Making Artificial Intelligence Real
“We need to be an AI-enabled company.” Replace the “AI” with any technology from history and this comment becomes a common refrain across businesses lured by the promises of new technology and fuelled by FOMO (a fear of missing out). As enterprise strategists and former CXOs who have lived through many “technology is the solution, now what was the problem?” conversations, we talk extensively about this issue. To paraphrase Roy Amara, we overestimate the impact of a new technology early on. When it falls short of our expectations, our disappointment means we are less willing to adopt it when it is truly ready. The reality is that often it is only when we change how we operate and organise that many new technologies come into their own.
But what happens when we really combine organisational change and technology in the guise of artificial intelligence? Whilst companies are only starting to see what’s truly possible, they are discovering previously unattainable insights into their customers, employees, financials, environment, new products, and more. The accessibility of cloud-based technology and the nexus of engineering, mathematics, and organisational change promise solutions or improvements to our most complex issues ranging from climate change and poverty to criminal exploitation and disease. Many of our current use cases seem mundane in comparison with what the future holds, but new technology is still fuelling business value into the trillions of dollars. Whether it’s a restaurant’s menu board customised to your locality, precise metrics about your favourite sports team, customised retail or banking offerings delivered to your phone, or the faster discovery of medical interventions, the promise of applying sophisticated technology to data has the power to amaze us and our customers.
So if we take a rational, step-by-step approach to AI, where should we start so that we can truly capitalise on its potential? How do we position ourselves to identify and scale opportunities, while preventing ourselves from going down blind alleys? At the risk of sounding repetitive, many of the steps are the same as with any new technology. Let me share some basic steps any company can take—and are probably already taking—to learn how AI can supercharge their business by fully capitalising on the data they already have.
History is littered with examples of new technologies that were either treated as a dark art that must be left to the wizard we now call the CIO to figure out, or as a magic spell that with a sprinkling of pixie dust will transform an organisation merely by uttering its name. To go back to a similar theme in my blog post about data, and understanding that AI is an enabler, not an outcome, education from the C-suite down is important. Dispelling myths behind AI and foundational components helps ground everyone in the considerations that are needed when thinking of applying AI to a business problem or opportunity. Don’t take for granted that everyone (or anyone) knows what AI and its sub-disciplines are. For the sake of clarity, AI encompasses the use of systems to perform tasks that usually require human intelligence. This is defined in a very narrow way, unlike what is referred to as “artificial general intelligence,” which aims to replicate human behaviour but is just a distant dream. Most AI is based on machine learning to create a model that represents the decision process required often by divining patterns in data.
Helping your stakeholders understand terms like robotic process automation (RPA), data cleansing, big data, model training, ML ethics, and supervised and unsupervised learning helps replace mystery with understanding. It will help your stakeholders hold meaningful, grounded conversations about AI, and focus on the “what” to solve rather than the jumping straight to a technology solution.
Beyond awareness and education, organisations need various skills demanded by AI to be effective. While you might be tempted to go hire a couple of data scientists and expect success, this won’t get you anywhere fast. The skills required extend beyond technology and include deep business domain understanding, modelling skills, and the ability to create, run, and measure experiments.
Partners and external hires can help jumpstart this learning. However, making a concerted effort to upskill the entire organisation rather than making knowledge the purview of an elite new team will take the fear out of the new technology and democratise the ability for every employee to identify opportunities to apply AI. The same techniques we advocate for upskilling employees on the cloud apply to AI. Many organisations already have highly skilled staff members who just need a bit of training to be successful with AI. Training your existing workforce has a double whammy effect of both creating internal talent that understands your business more cost effectively and creating a positive reception from employees who might be fearing for their jobs in this new world. I’m an advocate of leading by example in order to take others with you. In my case, I earned the AWS Machine Learning certification and returned to school to understand data science.
Champion a Cultural Change
Successfully adopting AI hinges on your ability to be an agile organisation. In other words, your organization must be willing and able to experiment and learn fast using cross-functional teams. When talking about the success factors in agile, resilient organisations, I find myself spending about 70% of the conversation on the cultural imperatives. Like any effective technology or initiative, the power of AI comes through its application to real-world problems and the turning of insights into action. Ultimately this requires changes to business processes, operating norms, and even the roles employees fill.
With AI, you need to know how to accept and embrace experimentation. Gone are the days, if they ever existed, of the omnipotent leader with all the answers. Instead, agile, autonomous teams create a culture of learning, run small experiments with AI, measure the results, and then rapidly scale successful experiments, or cost-effectively shut down ones that didn’t work.
Another critical culture change for many traditional organisations is setting these teams up as truly cross-business teams. One recent research paper into companies that successfully scale AI-based initiatives showed that over 90% of them were using cross-disciplinary teams1. While this shouldn’t be a major surprise, it is surprising how many companies still struggle to fully understand the elements of doing this successfully, such as clarity of goals, appropriate career paths and reward mechanisms, and changes to leadership behaviour.
Understand your Data Strategy
Foundational to sophisticated ML models is the need to collect, clean, and make accessible the underlying data. The traditional statistic that is trotted out in these discussions is that well over half of the time spent “doing AI” or even just getting value from data is actually spent on mundane processes such as data acquisition and wrangling. Incremental systemisation of this by, for instance, improving data quality at the source or appointing data stewards is helpful, but foundational elements are equally as critical. Having a cost-effective highly scalable data lake is a good starting point, as is increasing data literacy and data accessibility across your organisation. None of these are silver bullets, though. Ensuring you extract value from the data on every step of this journey helps you maintain confidence and momentum in moving towards a more data-enabled enterprise. I also urge you to have an open mind towards organisational structures that support the use of data. Your chief data officer, if you choose to have one, should primarily focus on making data ubiquitously available and ensuring individual functions and lines of business can extract value from it, not building a new, siloed empire. Similarly, some leaders choose to overcentralise or even over-decentralise activities in more complex organisations, all with the same effect of slowing down data value realisation.
Rinse and Repeat
For the agile aficionados out there, the next step will be familiar. Pick one or more problems that are important to your business, empower your new cross-functional teams to own a desired outcome, measure the results, and address as appropriate. Be wary of making problems more complicated than they need to be or picking intractable issues. Aim to create a virtuous environment of learning and excitement. This is not the time to look for long-term, expensive initiatives.
I recommend choosing those problems where data is readily available, where the potential solutions are complex enough to warrant using ML, and where the end results can be scaled in a business unit when a solution is discovered. Some companies start this process and discover problems they didn’t even know they had by using data, but I suggest starting off with known problems that haven’t been solved sustainably by traditional methods. For example, forecasting and pricing tend to be sticking points for restaurants and retail businesses, but they can be improved with machine learning.
Through each iteration of the experiment, ensure that the teams continue to make progress towards a measurable outcome and that they can show this through data.
Operationalise and Scale
As I discuss in my data blog post, these investments can only manifest value when good ideas are operationalised and scaled. The beauty of working in the cloud is that scaling great ideas is so much easier than it used to be. That leaves the human issue: how do you change processes, roles, and even businesses based on scaling machine learning? It’s a topic that deserves a book to itself, but think on it as you endeavour to harness AI to improve and grow your business. Whatever you invest in, ensure the people most able to implement the results are intimately engaged so benefits can be quickly scaled.
As a leader, be excited about the possibilities that AI opens up, but be grounded and deliberate too. Be clear in your outcomes, push and support your teams to experiment, and learn quickly.
The Data Flywheel, AWS
The Technology Trap, Phil Le-Brun