Artificial Intelligence (AI) holds great promise for creating super-optimized selling, but what does it actually take to make it a reality and deliver those results?
In previous blogs we looked at the potential of Artificial Intelligence (AI) for sales and how AI can be used to drive sales performance. Great potential, but at what cost?
According to a Teradata report on AI, 30% of companies believe that their organization isn’t investing enough in AI technologies and will need to invest more over the next 36 months to keep up with competitors in their industry, with 91% expecting to see the following barriers to AI realization.
Barriers to AI
In a recent webcast, Lily Scanlon from Korn Ferry’s Sales Force Effectiveness Practice described the barriers to AI adoption. According to Scanlon, it boils down to three areas: talent, data, and bias.
Finding and Organizing Talent
When it comes to AI, talent can be in scarce supply and expensive. Applying AI requires data scientists. Per the Teradata survey, only 28% of respondents believe their organizations have enough trained people internally to buy, build, and deploy AI. Thirty-four percent cite a lack of access to talent and understanding.
Yet simply hiring some MIT/EPFL grads is not enough. A company needs an underlying strategy that will support the teams who are operationalizing the AI. So, first the organization must have the right strategy and culture to help people be successful, be clear about what they are trying to do, what questions are they trying to answer, for what purpose, in what environment, and how these data scientists will be supported.
That means there is a strong change management component to AI as well as an educational component. Existing resources in IT and on the business side need to understand what AI can do and where it makes sense, so that they can identify opportunities where AI can have an impact.
Then comes the challenge of finding scarce talent. Companies may use their rewards strategies to draw the right talent, others may decide to develop internal talent, or still others may opt to buy the expertise wholesale. For example, Salesforce acquired Tableau, perhaps as much to get the data science talent as the technology.
Ensuring Data Access, Quality, and Safety
Data seems to be headache no matter what type of IT project you have in mind, but with AI, it’s on a different order of magnitude. Some key data challenges include:
- Amount of data: the Prescriptive and Predictive approaches rely on access to data— and lots of it. Massive data sets are required to “train” the AI system (on the scale of trillions of data points).
- Defined processes: the Robotic Process Automation (RPA) approach requires a definable process and decision tree, which may take significant work to produce (and requires access to data as well).
- Preparation: a massive amount of data prep is needed; anyone who works with data knows that to consume data for a purpose, it needs to be clean and in a form usable for that purpose. For AI, you also need to write code and train algorithms to tell the system what to do with the data and make sound decisions. In short, the end result can only be as good as the quality of the data being fed into the system and the quality of the algorithms being applied to that data.
- Managing Risk: With data comes great risk: how do you protect against data breach, and how do you respond when there is a breach? Depending on the types of data involved, you may need to address security or privacy issues related to the data, requiring robust data maintenance and governance processes.
- Process optimization: this all has to fit within your bigger picture processes; you will need to integrate the AI models into your business process workflows.
Avoiding Built-in Bias
While a strength of AI is that it can take the human bias out of decision making by using data to drive decisions, AI also is susceptible to the risk of propagating a built-in bias. Why? As noted above, AI relies on massive data sets. But data by definition reflects the past, so past prejudices will tend to be re-created.
Amazon discovered this when it trained an AI system to select job candidates. It trained the system on resumes received over a 10-year period. As a result, the selection criteria replicated the biases around gender and race displayed by the real-world tech industry over the previous decade. Efforts to mitigate the historical bias failed, as the AI system was smart enough to figure out that an applicant was a type to be excluded (e.g., a woman) using proxy clues, even if admins told it not to penalize women per se.
Why do AI projects fail?
To know how to make an AI project succeed, it’s useful to understand why AI projects fail. Of course, any project needs to overcome the barriers just described. Beyond that, however, the most common reason such projects run aground, according to Scanlon, is that companies try to do too much, to “boil the ocean” as it were. An alternate approach that is often more practical and successful is to seek incremental, quick wins. This means identifying specific opportunities — the low hanging fruit — and using the right technology for the job.
No doubt the barriers to AI will reduce over time as applications are developed to deliver solutions that are more accessible to business users. Already, purpose-built AI functionality is beginning to be productized; for example, Salesforce CRM has Einstein to do predictive lead scoring and forecasting. But as always, the key to success will be using the strategy and technology that gets you the most bang for the buck, the best and fastest ROI.
For more insights on using AI and a data-driven approach to enhance sales performance, watch the free on-demand webcast Can AI Transform Sales Performance Management?
