The potential for AI to deliver variable value is almost limitless. And yet, access to that number is not provided. So how do we decipher the code?
As someone who has been in the business of implementing business-level AI solutions from the earliest days of AI – from within, as a CIO in Verizon, and externally, as a company consultant for AI ASAPP – I know it is our job as CIOs to find value in change technology. And as recently as 2020, McKinsey reported that less than 25 percent of companies “see a significant line-up impact” from AI.
I believe there are at least three ways we need to change our thinking if our organizations are to unleash the full potential of AI conversion:
Instead of: Attempting to bind AI to the way you already work
Try: Designing natural AI systems and structures
Attaching new technologies to existing systems always seems easy. After all, once you have invested so much in building the processes and structures that you have in place, how can it be costly to prevent it? But when it comes to truly transformative technologies (in contrast, for example, the kind of incremental improvements we expect in this year’s processor development), innovation is key to seeing the value of real change. If you want to give AI the opportunity to change the game for your business, you have to let it actually change the game — meaning you don’t stick to the same old game board. Integrating the full potential of AI means baking it in the context of your business operations.
Example: As we make changes to the development of traditional cloud systems, we do things like “lift and move” our applications into the cloud – and ask ourselves why we are not getting all the benefits we were promised, such as continuous delivery and automatic testing. But this was magical thinking. It wasn’t until we realized we had to rethink our life cycle of software development and building programs where we began to see real benefits in our cloud software business. The same thing happens when we move on to mobile app development and make the mistake of thinking we can simply show our web applications on mobile phones. We all know how that happened.
Instead of: Focusing on changing people with machines
Try this: Find out how machines can help people to do their jobs better
When it comes to AI, companies have historically built their options as complete automation, on the one hand, or failure to automate, and hand-delivered to man, on the other. But this barrier ignores the fact that technology in general, and especially AI, has always been about allowing people to do superhuman things, with machines that incorporate an external genre that enhances our skills — as individuals and as groups. By embracing the unique strengths that exist at this crossroads of human and mechanical design, we can open up an unprecedented amount of business. In other words, at work, machines work together, not competitors.
Example: In the customer service, businesses with large call centers have spent decades trying to change points a few percent of their customer service calls – spending billions and even our customers saying they hate what happened and asking to ‘talk to an agent’. Instead, let’s use powerful machine learning methods to learn from the best agents and help make every customer service agent look like yours. The proposal to make 100 percent of your agents even better than the default price is a small percentage of simple changes – and it’s hard to set a price to keep your customers happy, rather than keeping them at arm’s length. The so-called “extension agent” is currently in turmoil, but many businesses are not fully embracing the idea, and many technology providers are not yet focused on how to do this on a scale.
Instead: Relying data only trains our way out of the traps of algorithmic bias
Try: Prioritizing priorities for people in the traditional design table who are less or less representative of data
We know that the racism embedded in the training data causes our AI programs to strengthen gender-based, racial and social and economic discrimination. In addition, the lack of diversity of AI influences what problems are gaining the attention of technology investors, and whether the solutions being developed are sustainable and ethical in themselves. To ensure that we see the full value of AI, we must make it an irreplaceable metrics for our businesses to hire a variety of professional AI staff and empower them to influence critical design decisions. As industry leaders, this means we need to rely on solving a problem called “pipeline problem.” We will not make progress if we simply blame academics for not sending enough AI students to the staff; instead, we need to work with partners in the field to understand the barriers to entry that keep more women, students of color, and students from low-income communities in pursuit of education and careers in AI.
Example: Complete reports such as the AI Now Institute on Discriminating Systems report: Gender, Race, and Power in AI make it clear that we cannot train our way out of a crisis. There is no way around it: hiring a variety of staff with AI expertise and then, critically, empowering them to influence critical design decisions, is something that cannot be negotiated. In my last article, I gave you a guide on how to bring “people to the loop” to fight algorithmic discrimination in your organization.
AI is a powerful technology that can create a number of changes, but to achieve that number, for our customers and our stakeholders, we must be willing to change them. Hiring the same people and following the same business practices that were used in the past will not bring us to the promised land.