Watch all the Transform 2020 sessions on-demand here.
Artificial intelligence is having a moment across the business world. Data and technology have become valuable differentiators for many industries, and companies that want to remain on the cutting edge are exploring the potential of what AI can do for them.
And, as with many buzzed-about business trends, many enterprises are throwing around the term “AI” without really understanding what it is, what’s needed, what it’s actually capable of, and how it’s different from data science. AI is not something a company can simply add to its business and expect immediate breakthrough results. It’s also much more complicated than simply hiring one person with a Ph.D. to oversee an AI project and expecting them to succeed.
While some highly innovative companies have devoted resources to getting their data sorted and organized, most haven’t, in which case AI isn’t a realistic next step. For true AI success, your enterprise must consider the following three things.
1. Determine what AI is good for and what it isn’t
AI works best when a business has a clearly defined problem that also falls within the class of problems that current AI processes are suited to. AI is not going to fix nebulously defined issues that require emotional intelligence or a lot of context. What AI is good at is solving problems with a lot of data and/or complicated math, like diagnosing diseases or testing and optimizing advertising creatives.
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
One common misuse that surfaces across industries is the idea of using AI to improve the “sales productivity” of internal B2B sales teams. In one instance, I’ve seen a company consider AI to help identify the characteristics of a successful sales rep, which it would use later in the hiring process. Because hiring is a complicated, ambiguous process that often relies on context, it’s a poor fit for AI. While an organization may do a lot of interesting, and even accurate, analysis, this is unlikely to result in anything applicable.
A better fit for AI is to identify certain patterns that enable sales reps to be more successful. For example, a business may find data that suggest prospects called at least three times without a response are less likely to respond in the future. A rep armed with this knowledge will ultimately spend less time on low-potential prospects.
Before seriously pursuing AI and hiring a team of experts, a business needs to look at how similar companies have successfully implemented AI into their operations as examples of what’s possible. Then, it must outline specific problems, with a sense of how AI can solve each of those problems. As with any new, innovative project, you should always start small and iterate rather than try to “boil the ocean” with a massive first project.
2. AI = Massive data = Massive infrastructure
Amid the AI hype, many companies are chasing talent without first assessing their internal data infrastructure. Without this in place, the project is likely to be a bust before it ever gets off the ground.
The first step for any organization is to make sure it has the data that supports its goals, or at the very least has plans to acquire the necessary data. Then, the company must ensure that data can be related to the relevant context. In the sales example outlined above, the company would want to make sure that all sales rep actions can be related to both the rep involved and the end sale.
From there, the company must store the data for appropriate accessibility. Data is only useful if the analysts and data scientists can access and use it, via SQL-like querying or .csv data dump. Finally, the infrastructure must be able to handle the volume and frequency of the data.
In most cases, a business needs an additional full-time data science and/or data engineering resource to ensure it meets these requirements, and possibly even to model, clean, and manage the data before the company can use it to build an AI system. This may be a hire that the company makes concurrently with the AI lead, or even well before the AI operation officially starts.
3. You need doers, not just thinkers
The most difficult part of this process is actually evaluating the talent for AI leadership. Because AI is so new, few people know what an ideal candidate actually looks like on paper or in practice. It’s possible to hire a mathematical genius for the role, but sometimes those kinds of thinkers have trouble expressing their ideas in layman’s terms or code.
The best data scientists can balance the academic side of AI with the applicable. They are practical thinkers who can roll up their sleeves and get a good amount of work done by themselves. Early on, the AI lead will be very busy dealing with coding bugs, data issues, and performance tuning, not writing papers and presenting at conferences. Hiring someone with an impressive academic resume can backfire if that person is unable to rise to the task or produce applicable work. In the end, their attitude and inaction will only demoralize others.
It’s therefore important to conduct a practical interview where the candidate demonstrates their skills coding and solving real problems. Implement this practice by enlisting someone within your network who has experience with AI and ask them to oversee the practical portion, as it can be difficult to evaluate a set of skills if you are not also an expert in the field. In order to make sure you get this very important hire correct, it might also be worth paying a trusted third party, such as a consulting firm, for help.
Once a company has completed the hiring process, it will still need to align its AI team with the overall company strategy to ensure it’s on the way to achieving its goals. One option is a consultative model, where a centralized data science team consults various departments on relevant initiatives. Companies can also “embed” data scientists within departments, such as marketing or engineering, so that they are part of cross-functional teams with clear goals. Before doing this, it’s important to understand the different AI and data roles. Data analysis is different from data engineering, and those disciplines will have an impact on how those individuals and their work fit within other teams.
Overall, the opportunity to implement AI will always depend heavily on the nature of the business. But as companies across different sectors continue to rely on technological infrastructure and generate massive volumes of data, there will be more opportunities for optimization through AI. A company considering adoption needs a clear idea of what it will use the AI for and how its team will use those outcomes across the rest of the business. If a company starts by outlining these goals and then examining its existing technical capabilities, it should be able to set a new AI hire up for success.
John Merryman is the chief technology officer at Goodway Group, a programmatic partner to agencies and brand marketers.
