The top 10 AI enterprise strategy trends for 2022

CIOs, CTOs and technology leaders agree: AI will be the main driver of innovation across industries throughout the next five years. Nearly half of CIOs say they’ve either already started using AI or plan to implement it in 2022, and two-thirds of workers surveyed recently said they want employers to add more AI-based solutions to their workplace.

(AI-generated illustration based on image by Computerizer via Pixabay.)

As more businesses look for ways to implement AI, industry leaders must consider how they’ll prepare for the next big thing. Here are the top 10 AI enterprise strategy trends we’ll see this year.

1. Ethics will be a priority in high-stakes scenarios.

If you’re contemplating building your own AI algorithm for a high-stakes purpose, get your ethics game in order.

Will your algorithm make recommendations about hiring and firing? Granting or denying a loan? Diagnosing a medical condition? In these cases, the risk of inaccurate predictions from a machine learning model is serious, so your team has to be equally serious about the ethical considerations involved. If these use cases have sufficient volume and centrality to your business to justify proceeding with an AI solution, make a careful, ethical investigation of the possible outcomes for all stakeholders and participants.

Start with the trusted resources that are already available in your industry. For instance, the World Health Organization in 2021 released its six key principles for ethics in healthcare-focused AI. Those principles laid out the importance of protecting autonomy, promoting human safety, ensuring transparency and safety and fostering accountability.

And it’s not enough for just your development team to understand the ethics around your algorithm. Make sure your whole organization is part of the conversation.

2. A robust human review process will accompany that ethical dialogue.

Along with a deliberate consideration of ethics, one practice I think we’ll see more of is a heavy reliance on a “human in the loop” process. While piloting and standing up a new AI system, and even in steady-state production, a robust human review process ensures accuracy and goes a long way to improve machine learning. For instance, in support automation, an AI-powered platform should easily escalate to a human expert when a customer question presents ambiguity. Each time the AI reaches for a human source, it learns from the exchange, making the need less frequent over time.

For teams starting out with AI, the human in the loop is crucial; they should only allow the algorithm to take action without review in cases where they have absolute confidence.

3. The use of chatbots will expand to more than a website’s landing page.

We’ll see businesses deploy a chatbot with neural search anywhere their websites or support teams are getting a lot of traffic. The natural language technology for this use case has gone from fumbling to mature in just the past few years, and if you don’t have AI running defense to offload and augment your teams, you are going to get smoked.

Support automation, in particular, gives an edge to helpdesks and support teams that incorporate conversational chatbots. Powered by comprehensive, centralized knowledge bases and able to launch in a variety of communication interfaces in addition to a company website, these chatbots give users the self-serve options they’re seeking, wherever they are.

(AI-generated illustration based on image by Stefan Dr. Schulz via Pixabay.)

4. We’ll see more instances of humans working alongside AI.

AI is quickly diffusing into the abstracted base layer of our everyday life, much like email, e-commerce, mobile and cloud computing trends have done in the recent past. AI behind the scenes powering search, customer interactions and recommendation engines in our favorite platforms.

But at other times, the way humans use it is more prominent and intentional. My favorite example of this is writing computer code with the help of new generative models like OpenAI’s Codex and DeepMind’s AlphaCode. You can literally describe the outcome as a prompt for these systems, and they will generate a functioning scaffold of the code for developers to review and fine-tune for their purposes. It’s an amazing way to amplify a human coder, and other similar models are sure to pop up.

5. New business models will emerge for robots in the physical world.

Robots in the physical world have begun to experiment with new business models. Now, rather than buying a manufacturing robot with a large capital expense, companies are leasing them by paying an hourly wage that is competitive with or lower than what they would pay a human worker. This could accelerate the adoption of robotics considerably in small to medium businesses.

More than 80 percent of CIOs, CTOS and IT directors surveyed agreed that robots would enhance 25 percent of what they do — and businesses would deploy robots across business functions — all within the next five years.

6. The packaging of AI tools will keep maturing.

The packaging of AI tools and services is maturing impressively. Containerization, for example, provides portability, efficiency and agility as pressure mounts for swifter delivery of enhancements and easier scaling of AI tools.

Think about the evolution of any technology. For example: When keyboards first emerged in the typewriter era, there were several competing layouts from alphabetical to ergonomic to letter frequencies. Eventually, the QWERTY layout became a standard with an odd combination of all those factors, and now we don’t give a second thought to it.

Similarly, AI tools are beginning to converge on a consistent set of forms and functions, like cloud databases, containerized models, benchmarks and APIs built with (or wrapped inside of) Python. I’m encouraged by the teeming and productive ecosystem that continues the self-organizing and constant march of innovation in AI.

7. More teams will start small and avoid overdesigning. 

Teams will embrace the wisdom of starting small with AI. Just get something out the door: a proof of concept or a pilot. Because it’s estimated that nearly 90 percent of pilots don’t make it to production, avoid overdesigning.

Instead, prepare to iterate. Be ready to learn from the inevitable mix of successes and failures, documenting each version of the AI to evaluate performance and consider training data and techniques.

8. A renewed focus on the right people to manage data pipelines.

Remember that getting the data, cleaning the data and understanding the data is still hard, hands-on work. Many will promise a plug-and-play, automatic solution for this. In my experience, you still need competent people to put data pipelines together and manage them so that AI tools and services can leverage your data.

A survey of data scientists found they spend about 45 percent of their time preparing data — including loading and cleaning it. Especially in its early phases, AI relies on high-quality data to learn and develop. Putting the right people at the helm of your data pipeline offers the potential to instill a better data culture throughout your organization, promising a better return on efforts.

9.: The role of the chief data officer will evolve alongside operations.

The role of the chief data officer must keep up with the dynamic evolution of traditional data functions, such as storage, architecture, modeling, forecasting, business intelligence and analytics, especially with the current digital-first landscape. As companies continue to implement this critical role in their operations, the chief data officer will increasingly need to be involved in additional areas, including strategy, product, ethics and legal.

10. The macro environment will drive the business case for automation higher.

Over the past few years, we have seen a steady uptick in the adoption of AI and automation to enhance products and services, despite a period of historically low inflation. With labor and other costs forecast to increase more dramatically in the coming months and years, we will see the business case strengthen even more for alternatives like automation, making the flywheel of AI innovation spin that much faster.

Effective AI depends on you

AI is swiftly moving from “nice to have” tech to an essential asset to optimize your business. But it’s not enough to implement AI — successfully deploying AI depends on cooperation across your organization, a focus on the humans involved, your industry’s ethical concerns, and a dedication to constant innovation.

Dave Costenaro
Latest posts by Dave Costenaro (see all)