By John P. Desmond, AI Trends Editor
Companies need to align the AI strategy with the business strategy.
First the company needs a business strategy. Can the AI help with that?
Maybe so. AI is being applied to the model decision-making of governments and corporations. A recent article in Forbes described the Real Time Strategy (RTS) technology involved in Google DeepMind’s gaming software that works with “imperfect information.” RTS is a deep neural network trained on past games and which evolves by playing itself to get better.
“The core of the approach is a system that uses a mash-up of supervised learning and reinforcement learning, but the key concept to absorb is that a complicated RTS game only humans could play is now being solved very comprehensively by an artificial intelligence system,” states author Dan Shapiro, PhD CTO and co-founder of AuditMap.ai, offering compliance AI for the enterprise.
Most AI projects today are limited in scope to model one part of corporate strategy, such as credit risk models, recommender systems, and trading algorithms. “Corporate strategy development is going to benefit from more automation, and specifically from artificial intelligence,” Shapiro states. “Companies are organized a certain way, with specific Key Performance Indicators (KPIs), and plans for improving the KPIs. Expect artificial intelligence to creep into corporate strategy discussions.”
Identify the Business Benefit of the AI Project and Make it Achievable
Beware of Bad AI strategy.
“Bad AI strategies are hype-driven, focus on technology over impact, and employ 2–3 data scientists scrambling for projects. Try to stray away from the latter,” suggests Jan Zawadzki in a recent article in Towards Data Science. Zawadzki is a management consultant and data scientist currently working as a Project Lead Data & AI for Carmeq GmbH, the innovation arm of Volkswagen AG.
Each business will have its own AI strategy that might not look like another company’s AI Strategy, but all AI strategies need to answer similar questions. “The core components of any AI strategy concern its holy trinity of data, infrastructure and algorithms, surrounded by the pillars of skills and organization,” Zawadzki suggests.
How to organize for AI is another set of decisions to be made. Do you build an in-house team or outsource tasks? Do you invest in educating management and employees about AI? Andrew Ng, the renowned computer scientist who founded Coursera, recommends building an in-house team, Zawadzki stated. Outside consultants would not know the data, infrastructure, and issues at the company as well as its employees. And learning about AI makes employees more enthusiastic.
Ng also recommends establishing a separate unit which becomes the central enabling point of AI across the company. This unit works with existing departments to find high-impact AI projects and support their development.
Align your organization development processes with machine learning workflow, for best results. Machine learning follows an iterative process, with outcomes far from certain at the outset. Its exploratory nature make it difficult to fit into company-wide goal measurements.
“You can’t promise a working model without thoroughly evaluating the data. Thus, it is difficult to estimate the concrete business impact of AI projects without first investing in ETL and initial data analysis,” stated Rachel Berryman, Co-Founder of todoku.ai, The company offers workshops for non-technical decision-makers to exploit the benefits of data science.
Include a Plan for Putting the Machine Learning Model into Operation
This process of bringing machine learning into the operations requires a shrewd business perspective, suggests Eric Siegel in an article in Machine Learning Times, where he is an editor.
“The most wicked and pervasive pitfall of predictive analytics is organizational in nature, not technical: Predictive models often fail to launch,” he states. Many ideas that seem appealing in the lab can never actually be deployed in the operation. The team may be able to create an elegant predictive model, but the business may not be ready to act on it. Maybe it was lack of management buy-in or unforeseen business constraints.
This is where being shrewd comes in. The team needs to aim at prediction goals that are both achievable and usable.
Oh, and before you get to that, you need to fix the corporate culture and get it focused on succeeding with AI. From the blog of KungFu.ai, a company focused on helping organizations adopt AI, comes a suggestion to invest in six areas: reskilling, incentivized learning, identifying and removing silos, clear communication, ongoing conversations on ethics, and prioritizing diverse teams.
Known reskilling efforts include an investment of $700 million by Amazon to retool a third of its workforce; an experiment by Walmart to offer college for “$1 per day,” an investment by PwC of $3 billion globally to train employees, and a collaboration between San Jose State University and IBM to train students in high tech skills needed for jobs that may not yet exist.
Practices to accelerate learning are being explored; some call it a “scalable learning model.”
The authors outline the choices they see: “You can defend your current position and cease to be relevant. You can try to change the future—but exponential technology is here, like it or not. Or you can create the future you want to see. You get to choose which to accept—but there’s no sitting on the sidelines anymore.”
Read the source articles in Forbes, Towards Data Science, Machine Learning Times and in the blog of KungFu.ai.
