How do you know something is at the top of the hype cycle? It’s simple. Everybody’s talking about it, but very few are really using it. That’s the situation with artificial intelligence in the enterprise. If you were to take everything you have read related to AI in recent years at face value, from the doomsday scenarios about impending job losses to the “evil AI” that might take over the world, you’d probably conclude that AI has already been widely adopted in companies around the world.
The reality is that very few companies have deployed it extensively across their operations already (Amazon is probably the most prominent example), but a lot of companies are actively engaged in pilots to figure out where and how AI might benefit them.
Several surveys have indicated that not knowing where to start with their enterprise AI efforts and not having enough data scientists to help them execute on their plans are two of the biggest barriers to wide adoption of AI technologies in the enterprise. This article is our attempt to help companies figure out how to get started with their enterprise AI initiatives. It describes five valuable best practices we’ve learned over the last couple of years by observing, helping and being the visionaries and innovators, who have successfully navigated AI projects through enterprise approval processes.
1. Get a senior executive on board to sponsor
Mid-level executives who are jazzed about enterprise AI aren’t generally enough to ensure eventual success. In many of the cases, AI aims to transform the way the company works, how it is organized and how work is being done. This is not merely a technology initiative, but the start of a business transformation journey.
The level of vision required to weave AI into various aspects of a company’s operations requires a CXO attached to the initiative. This is the cold, hard truth – we’ve witnessed countless efforts fizzle out because the champion wasn’t senior enough.
So, if that’s you, find a CXO, GM, or an influential functional head to play the sponsor role at an early stage. They’re going to need political capital and management resilience to take the organization on a learning journey.
2. Pick your initial enterprise AI projects thoughtfully
The most frequent question we get in our conversations with executives at Fortune 1000 companies is that they don’t know where exactly to start with enterprise AI.
We suggest they start by making a list of the significant operational planning and execution challenges they face on the supply as well as the demand side.
Excess inventory? Delays in assembling and shipping your industrial goods? Not able to predict energy usage in your factories? Trouble forecasting demand? Price bundling? Promotions not effective? Total fleet optimization? This initial list of business issues can then be prioritized by using a variety of analytical frameworks.
In industries where physical goods are manufactured and distributed, we have observed that demand forecasting is a suitable place to start. It’s crucial to give some thought to picking your first project well. You want to score a win, so you can move onto bigger projects that are harder, involve more stakeholders, and will also have a bigger impact on your business. As John Wooden has famously said, “It’s the little details that are vital. Little things make big things happen.”
3. Begin with a small pilot
Unlike traditional projects, where a business case can be made upfront to gain approval and budgets, ROI from an enterprise AI project will only become clear after doing some exploratory work on your data.
Anyone who tells you that they can reduce your costs by x% (or can improve another crucial KPI by a certain amount) before doing this kind of exercise with your data is smoking something, or is hoping that you are.
Based on our experience, we recommend starting with a 3-month pilot that reveals whether the right kind of data is available for the problem you’re trying to solve. Set your goal for ending the pilot with identifying the first business problem that is valuable (and feasible) to solve along with a clear view of the projected ROI. Don’t forget to track what you learn during the exploration about things like previously unknown internal data sources, lack of enough historical data, dirty data, or novel external data sources you hadn’t been using. These learnings are an important intangible benefit that you’ll re-use on your next AI project.
4. Assign the right cross-functional team
This is important because of the potentially pervasive impact AI can have on a company over time. While internal subject matter experts around the pilot topic are key (e.g., revenue managers for a dynamic pricing pilot), it’s also important to involve representatives from other business functions who can contribute to the pilot’s success. AI is going to be very data intensive, so call in Legal, Information Security, Marketing, and HR to give direction on topics like GDPR compliance, data security, and employee training.
Building the right cross-functional team from the beginning will not only accelerate speed of execution, but it will also start building an enterprise-wide awareness and capability around AI. Another obvious success factor is finding the right external partner, one with domain expertise in your industry and solid understanding of data and regulatory requirements.
5. Pay careful attention to change management issues
When you get started on the enterprise AI journey, make sure that people in your company understand that there are many unknowns to discover (data quality, time to train AI, external data sources, etc.) so it will take a few iterations to get the output of the AI to be “correct.” As the pilot progresses, start to manage the change required in people; e.g., if it is a demand forecasting-focused pilot, discuss how to redefine the role of demand planners in the future.
Think ‘machines augmenting humans’ instead of ‘machines replacing humans,’ so the pilot surfaces education opportunities as well as new roles (e.g., AI quality assurance, AI training, etc.) Your internal communications need to be transparent and candid. Let the team involved in the pilot do the storytelling and if there is automation involved, allow the people affected to take on a more strategic role. A big win when it comes to adopting AI in the enterprise involves not only tackling an important business challenge, but also designing a solution that will be accepted by decision makers as an integral part of their workflow.
This article was co-authored with Marcelo De Santis, the CIO of Pirelli and was originally published on LinkedIn.
Enterprise AI is the ability to embed AI methodology into the very core of the organization and into the data governance strategy. I read the above post and good knowledges about Enterprise AI. thanks for sharing the good information.