Throughout the previous eight posts in this series, we’ve seen how ML presents a new set of problems and demands a new approach to that of traditional software. In this final post, I’m going to take a look at what companies need to do, not just to survive, but to flourish within the new landscape forged by the advent of ML.
What does machine learning mean for businesses?
100 years ago, when people like Edison and Westinghouse were striving to electrify the US, there were the first few killer apps (electric lighting, the telegram) that generated early buzz, but these only hinted at the effect electricity would have on the world. No one could predict all these changes, but the effects were ubiquitous. Electric refrigeration reinvented the food supply chain. Electric motors restructured societies.
In the beginning, electricity was a complex and dangerous technology. Companies hired chief electricity officers to sprinkle electricity throughout their organizations in attempts to reinvigorate the company’s DNA. President Harrison refused to operate the newly installed light switches in the White House, forcing an aide to follow him through the rooms and corridors of the building, operating the switches for him as he went.
History tells us that companies that failed to adopt the new technology perished. Companies like General Electric that adopted and harnessed electricity survived, ushering in a new era of innovation and profitability.
An equally transformative technology was the internet. The internet changed the way we interact with our friends and family, as well as complete strangers, but it also reinvented the way we deal with businesses. The advent of the internet produced the internet company, with new business models like network effects and platforms. Internet companies like Google and Facebook have a fundamentally new DNA unlike that of any company that came before them. The internet company produced the chief information officer, whose job it is to sprinkle internet technologies throughout the company. And again, the Fortune 500 companies that failed to adapt to the internet in time are regretting it as they begin to fade away.
Machine learning and artificial intelligence will have a similar impact on society as electricity and the internet. The advent of ML brings with it companies with a new DNA. There will again be companies that do not adapt, and they will suffer the same fate as those companies that failed to adapt to electricity or the internet.
The power of machine learning and artificial intelligence presents new challenges. The difficulty with machine learning is that, as we’ve seen, it forms leaky abstractions and weak contracts. The chief data officer cannot just sprinkle machine learning across the organization like the chief information officer did with the internet.
What is cultural debt?
This leads us to the final form of technical debt created by machine learning: cultural debt. When moving from an internet company to an ML company, there is a DNA shift that needs to occur and this is inherently a culture shift.
When you set about testing a website, the process is inherently deterministic. When you use a website, it’s immediately clear if it works or not. Machine learning, on the other hand, is probabilistic. Instead of writing unit and integration tests, you now need statistics and probability to check if things work.
This isn’t just the case for data scientists and ML engineers. This new DNA percolates through to the sales team, who need a new approach to how they demo and sell the product, to the QA team, who now need statistics and probability to test the product, all the way to the executives and the CEO, who need to make decisions for the company in the face of a new world.
A new business paradigm
Machine learning is upending the business landscape, and it’s essential for any successful business to integrate ML into their company’s DNA in order to flourish. The Harvard Business Review just this month highlighted the potential scale and speed at which AI companies can “outstrip traditional firms”:
You don’t have to be a software start-up to digitize critical elements of your business—but you do have to confront silos and fragmented legacy systems, add capabilities, and retool your culture.
While the complexity of machine learning and the problems it presents may at first seem overwhelming, we’ve now seen the 7 main costly surprises that companies commonly encounter when adopting ML technologies and identified solutions at every step of the way. You can employ these solutions to not just mitigate but, if done early enough, avoid these pitfalls. ML presents not just an opportunity to improve company or product performance—the technology represents a seismic shift in the business landscape, one which many businesses will ultimately live or die by.