Laura Albert: Designing Decision Systems for Trade-Offs and Disruption

Laura Albert is a professor of industrial and systems engineering at the University of Wisconsin, a past president of INFORMS, and one of the authors of a recent paper on the grand challenges facing industrial and systems engineering. Much of her work covers operations research, resilience, cybersecurity, and decision-making.

One of Albert’s central concerns is resilience. Operations research has historically been drawn to optimal solutions, but optimal systems can also be fragile. A plan that performs exceptionally well under expected conditions may break down quickly when labor disappears, supply is disrupted, infrastructure fails, or assumptions change. Resilience therefore requires more than restoring a system to its pre-disruption state. It also requires thinking about how systems adapt when the conditions around them no longer match the assumptions used to design them.

Albert argues that the field should focus on larger problems. Industrial engineering and operations research often concentrate on tightly defined technical questions, even though the systems they influence are shaped by cybersecurity, supply chains, AI, infrastructure, and broader societal risks. Framing these as grand challenges pushes researchers and practitioners to work across those boundaries. It can also shape where funding goes, what universities teach, and which problems the next generation of engineers is prepared to tackle.

Education, therefore, needs to change to address the problems. Albert argues that new subjects cannot simply be added indefinitely to engineering programs. Required undergraduate courses have to cover topics such as AI, resilience, governance, and ethics through practical case studies. She has been developing AI and ethics cases for her own courses and plans to bring the economics of AI into engineering economics, including the cost of better data, testing, governance, and model deployment.

Generative AI raises an even more fundamental teaching problem. Albert expects AI systems to become substantially more capable over the next few years, particularly at tasks such as programming. That changes what engineers need to learn and how they should be assessed. Her concern is that using generative AI effectively still depends on domain expertise. An engineer has to know enough to distinguish a strong answer from a hallucination. If students outsource too much of the underlying work before developing that knowledge, they may gain access to better tools while becoming less capable of judging their output.

This is also why Albert sees AI governance as one of the largest opportunities for industrial and systems engineers. Governance determines who has authority, how information moves through an organization, how rules are enforced, and who is accountable. Industrial engineers are well suited to this because they are already trained to work across systems.

The hard part is managing trade-offs. AI audits, testing, governance, and compliance consume time and money, while organizations still have operational work to complete. Albert argues that systems engineers have an advantage because they are accustomed to making those trade-offs explicit. Every operating objective competes with other priorities, and treating every requirement as equally important eventually makes prioritization impossible. Good governance depends on understanding those competing objectives and designing rules that strike a workable balance.

Cybersecurity illustrates the same system problem. Much of cybersecurity research concentrates on individual exploits, software weaknesses, or defenses. Albert sees a large opportunity at the system level, where organizations must decide how to allocate limited resources across many possible risks. Those risks extend across software and physical supply chains, meaning an organization must protect its own systems while also accounting for vulnerabilities introduced by suppliers and connected infrastructure.

For Albert, these examples point to the same conclusion. The value of decision intelligence will depend on designing systems that remain useful when assumptions change, risks interact, AI becomes more capable, and organizational constraints get in the way. The opportunity for the field is to think at that larger level: resilience, governance, deployment, security, education, and the trade-offs connecting them.

Declarative AI 2026 will take place in Vilnius, Lithuania, from August 24-30, bringing together RuleML+RR 2026, DecisionCAMP 2026, and the Reasoning Web Summer School. The program covers rule-based reasoning, decision management, explainable AI, knowledge representation, and the use of declarative methods in AI systems. DecisionCAMP will focus specifically on decision management, the DMN standard, real-world implementations, and decision intelligence frameworks. Details →

The Institute of Business Forecasting & Planning will hold its European Business Planning, Forecasting & S&OP/IBP Conference in Amsterdam from November 18-20, 2026. The event brings together practitioners across demand planning, forecasting, supply chain, analytics, finance, sales, and operations, with two maturity tracks covering both S&OP/IBP implementation and more advanced planning practices. A pre-conference tutorial on demand planning and forecasting is scheduled for November 18. Details →

Stephanie Fleming and Janna Broaddus introduce Decision Risk Pattern Intelligence (DRPI), a governance framework for documenting recurring patterns in how people make decisions around AI-supported systems in courts, corrections, law enforcement, and other justice environments. DRPI is designed to examine decision conditions across multiple cases over time, including persistent gaps between policy, human judgment, and operational practice. It explicitly differs from predictive analytics, algorithmic risk scoring, compliance assessment, or post-incident investigation. The paper argues that longitudinal documentation can help institutions identify recurring governance weaknesses, preserve institutional knowledge, and improve oversight as AI becomes more embedded in high-stakes decisions. Read →

Benjamin Baer, Founding Member of the DecideWise community, argues that enterprises are sitting on valuable “dark data” spread across contracts, emails, financial records, and internal systems. The challenge is making that information usable without giving autonomous AI agents unrestricted access across the organization. Baer sees decision intelligence as the layer that can connect these fragmented sources while keeping decisions governed and measurable. Once safely incorporated, dark data can reveal how an organization actually operates and create an intelligence advantage that competitors cannot easily replicate. Read More →

Vijay Mehrotra and Michael Watson reflect on the first year of the Decision Intelligence Lab podcast and the themes that kept resurfacing across 19 guest conversations. They argue that successful decision intelligence depends on more than models, with data quality, testing, deployment, user adoption, and frontline knowledge all shaping whether a system creates value. They revisit examples from healthcare, sports, supply chains, and optimization to show why more data is not always better and why human judgment still matters when models miss local context. The episode also looks ahead to generative and agentic AI, where trust, controls, and clear ownership of decisions will become harder to maintain as more work is automated. Their conclusion is that technology alone does not capture business value. Organizations, incentives, collaboration, and accountability determine whether it reaches production and delivers measurable results.