, unified analytics environments, enterprise integration patterns and practical applications in regulated sectors. The content reflects a discipline transitioning from experimental implementations to production-grade, business-critical infrastructure.
Strategic Platform Modernisation
A presentation from DNB Bank detailed the organisation's migration from SAS 9.4 to SAS Viya on Microsoft Azure. The strategic approach proved counter-intuitive: whilst SAS Viya supports both SAS and Python code seamlessly, DNB deliberately chose to rewrite their legacy SAS code library into Python. The rationale combined two business objectives. First, expanding the addressable talent market by tapping into the global Python developer pool. Second, creating a viable exit strategy from their primary analytics vendor, ensuring compliance with financial regulatory requirements to demonstrate realistic vendor transition options within 30 to 90 days.
This decision represents a fundamental shift in enterprise software value propositions. Competitive advantage no longer derives from creating vendor lock-in, but from providing powerful, stable and governed environments that fully embrace open-source tools. The winning strategy involves convincing customers to remain because the platform delivers undeniable value, not because departure presents insurmountable difficulty. This is something that signals the maturing of a market, where value flows through partnership rather than proprietary constraints.
Unified Analytics Environments
A healthcare analytics presentation addressed the persistent debate between low-code/no-code interfaces for business users and professional coding environments for data scientists. Two analysts tackled identical problems (predicting diabetes risk factors using a public CDC dataset) using different approaches within the same platform.
The low-code user employed SAS Viya's Model Studio, a visual interface. This analyst assessed the model for statistical bias against variables such as age and gender by selecting a configuration option, whereupon the platform automatically generated fairness statistics and visualisations.
The professional coder used SAS Viya Workbench, a code-first environment similar to Visual Studio Code. This analyst manually wrote code to perform identical bias assessments. However, direct code access enabled fine-tuning of variable interactions (such as age and cholesterol), ultimately producing a logistic regression model with marginally superior performance compared to the low-code approach.
The demonstration illustrated that the debate presents a false dichotomy. The actual value resides in unified platforms, enabling both personas to achieve exceptional productivity. Citizen data scientists can rapidly build and validate baseline models, whilst expert coders can refine those same models with advanced techniques and deploy them, all within a single ecosystem. This unified approach characterises disciplinary development, where focus shifts from tribal tool debates to collective problem-solving.
Analytics as Enterprise Infrastructure
Multiple architectural demonstrations illustrated analytics platforms evolving beyond sophisticated workbenches for specialists into the central nervous system of enterprise operations. Three distinct patterns emerged:
The AI Assistant Architecture: A demonstration featured a customer-facing AI assistant built with Azure OpenAI. When users interacted with the chatbot regarding credit risk, requests routed through Azure Logic App not to the large language model for decisions but to a SAS Intelligent Decisioning engine. The SAS engine functioned as the trusted decision core, executing business rules and models to generate real-time risk assessments, which returned to the chatbot for customer delivery. SAS provided not the interface but the automated decision engine.
The Digital Twin Pattern: A pharmaceutical use case described using historical data from penicillin manufacturing batches to train machine learning models. These models became digital twins of physical bioreactors. Rather than conducting costly and time-consuming physical experiments, researchers executed thousands of in silico simulated experiments, adjusting parameters in the model to discover the optimal recipe for maximising yield (the "Golden Batch").
The Microsoft 365 Automation Hub: A workflow demonstration showed SAS programmes functioning as critical nodes in Microsoft 365 ecosystems. The automated process involved SAS code accessing SharePoint folders, retrieving Excel files, executing analyses, generating new reports as Excel files and delivering those reports directly into Microsoft Teams channels for business users.
These patterns mark profound evolution. Analytics platforms are moving beyond sophisticated calculators for experts, becoming foundational infrastructure: the connective tissue enabling intelligent automation and integrating disparate systems such as cloud office suites, AI interfaces and industrial hardware into cohesive business processes. This evolution from specialised tool to core infrastructure clearly indicates analytics' growing maturity within enterprise contexts.
Applied Data Science in High-Stakes Environments
Whilst much data science narrative focuses on e-commerce recommendations or marketing optimisation, compelling applications tackle intensely human, high-stakes operational challenges. Heather Hallett, a former ICU nurse and healthcare industry consultant at SAS, presented on improving hospital efficiency.
