How to Succeed with AI & ML in 2020

It’s clear that Artificial Intelligence (AI) and Machine Learning (ML) are going to have a huge impact on businesses in the coming years. There is near unanimity among firms about investing in AI and a widespread belief that AI is going to transform businesses. There are estimates that AI will add trillions of dollars to the economy. Yet most companies are still spending money on AI, not making money with it. There is a wide gap between intent or desire to use AI on the one hand and successful deployments to production on the other. Companies – likely including yours – are trying and failing to create value from AI, ML and other kinds of advanced analytics. 

If you want to succeed with AI and ML in 2020 you need to add a new phrase to your lexicon – digital decisioning. 

What is Digital Decisioning?

Your business must make choices, business decisions, on a regular basis. The decisions your organization makes have an observable impact on the behavior of your customers, your organization and on its results. A digital decision is a decision made by a computer.

Digital decisions are not executive or management decisions but repeatable, operational decisions about a single customer or a single transaction. While each individually of limited impact, you make so many of these decisions that the cumulative impact of even small improvements in these decisions can fundamentally change your organization. Digital decisions are a key element of a digital business, allowing customization at scale, self-service, and real-time responses. Digital decisioning is especially effective where you have to handle large numbers of transactions, when it’s not easy to decide how to handle them, and when responsiveness is critical.

Digital decisioning delivers a return on your investment in your data by operationalizing machine learning and artificial intelligence algorithms. It uses business rules to guarantee agility, transparency and compliance. And it supports continuous learning and improvement.

How does Digital Decisioning drive AI and ML Success?

First, let’s identify the key reasons AI and ML projects fail. Lots of surveys have been done on this but two main things come up again and again.

1. The AI or ML project is focused on the wrong problem. Defining or framing the problem and aligning the ML or AI being developed to the business is hard. Projects that start by analyzing data and trying to “find something interesting” create insight that cannot be meshed with how an organization actually operates. There’s no alignment between the insight being generated and real business problems.
2. Assuming useful insight is found, deploying ML or AI into production systems and front-line workflows is hard. These “last mile” problems mean potentially useful insights don’t get deployed and used as often as they should. The insights may be great, but they don’t change the way the organization decides.

Both of these are issues with making AI or ML actionable – creating an AI or ML algorithm that can be actioned and then ensuring it really is actioned - used in production.

Digital decisioning addresses both these challenges by focusing on the goal – improved decision-making. Digital decisioning has a decision as its goal, not an algorithm. The focus on digitization puts production integration and operationalization front and center. Digital decisioning focuses AI and ML projects on business outcomes and puts the last mile first, getting teams to think about how to put the results into production right at the start.

What are the Keys to Success for 2020?

DecisionsFirst Thinking. 

Success with digital decisioning – and therefore with AI and ML - requires a focus on the decision-making itself. You’ll get the best results if you think about your decision separately from the processes it influences or the data that supports it. Putting decisions first helps establish decisions as first class citizens and ensures that your AI and ML investments are focused on improving a decision.

Decisions provide the business context for your AI and ML investments.

Experience working with thousands of business decisions around the world shows that the most effective way to focus on your decision making is to work directly with the business owners to build a decision model. Decision models, like the one shown below, let you break down even very complex decisions into easier to manage and describe components. They show how pieces of decision-making can be reused. They show you exactly what data you need, and why. And they enable you to more easily identify the policies, regulations, best practices and analytic insight – the knowledge – that you need to make the decisions you want to improve.

Because decision models can be built for any repeatable decision and because they don’t require you to pick a technology, you can always begin with decision modeling. Decision models provide the blueprint you need for success with digital decisioning and in the application of AI and ML to real-world business problems.

Mix and Match Technology.

The second key to success is to mix and match the available technologies. It’s not about using deep learning or replacing “old” business rules with “new” machine learning. It’s about solving your business problem in the most appropriate and effective way. And when your business problem is decision-making, a wide range of technologies can be useful. Business rules and a Business Rules Management System (BRMS) are effective for managing decisions driven by regulations or where you have to capture human judgment in a manageable format. Data mining and predictive analytic techniques applied to data can be used to create analytic models that make decisions – predictions – in ways that are easy to explain and document. Modern machine learning techniques can be applied when there is a lot of data and where an explanation of why you decided the way you decided is less important than how precisely you decided. Optimization and solver technology can be applied to decisions that require tradeoffs between complex options. And people can be relied on to make decisions too.

All these different approaches can be applied to specific decisions in the model. The decision model shows how to orchestrate these technologies under a decision umbrella, allowing you to deploy a coherent decision service for use in your IT architecture. And when you need to include people in the decision-making, the decision model lets them focus on specific pieces of decision-making and avoids the inefficiency of simply throwing the whole transaction “over the fence” for manual review.

It’s not about using ML or AI, it’s about solving a business problem. Pick the technologies that will best solve your problem.

Continuous Improvement.

The final key is to focus not on how good version 1 of a digital decision is but on how well you can improve it. Experience shows that the journey matters. The rate of improvement and the engagement of the business in continuous improvement are what drive success. It matters less what the first version does.

In fact, focusing on making version 1 better often has negative side effects. The more the first version tries to do, the harder the organizational change will be. The more it does, the riskier it is likely to be. And the more it does, the more accurate any ML or AI will have to be, increasing your upfront investment and delaying your return.

It’s much more effective to get a minimum viable digital decision deployed and to invest in making sure you generate good data about how that digital decision was made. Capturing a rich set of information about how your system decided allows the business to review decisions made and identify ways to improve them. Tying decisions made to business outcomes allows for targeted improvements to business rules, analytic models, ML or AI algorithms. Regular updates empower the business to really own their decision making and think analytically about how to use data to get better.

Adopting Digital Decisioning in 2020

2020 will be a make or break year for many of you when it comes to adopting ML and AI. Will you start to show an ROI or will you find that you are continuing to spend with just science experiments and pilots to show for it? As John Rymer and Mike Gualtieri of Forrester Research say in “The Dawn of Digital Decisioning: New Software Automates Immediate Insight-To-Action Cycles Crucial for Digital Business”

“Enterprises waste time and money on unactionable analytics.

Digital decisioning can stop this insanity”

Digital Decisioning is your path to success with AI and ML. Digital Decisioning and using Decision Management is how you can deliver business impact from AI. And that, it turns out, is the title of my new book, a completely updated version of my most popular title.

But don’t take my word for it – check out the reviews and decide for yourself.

“…the only approach that has actually allowed me to operationalize predictive models and deliver real ROI!”

“Essential reading for COOs looking to rigorously improve automation through AI.” 

“Anyone trying to automate and embed analytics to support decisions should read this book.”

“Nothing but solid knowledge, sage advice, and great examples without an ounce of hyperbole or fluff.”

Forewords by Tom Davenport and Eric Siegel.

Connect or drop me a line if you'd like to discuss or if we can help.