Sponsored by

At Gurobi, we help organizations make better decisions with mathematical optimization, a proven technology that identifies the best possible solutions to complex business challenges. From supply chains to energy and beyond, our technology powers critical decisions across industries. Explore our 2026 State of Mathematical Optimization Report to see how the field is evolving.
Decisions in the Wild
Soren Little: Optimization Inside the Database
Soren Little, co-founder of OSM Data, is working to make mathematical optimization a standard part of database systems. He studied computer science at the University of Munich, where his interest in optimization began through quantum computing research. After finding that quantum computing offered limited practical value for the problems he was working on, he shifted his attention to conventional optimization methods. He later co-founded OSM Data to help businesses use their existing data to make better operational decisions.
Few believe companies have invested heavily in systems that analyze past performance and predict future outcomes. Yet turning those insights into decisions often requires a separate optimization process. Data must be extracted, prepared, and converted into mathematical models before an optimization solver can use it. OSM Data wants to bring these steps into the database, allowing teams to work directly with the data they already maintain.
The company's proposed architecture, called the Decision House, builds on the idea that databases should support three types of questions. SELECT retrieves information, PREDICT estimates future outcomes, and DECIDE determines the best action given business objectives and constraints. Little wants users to describe what they need to achieve without specifying the mathematical technique used to solve the problem. The system would identify an appropriate formulation and pass it to an existing optimization solver.
Little sees this separation as an important way to reduce the technical work involved in optimization. For example, a company allocating GPU workloads might normally build a mixed-integer programming model. However, some allocation problems can be solved more efficiently using specialized matching algorithms. OSM Data is developing methods to automatically recognize these problem structures, letting users focus on business requirements while the system handles the technical formulation.
The company has explored applications in airline workforce scheduling and industrial battery storage. In the energy sector, Little describes how businesses can schedule battery charging when electricity is inexpensive and discharge during expensive periods. The challenge becomes more complicated when operators also participate in electricity markets. They must balance potential trading revenue against the need to keep enough energy available for industrial customers. Little believes optimization systems must account for uncertain future prices and demand, since decisions based only on average forecasts can perform poorly when conditions change.
Little believes bringing optimization into databases could make advanced decision-making available to organizations without dedicated operations research teams. Data scientists who currently produce reports, dashboards, and forecasts could also recommend specific actions based on business objectives and constraints. He sees this as the next stage in making data systems more useful for everyday operational decisions.
The DecideWise Insider
Benjamin Baer, Founding Member of the DecideWise community, argues that organizations need to rethink how they measure success as technology changes. He uses smartphone battery life as an example: battery capacity alone tells consumers little about how long a device will actually last, which is why Apple emphasizes usage time instead. Baer believes businesses face a similar challenge in measuring performance, decision quality, and operational effectiveness. As AI changes how companies collect and analyze information, existing metrics may become less useful or fail to capture what matters. He points to advances in astronomy, where improved instruments revealed thousands of previously undetected planets, to show how better measurement can change our understanding without changing the underlying reality. Baer encourages decision professionals to examine how AI is changing the metrics they rely on, particularly those affecting the speed of analysis, decision-making, and communication across organizations.
Market Pulse
Profitmind made Awase, its AI product matching model, generally available. Awase reads product images, descriptions, and specifications together to find the same item, or its closest substitute, in competing online catalogs that share no barcode. Batteries Plus now pairs about 11,000 of its products with roughly 75,000 competitor listings, up from about 2,000. Those products account for more than 95% of its consumer revenue, and the retailer uses the matches to decide when and how far to adjust prices.
Finic open-sourced Kairos, a fraud and risk platform that rules on events such as withdrawals, sign-ups, and merchant payouts. Teams frame each check as a plain-language yes-or-no question with optional signals and a policy. Kairos pulls the subject’s recent history, counts how many other users on the same device or payee carry fraud labels, queries Finic’s Grayson model, and applies the policy, typically within a few hundred milliseconds.
On November 4, AWS published a guide showing how smaller decision models can work with Amazon Bedrock AgentCore Evaluations to assess AI agents. The approach uses models such as TypeSafe's Jev and the open-source Strands Decider 2B to check whether an agent's response is accurate, helpful, or completes a user's goal. These models return scores and confidence levels instead of lengthy explanations. AWS says decision models can deliver results in hundreds of milliseconds, compared with 2-10 seconds for larger AI evaluators. Developers can use them for routine checks and send uncertain results to larger models for deeper review.
Fleetworthy introduced RADAR, an AI-powered tool built into its fleet management platform. Launching first in Toll Management, RADAR lets fleet managers ask questions about toll charges and transponder performance using plain language. It also flags unusual charges, spending trends, and potential savings, with recommended next steps. The tool uses customers' operational data to generate insights, while managers retain control over decisions.
Resources and Events
📅 How Knowledge Graphs Help Identify Bad Financial Actors and Commercial Opportunities at Scale (Online - November 17, 2026)
Quantexa is hosting a webinar on using knowledge graphs to detect financial crime risks and identify business opportunities. Lead Data Scientist Matt Jackson and Technical Product Marketing Manager Steve Wilcockson will demonstrate how multi-hop graph analysis reveals hidden connections between customers, transactions, and suspicious entities. The session will cover financial risk detection, customer intelligence, and techniques for analyzing large datasets. Attendees will also learn how to integrate graph analytics into existing applications and decision-making workflows.
📅 60th Hawaii International Conference on System Sciences (Waikoloa Village, HI - January 5-8, 2027)
The 60th Hawaii International Conference on System Sciences brings together researchers and practitioners to discuss information systems, AI, analytics, and digital technologies. The conference includes research tracks on decision-making, digital government, machine learning, and business applications of technology. Sessions will explore how AI, simulation, digital twins, and strategic foresight can improve organizational decisions. The program also includes workshops, tutorials, research presentations, and networking opportunities for academics and industry professionals.
📊 Report Spotlight: 2026 State of Mathematical Optimization Report (Gurobi)
Gurobi's 2026 State of Mathematical Optimization Report surveyed 382 professionals on how organizations use optimization and AI. It found that 79% now use generative AI or large language models in their optimization work, up from 30% in 2025. Another 90% combine machine learning with optimization on at least one project, up from 81%. Meanwhile, 52% use optimization for ongoing business decisions, and 45% run models automatically in production. Respondents identified AI integration (67%) and faster real-time decision-making (57%) as the leading trends shaping their work.
Join the conversation: Can superintelligent AI be trusted to police itself?
Join the DecideWise discussion on whether superintelligent AI can safely monitor and control other superintelligent AI systems. Explore the idea that equally capable AI could prevent dangerous behavior, the risks of relying on AI to police AI, and whether such oversight would be enough to protect society. Share your perspective on who or what should remain in control.
For the Commute
Building Digital Twins Executives Can Trust (Value Driven Data Science)
In this episode, host Dr. Genevieve Hayes speaks with Dr. Greg Biegel, AWS Solutions Architect and author of Digital Twins in Action, about using digital twins to improve business decisions. Biegel explains how digital twins differ from traditional simulations and describes five levels of capability using Google Maps as an example. They discuss how organizations can turn operational data into models that help executives evaluate decisions before implementing them, particularly when mistakes are costly or dangerous. Biegel emphasizes starting with the business decision, then determining the data and technology needed to support it.





