We are near the end of the hype cycle for artificial intelligence (AI). The human champion of the game of Go decided to retire, saying AI cannot be beaten after AlphaGo defeated him. Domain-specific chatbots are engaging with customers and providing them with the answers they need. AI is about to revolutionize our broken health-care system. Is your company ready for AI?
Anyone with deep data claims to be using AI. Credible pilots and use cases have succeeded in many different sectors. We’re on the productivity end of the cycle and likely to show ROI very soon. In this piece, I’ll share the current state of AI, barriers to entry and a road map for your company.
Breakthroughs Last Year
Major breakthroughs have been driving AI pilots and adoption. Microsoft paid $1B for equity in OpenAI. Samsung enabled the “deep fake” by creating videos of a person speaking using a single photo. Computer vision that helped identify medical issues can now be used to analyze customer, sales associate and crowd movements. Imagine a personalized learning coach helping you work. Most telling is that AI models are beginning to be explained and validated. AI explainability will be a key factor in convincing management that risks can be managed and bold steps taken forward.
Major advances in gaming by AlphaGo and OpenAI, natural language processing (NLP) and near-human text generation by OpenAI’s GPT-2, computer vision and biomedical research by Google DeepMind and data analysis by TensorFlow are opening up very real possibilities.
Barriers To Entry
So, what is the hesitation? Apart from the usual data considerations around security, privacy and compliance, companies are also stymied by a lack of system design, a strategic approach and a general shortage of talent. Challenges around data apply regardless of AI and need to be dealt with otherwise.
Also, all newly hot technologies experience talent shortage and can be addressed with money and time. Here is a list of AI skill sets highest in demand. The real barriers are strategy and system design. Below are stages of readiness and a phased approach for strategic AI.
Readiness And Road Map
Organizations might be in one of the below four stages. Assess where your company is with technology, process and cultural sophistication to plan the next stages.
1. Robotic Process Automation (RPA)
Most companies already have some process automation. Technology is used to shorten repetitive, error-prone, slow human processes. Areas such as customer service, invoice processing, order management, payroll, storing customer or HR information, processing refunds, etc., usually incorporate some automation. Typically these have little basis in data and are conditional processes with low risk that require no greater authority than a manager or a department head to sanction. However, this is the basis for greater sophistication.
While it is possible to go directly to other phases, in my experience, companies that have started here are better able to bridge the technology and cultural gaps that come with the next phases. There is little fear of job loss in this stage and workers are eager to stop doing things manually.
RPA is in fashion today and Covid-19 has forced many companies to supercharge manual processes to accommodate the urgency created. For example, one major airline received 120,000 cancellations in the first few weeks of Covid-19 and built an RPA in six days to reduce up to 6,000 manhours of work by 80%.
2. Intelligent Process Automation (IPA)
Conditional behavior of an RPA can be augmented with intelligence. It is anticipated that by 2022, as much as 75% of enterprise companies will embed AI into processes to discover operational insights heretofore unknown in their processes. The integration of optical character recognition (OCR), NLP, computer vision and chatbots into sales processes will enable one-to-one personalization and more relevant sales. Procurement and finance processes can be tuned much more effectively with some machine learning (ML) driving approval.
This type of automation requires data and training and some expense on AI technology, along with sophisticated skill sets. This is likely to require departmental approval at the VP level and some level of change management. Employees will require training and even retraining. This is the first stage of AI adoption.
3. Point AI Applications
Individual initiatives toward greater automation, optimization and insights within companies will drive the creation of specialized applications of ML and AI. Large departmental operations such as supply chain optimization, customer data insights and customer support operations will create their own solutions for efficiency and new possibilities. Huge data sets will train models to predict one-off outcomes in financials, healthcare monitoring, fraud detection and even employee learning.
In this stage, there is a heavy dependence on sizeable data for training AI models. Dedicated skill sets and sizeable expenses are needed. Some capital expenses might need to be justified to the C-suite. Intra-departmental change management is key to success. Entrusting substantive operations to ML and AI even within a department will require planning for both technology and culture.
4. Strategic AI Platform
When used across a company in a strategic manner, AI can drive never-before-seen success. Only companies that manage to organize and connect large volumes of data across different operational areas can attempt and benefit from this stage. Data lakes and repositories from different departments can be joined in ways AI can exploit to create market differentiating advantages across the supply/production and sales/customer chasm. Add in predictive analysis of market and externals and you have market domination.
Such success will need a constant evolution of employees learning and working in new ways. These are deep transformations and require interdepartmental collaboration, very specialized skill sets and major capital expense. This will require approval from an educated C-suite and board. Such sweeping change can only be accomplished after a company passes through the prior stages.
Depending on which of these stages of automation you are in, it will take two to seven years to reach the full potential of your company. Forward-thinking leadership has already started on this path or, if not, must start immediately.