Narrow AI for Decision Intelligence
Can a business that doesn’t use Artificial Intelligence (AI) survive the competition?
AI beat humans at chess years ago, then at the more complex games of Jeopardy and Go. How long before it beats business managers? The complexity of running a business stands in the way. On the side of AI are the forces of data accumulation and algorithms that convert human expertise into software. AI is already available in many parts of the business, optimizing inventory, scheduling work, etc., and now it is optimizing decisions across silos and business units.
The game is on, let’s call it Decision Intelligence. Who, or what, will make better decisions to run a business? AI, humans, or a hybrid?
I’m betting on the hybrid. A community of professionals is arrayed towards putting AI to work on helping with business decisions, and a hybrid approach enables people to leverage the cold-blooded analytical skills of algorithms. The hybrid combines the intelligence of human decision-makers with the AI’s analysis of the data and implementations of decision modeling.
So, are we rebranding Analytics?
No it’s not a rebranding. Decision Intelligence provides a systematic way to improve decision-making. Think of it as the fourth kind of analytics system.
This is the logical next step for professionals who are already engaged in building decision models and providing decision analysis and support. The technology to enable Decision Intelligence systems has become cost-effective in terms of compute and storage, making it feasible to jump from special-purpose “expert systems” that automate a few decision-support models to enterprise-scale systems that coordinate and optimize decision-making across silos.
The premier professional association for these professionals, INFORMS, recently updated its vision and mission to point to Decision Intelligence:
- Vision: better decision making for a just, prosperous, and sustainable world.
- Mission: INFORMS advances and promotes the science and technology of decision making to save lives, save money, and solve problems.
The improvement of decision making is in the spotlight, and the science and technology of decision making must be deployed as the organizational capability for Decision Intelligence. This capability is a combination of narrow AI and Data Supply Chain technologies delivered in AI platforms such as FrogData, along with the people and processes required for continuously improving decisions.
What is Narrow AI
Artificial Intelligence (AI) systems perform tasks that typically require human intelligence, for example: planning, problem solving, and decision making. The management functions that strive to be objective and goal-seeking intend to remove human frailties from the decision process and align to algorithmic approaches. Narrow AI is an algorithmic approach that uses software to accomplish specific pre-learned problem-solving tasks. It does not use human-like cognitive abilities, awareness, or consciousness. It stands in contrast to General AI, a machine that must also be self-aware and conscious.
A narrow AI has the problem-identification and problem-solving methods used by expert humans in performing a specific job. It beats human performance on three fronts. First, it has the combined knowledge of all its teachers. Second: it reliably and consistently applies that knowledge. Third: it can transparently inform decision-makers about its limits and methods so that people can assess when and how to override its advice.
Building a Narrow AI to provide Decision Intelligence
To enable a business to run with intelligent decision-making, we need to convert decision-making into a science. Each decision has to be precisely defined, so that it can be run as a software algorithm (a narrow AI). Decisions, however, are made in a decision process in a business context, both of which must also be scientifically defined.
Such conversions of behaviors into software algorithms have occurred before. Transaction automation gave rise to tabulating machines, pick/pack software, and call-center automation. Business process automation connected sets of transactions into threads that stretched across organizational boundaries into supply chains and business ecosystems. In the next step, to decision intelligence, we must define decision-making in a scientific way. This is done by the use of decision models embedded in decision cycles that drive the decision layers of strategy, planning, scheduling, and execution.
The Decision Cycle
The decision cycle is the scientific algorithm for improving decision-making. Start with a business need: how can the business be improved? Which decisions drive the results? Select a decision to improve. Then define, design, develop, and use the decision cycle.
The decision cycle puts the focus of analytics on creating value from analytics. It is a continuous improvement cycle. Data is managed in a hub that supports the cycle.
The decision cycle is the science of applying a Narrow AI for making better decisions. The technology for implementing it requires a melding of the data supply chain and the decision cycle.
The Data Supply Chain
The Data Supply Chain includes sourcing, moving, and transforming data to enable the Decision Intelligence functionality. Data supply chains can operate at different scales, each enabling a different level of knowledge-sharing into the AI. First, a department-level data supply chain. Second: the enterprise data supply chain that provides a cross-functional data for an entire business entity. Third: an industry data supply chain that can enable pooling of expertise across business entities.
In the data supply chain, data flows from source systems, is aggregated in the data warehouses and marts, and then flows out into five kinds of uses.
Data exploration, reporting (filter, aggregate, slice/dice, visualize, etc.), and algorithm development are costs that the decision intelligence capability must incur to provide the decision support and quality data that drives value.
People, Processes, and Roles
Analysts and software engineers collaborate to build and run the Decision Intelligence system, in such a way that human and machine strengths are best aligned to the roles assigned.
People can use tools to quickly make sense of data and determine the right ways of analyzing, modeling, and presenting it. Machines can enable people to be better and faster at their jobs, and then take over the parts that are amenable to repetitive execution and machine learning.
Decision advice is a hallmark of Decision Intelligence
How can you recognize a Decision Intelligence system? It must provide decision advice, as opposed to the mislabeled Business Intelligence systems that enable analysts to build tables and graphs as they like. There can be no decision without a set of feasible options. Advice consists of evaluating (scoring) the options against decision criteria, and providing recommendations based on those scores. A set of scenarios may be presented in the cases where the recommendation changes because of a change in the scenario (e.g., for inventory levels in different market growth scenarios).
scores, and recommendations
The advice is provided to the people who are involved in the decision cycle, when needed in the decision process: on schedule, on user request, or triggered by an event that requires that decision.
Decision Intelligence systems orient around the central idea of supporting a set of decisions, and thus they will address a list of decisions and support each decision with a set of reports that address the needs of the decision makers in evaluating their options, arriving at a decision, and monitoring outcomes.
Decision Intelligence systems promise the benefit of consistently making the best decisions, across the organization, at scale. We will see this capability grow from the current implementations in teams and groups to become pervasive, a routine part of the technology infrastructure of organizations. This is the path blazed by earlier technology waves such as databases, spreadsheets, book-keeping, ERP (Enterprise Resource Planning), and CRM (Customer Relationship Management) systems.
Decision making in the age of AI becomes a collaboration between human and machine intelligence.