She described the challenge of staffing intensive care units, where having appropriate nurse numbers with correct skills proves critical. Staffing decisions constitute "life and death decisions". Her team uses forecasting models (such as ARIMA) to predict patient demand and optimisation algorithms (including mixed-integer programming) to create optimal nurse schedules. The optimisation addresses more than headcount; it matches nurses' specific skills, such as certifications for complex assistive devices like intra-aortic balloon pumps, to forecasted needs of the sickest patients.
A second use case applied identical operational rigour to community care. Using the classic "travelling salesman solver" from optimisation theory, the team planned efficient daily routes for mobile care vans serving maximum numbers of patients in their homes, delivering essential services to those unable to reach hospitals easily.
These applications ground abstract concepts of forecasting and optimisation in deeply tangible human contexts. They demonstrate that beyond driving revenue or reducing costs, the ultimate purpose of data science and operational analytics can be directly improving and even saving human lives. This application of sophisticated mathematics to life preservation marks data science evolution from commercial tool to critical component of human-centred operations.
Transparency as Competitive Advantage
In highly regulated industries such as pharmaceuticals, generating trustworthy research proves paramount. A presentation from Japanese pharmaceutical company Shionogi detailed how they transform the transparency challenge in Real-World Evidence (RWE) into competitive advantage.
The core problem with RWE studies, which analyse data from sources such as electronic health records and insurance claims, involves their historical lack of standardisation and transparency compared to randomised clinical trials, leading regulators and peers to question validity. Shionogi's solution is an internal system called "AI SAS for RWE", addressing the challenge through two approaches:
Standardisation: The system transforms disparate Real-World Data from various vendors into a Shionogi-defined common data model based on OMOP principles, ensuring consistency where direct conversion of Japanese RWD proves challenging.
Semi-Automation: It semi-automates the entire analysis workflow, from defining research concepts to generating final tables, figures and reports.
The most innovative aspect involves its foundation in radical transparency. The system automatically records every research process step: from the initial concept suite where analysis is defined, through specification documents, final analysis programmes and resulting reports, directly into Git. This creates a complete, immutable and auditable history of exactly how evidence was generated.
This represents more than a clever technical solution; it constitutes profound strategic positioning. By building transparent, reproducible and efficient systems for generating RWE, Shionogi directly addresses core industry challenges. They work to increase research quality and trustworthiness, effectively transforming regulatory burden into competitive edge built on integrity. This move toward provable, auditable results hallmarks a discipline transitioning from experimental art to industrial-grade science.
User Experience as Productivity Multiplier
In complex data tool contexts, user experience (UX) has evolved beyond "nice-to-have" aesthetic features into a central product strategy pillar, directly tied to user productivity and talent acquisition. A detailed examination of the upcoming complete rewrite of SAS Studio illustrated this point.
The motivation for the massive undertaking proved straightforward: the old architecture was slow and becoming a drag on user productivity. The primary goal for the new version involved making a web-based application "feel like a desktop application" regarding speed and responsiveness. To achieve this, the team focused on improvements directly boosting productivity for coders and analysts:
A Modern Editor: Integrating the Monaco editor used in the widely popular Visual Studio Code, providing familiar and powerful coding experiences.
Smarter Assistance: Improving code completion and syntax help to reduce errors and time spent consulting documentation.
Better Navigation: Adding features such as code "mini-maps" enabling programmers to navigate thousands of lines of code instantly.
For modern technical software, UX has become a fundamental competitive differentiator. Faster, more intuitive and less frustrating tools do not merely improve existing user satisfaction; they enhance productivity. In competitive markets for top data science and engineering talent, providing a best-in-class user experience represents a key strategy for attracting and retaining exceptional people. The next leap in team productivity might derive not from new algorithms but from superior interfaces.
Conclusion
These observations from SAS Innovate 2024 illustrate a discipline maturing. Data science is moving beyond isolated experiments and "science projects", becoming pragmatic, integrated, transparent and deeply human business functionality. Focus shifts from algorithmic novelty to real-world application value (whether enabling better user experiences, building regulatory trust or making life-or-death decisions on ICU floors).
As analytics becomes more integrated and accessible, the challenge involves identifying where it might unexpectedly transform core processes within organisations, moving from specialist concern to foundational infrastructure enabling intelligent, automated and human-centred operations.