Management Artificial Intelligence (IntelliSys conference)
I presented Management AI architecture at IntelliSys London. It was one of the best conferences I have attended, short interview: https://youtu.be/oqK4vtdA-tw
Management AI (Mgmt-AI) aim is to improve organization performance by advising optimal management decisions. Mgmt-AI connects business economics to the behavioral science by game theoretical approach. It simulates the reality, thus it is important that the architecture is based on bona-fide science (not just AI black box!). The architecture includes theories of human capital production function and the theory of quality of working life index (QWL). QWL is intangible production parameter and enables the connection to Bayesian Game theory.
The Mgmt-AI includes real data from the company (business, HR, and staff survey). In addition, there are several evidence-based deductive rules and efficiency factors, which form the base for Neural-Network (NN). In the future, the NN verifies organization and leader specific factors and causalities to the simulation, making it even more accurate. Simulation includes 32 best management practices that form the action space matrix. The state space has 27 workplace problems. Each problem has certain tendency to reduce QWL and management actions have tendency to improve QWL. Each management action requires working time, thus decreasing the time for making revenue and profit.
The mgmt-AI simulates one year (12 months) scenario where problems illustrate the market situations (cash-cow, recession, growth). One simulation round takes about 20 minutes, so forms optimal micro-learning session. Mgmt-AI includes the management “DNA” where the core causalities are simulated without normal noise factors of real-life management. The idea is that agent gets one-year management experience in 20 minutes. The manager can also learn from mistakes without the fear of remorse and will experience the positive results of improved management policy.
The executive management may cause harmful social dilemma at supervisors. Short-term monthly profit requirement may force the leader to choose a strategy that minimize the collaborative management activities to maximize workers’ time for work. The leadership’s social dilemma is harmful when leader neglects the workers’ signals. The leader may know this profit maximizing strategy will reduce the workers motivation. Neglecting workers opinions may eventually change the game from cooperative to non-cooperative mode, where workers do not give verbal signals from work-related problems. The sustainable competitive advantage is very difficult to achieve at non-cooperative mode. Without workers signals the leader cannot choose optimal HR-actions to prevent a decrease in the QWL. When a negative spiral occurs, the leader will have to make decisions solely by the fiscal data from the past.
Management AI simulation is the most precise people analytics, because it
– is based on Human Capital Production Function that include QWL as a production factor
– consists deductive rules and causalities that are evidence-based and can be verified organization specific by Neural-Network
– includes market situation, strategy and organization problems meaning to business performance
– includes management behavior policy meaning to business performance
In the simulation when a player’s learning is jammed, AI will help choosing the most optimal management actions. AI advice is based on Markov algorithm with Bellman advantage function. It encourage the investment-based thinking where some monthly profit is wise to sacrifice to gain more profit later. AI can give timely suggestions and so create a guided learning experience. Solving the workplace problems in time, the leader maintain basic performance level in the team. Higher performance requires also systematic HR-practices that nurture employee motivation and innovativeness. In this way, the company can achieve competitive advantage where both well-being and business performance flourish.
Kesti, M. (2018). Architecture of Management Game for Reinforced Deep Learning, Intelligent Systems Conference 2018 6-7 September 2018 | London, UK.