Guest Post By: Prashant Momaya, Senior Director, Solution Engineering, India, Tableau at Salesforce
We’re seeing AI projects shift from hype to impact, largely because the right roles are getting involved to provide the business context that has been missing previously.
Domain expertise is key; machines don’t have the depth of context that people have, and people need to know the business and data well enough to understand which actions to take based on any insights or recommendations that are surfaced.
When it comes to scaling AI, many leaders think they have a people problem—specifically, not enough data scientists. But not every business problem is a data science problem. Or at least, not every business challenge should be thrown at your data science team. With the right approach, you can reap the benefits of AI without the challenges that come with traditional data science cycles.
To deploy and scale AI solutions, leaders need to shift the organization’s mindset to think of AI as a team sport. Some AI projects need a different set of people, tools, and expectations for what successful results look like. Knowing how to recognize these opportunities will help you approach more successful AI projects and deepen your bench of AI users, adding speed and power to decision making across the workforce. Let’s explore why and how.
Organizations are democratizing advanced analysis with AI
Using AI to solve business problems has largely been the purview of data scientists. Often, data science teams are reserved for an organization’s biggest opportunities and most complex challenges. Plenty of organizations have been successful in applying data science to specific use cases like fraud detection, personalization, and more, where deep technical expertise and finely-tuned models drive hugely successful outcomes.
However, scaling AI solutions through your data science team is challenging for organizations, for many reasons. Attracting and retaining talent is very expensive and can be difficult in a competitive market. Traditional data science projects can often take a lot of time to develop and deploy before the business sees value. And even the most experienced, robust data science teams can fail if they lack the necessary data or context to understand the nuances of the problem they’re asked to solve.
The 2021 Gartner® The State of Data Science and Machine Learning (DSML) report states that “client demand is shifting, with less-technical audiences wanting to apply DSML more easily, experts needing to improve productivity and enterprises requiring shorter time to value for their investments. While there may be many business problems that can benefit from the speed or thoroughness of analysis that AI can provide, a traditional data science approach may not always be the best plan of attack to see value quickly. In fact, the same Gartner report predicts that by 2025, a scarcity of data scientists will no longer hinder the adoption of data science and machine learning in organizations.
Domain expertise is critical for scaling AI across the business
AI is already helping bring advanced analysis capabilities to users who don’t have data science backgrounds. Machines can select from the best forecasting models and algorithms, and underlying models can be exposed, offering the ability to tune them and make sure that everything matches what the user is looking for.
These capabilities give analysts and skilled business domain experts the ability to design and leverage their own AI applications. Being closer to the data, these users have an advantage over many of their data scientist counterparts. Putting this power in the hands of those with domain expertise can help avoid the lengthy development times, resource burdens, and hidden costs associated with traditional data science cycles. Plus, folks with domain expertise should be the ones to decide whether or not an AI prediction or suggestion is even helpful.
With a more iterative, revise-and-redeploy model building processes, people with business context can get value from AI faster—even deploying new models to thousands of users within days to weeks, instead of weeks to months. This is especially powerful for those teams whose unique challenges may not be a high priority for data science teams, but can benefit from the speed and thoroughness of AI analysis.
However, it’s important to note that while these solutions can help address the skills gap between analysts and data scientists, it’s not a replacement for the latter. Data scientists remain a critical partner with business experts to validate the data being used in AI-enabled solutions. And in addition to this collaboration, education and data skills will be critical in using these kinds of tools successfully at scale.
Data literacy empowers more people to leverage AI
Your foundational data strategy plays a huge role in setting up your organization for success with AI, but bringing AI solutions to more people across the business will require a baseline of data literacy. Understanding what data is appropriate to apply to a business problem, as well as how to interpret the data and results of an AI recommendation will help people successfully trust and adopt AI as part of their decision making. A shared language of data within the organization also opens more doors for successful collaboration with experts.
McKinsey’s latest global survey on AI revealed that within 34% of high-performing organizations a dedicated training center develops nontechnical personnel’s AI skills through hand-on learning, compared to only 14% of all others surveyed. Additionally, in 39% of high-performing organizations there are designated channels of communications and touchpoints between AI users and the organization’s data science team, compared to only 20% of others.
Leaders can take a variety of approaches to build data literacy, from education and training, mentorship programs, community-building data contests, and more. Think about normalizing the access and sharing of data, as well as how you celebrate and promote successes, learnings, and decision making with data.
Data literacy and education about visualization and data science needs to be more prevalent, and taught sooner. There’s a sort of social and organizational responsibility that comes with the reliance on using data. People should be better equipped to understand, interpret, and make the most of data because AI will only get more sophisticated, and we should be a few steps ahead of the game.
Continuing to build your organization’s data culture creates powerful opportunities to nurture skills and foster new solutions across the business. Many organizations have already increased their investments in data and analytics in recent years, as digital transformation has accelerated. It’s not a reach to think of data as a team sport—and now we have the means to extend that mindset to AI.