Exclusive Interview: The Future of AI-Ready Data, Decision Intelligence, and Automation
In this article written by independent journalist Pat Brans with David Shannon, Head of Decisioning at SAS Northern Europe, we discuss how Automation is rapidly becoming the backbone of decision intelligence, enabling organisations to act on data in real time while safeguarding accuracy, ethics, and trust.
When organisations talk about transforming AI from research experiments into real business value, one theme comes up repeatedly: automation. In a recent interview, David Shannon, Head of Decisioning at SAS Northern Europe, explained how automation underpins what he calls “decision intelligence”—the discipline of putting insights into action.
“In a nutshell, decision intelligence is putting insight into action. It’s acting on our data,” Shannon said. “But in practice, it’s slightly broader than that. It’s a philosophy of real-time decisioning.”
According to Shannon, it has become much more difficult to make the decisions that lead to a customer experience that differentiates an organisations. LLMs need to be applied to take rich historical data and provide the best decisions—and RAG models should also be used.
Automation Across the AI Life Cycle
One of the strongest points Shannon made was the need to automate across the full AI life cycle—from ingesting data, to managing quality, training and validating models, deploying them, and continuously monitoring performance. This kind of end-to-end automation, he explained, allows companies to respond faster to change and free up employees for higher-value work.
“Have we seen organisations automate the full life cycle from data ingestion all the way to decision? Absolutely,” he explained. “The benefits are faster ability to respond to environmental change, better service to consumers, stronger relationships with suppliers, and a workforce that spends more time on innovation rather than repetitive tasks.”
Shannon was quick to note that this is not about replacing workers. “Organisations that re-skill their workforce to focus on the innovative and productive parts of the process are the ones that win,” he said.
This emphasis on trust and transparency naturally leads to the question of lineage—tracking how data flows and changes through the system. Shannon emphasised that automation provides the clarity needed to build confidence in AI outcomes.
“Organisations that re-skill their workforce to focus on the innovative
and productive parts of the process are the ones that win.”
“Any asset, be it a table of data, a model, or a decision process itself, is integral within that lineage, so everything is catalogued,” he said. “That does a number of things over time. It lets you see changes in behaviour, detect anomalies, and understand the impact of a change upstream or downstream. The graphical view of lineage provides transparency for auditors and helps ensure trust in AI outcomes.”
Another area where automation is proving vital is in the management of synthetic data. With organisations facing privacy concerns, cost pressures, and gaps in real-world data, synthetic datasets have become essential. Shannon explained how automation enables this at scale.
“The key difference today is the capability of algorithms to better
represent real-world data so we can train models more accurately.”
“Synthetic data really has evolved recently. The key difference today is the capability of algorithms to better represent real-world data so we can train models more accurately,” he said. He added that privacy and regulatory constraints often make real customer data unusable. Being able to synthesise data allows access to information that may otherwise require complex and expensive security controls.
But even with rich datasets, organisations face new risks as large language models grow in use. Hallucinations—outputs that are fabricated or misleading—remain a real danger. Shannon argued that automation has a central role in keeping these systems on track.
“It’s no longer just about monitoring analytical performance.
We also need to detect ethical bias and potential toxicity.”
“LLMs and generative AI provide many advantages, but they also produce hallucinations,” he said. “You have to put guardrails around that process. It’s not LLMs on their own, but composite AI—traditional machine learning models combined with LLMs—that provides more reliable outputs. Automation helps orchestrate those different models, govern prompts and outputs, and apply business logic so responses stay within safe bounds.”
Automation also extends into model monitoring, which has shifted far beyond traditional performance metrics. Shannon highlighted the importance of continuous oversight.
“It’s no longer just about monitoring analytical performance. We also need to detect ethical bias and potential toxicity, particularly with LLMs,” he said. “Organisations must decide at what point to automatically retrain a model or request human intervention. The ability to tune that balance—between touch-free automation and human-in-the-loop oversight—is a critical strength of an end-to-end process.”
These monitoring practices feed directly into measuring accuracy and business value. Without a system for tracking improvements in speed and precision, decision intelligence cannot deliver results.
“The more quickly you can respond to environmental change,
the more valuable your models are.”
“We often hear that many AI models never make it into production, or that it takes months to operationalise a model,” Shannon said. “Through automation, we’ve helped customers reduce that time from one to three months down to about three days for complex models—and even just a day in simple cases. That kind of impact is crucial, because the more quickly you can respond to environmental change, the more valuable your models are.”
Looking forward, Shannon described a future where automation tools continue to evolve, becoming more integrated with day-to-day operations and capable of learning from outcomes.
“The point of decision will get closer to the data origin or problem origin,” he predicted. “Decisioning systems will become integral to the processes interacting with suppliers, consumers, and partners. We’re already able to bring together different types of AI models in a single flow and publish them in a matter of days. The next frontier is making these systems self-improving—able to relearn from outcomes and become more automated over time.”
Broadcom Automic: Addressing the Challenges
Many of the challenges Shannon highlighted—lineage, monitoring, synthetic data, hallucinations, rapid operationalisation, and evolving automation tools—align well with the capabilities of Broadcom’s Automic platform as described in official documentation.
Automic Automation excels at end-to-end orchestration across business and IT workloads. It uses visual workflows, reusable components, and self-service portals to streamline scheduling, execution, and data pipeline synchronisation in hybrid environments automation.broadcom.com. Its strength lies in providing unified workflow control, visibility, and compliance through trend analytics and robust governance features automation.broadcom.com.
“When it comes to AI-augmented automation, Automic includes Automation.AI,
a generative AI component that enhances user interactions through
summarisation, troubleshooting guidance, and contextual assistance.”
When it comes to AI-augmented automation, Automic includes Automation.AI, a generative AI component that enhances user interactions through summarisation, troubleshooting guidance, and contextual assistance—delivered via conversational interfaces within the automation environment docs.automic.com. The ASK_AI functionality enables users to accomplish tasks like data extraction, structuring, and summarisation simply via natural language instructions—significantly reducing the complexity of scripting for data transformation workflows automation.broadcom.com.
On the synthetic data front, while this capability resides in a separate product—Broadcom Test Data Manager (TDM)—its Data Assistant Plus feature enables automated, rule-based generation of realistic synthetic data that mirrors production distributions, enforces relationships, and supports compliance with privacy regulations academy.broadcom.com.
Together, these capabilities make Broadcom Automic a strong fit for realising Shannon’s vision: a trustworthy, automated, AI-powered pipeline that handles lineage, compliance, monitoring, and adaptive decisioning at scale.
Your Next Steps
If you would like more information on how you can take the learnings above to help you manage change across your organisation, please complete the form below and one of our experts will be in touch.
