Executive Q&A: New Survey Reinforces the Importance of Data Science and AI/ML
The results of a new study by Domino Data Lab confirm the importance of data science and advanced analytics to modern enterprises. The company’s head of data science strategy and evangelism, Kjell Carlsson, drills down on the survey results.
- By Upside Staff
- August 8, 2022
Upside: A recent survey conducted by Domino Data Lab found that nearly four in five respondents agreed that data science, ML, and AI are critical to the overall future growth of their company. In fact, 36 percent said these were the single most critical factors. That’s a pretty strong statement. What’s driving such belief in these technologies?
Kjell Carlsson: It’s worth noting that these numbers are similar to the surveys I’ve run in the past. I did a survey at Forrester in 2021, where 25 percent said data science was the single most important factor for their competitiveness and expected that to rise to 51 percent in the next two years. In two surveys I did that year, 21–25 percent said data science/ML/AI was their largest investment area, rising to 49–54 percent in two years.
Arguably the reason why there are such expectations around data science is that more and more companies are seeing the results of their data science initiatives. In large part, this is because companies are becoming more mature in their use of these technologies and their data science teams are becoming more established, integrated with, and supported by the rest of the organization. In the surveys I’ve run, companies regularly report a 4–5x ROI from these investments. However, there is a significant (30 percent) gap between leaders and laggards and that divide is growing.
Management’s expectations for bottom-line benefits are also growing rapidly. Nearly half of respondents to your survey said their company’s leadership expects data science efforts to produce double-digit revenue growth -- which is up from just 25 percent in a similar survey you conducted last year. What accounts for this big leap in expectations?
Again, this is (happily) due to companies seeing results from existing projects. There has been a leap forward not just on the tools and technology side but, more important, on the people and process side. Organizations have been able to bridge the chasm between developing data science solutions and deploying them, which has required better integrated MLOps platforms as well as alignment between data science, data engineering, operations, and line of business leaders. The problem companies are running into is being able to scale these successes. Many teams are finding themselves victims of their own success in that they are now spending more time maintaining existing models and projects and struggling to take on new ones.
What challenges are enterprises facing when it comes to data science according to your survey?
When it comes to data science, there are several challenges that enterprises face.
The majority of respondents found that the most challenging technological issues related to scaling and operationalizing data science were accessing appropriate data science methods and tools (27 percent of respondents rated this their most significant challenge) and security considerations (26 percent of respondents rated this as their most significant challenge).
When it comes to people- and process-related challenges, most respondents ranked having enough data science talent (26 percent) as the most significant challenge. It’s no exaggeration that every fast-growing organization needs more data scientists -- they play a critical role in turning raw data into innovative new products and services.
However, it’s also important to understand that the notion of a “normal” data scientist is a myth. Today’s data scientists come from a wide array of backgrounds, ranging from computer science to applied physics, so when looking to hire and recruit data science talent, organizations should be ready to cast a wide net.
Speaking of staffing, having enough data science talent has been a problem for some time. Your survey found that over a quarter of respondents rated this as their most significant people-and-process challenge (implementing and/or managing infrastructure was a close second). How do you see enterprises coping with this talent issue?
One way for organizations to cope with this talent issue is through inclusivity. Every organization needs to embrace the notion of diversity of data scientists because there are not enough of them with a specific profile to hire exclusively. To start, they must tackle the people, process, and technology aspects of data science diversity.
On the people side, they need to recruit from diverse profiles, taking into account transferable skills and demonstrated ability to learn. When it comes to process, enterprises need to encourage a diverse community that is consistently nurtured to create a sense of belonging amongst data scientists. Lastly, enterprises need to empower their data scientists with a diverse toolkit to drive collaboration and productivity.
By taking into account people, processes, and technology, companies give data scientists the ability to freely express themselves through their work and give them a sense of belonging. Great data science work is a culmination of different ideas and backgrounds that encourage collaboration to reach strong outcomes.
Were any of these survey findings surprising?
It’s gratifying to see that in this survey, accessing data, data engineering, and data prep are no longer rising to the top as challenges to generating outcomes with data science. Indeed, “better access to data platforms” came dead last in terms of things that would be helpful for data science success. Organizations seem to be finally coming to grips with providing data to the organization, and teams can now turn to the challenges of developing, operationalizing, and managing production models instead.
How do you expect we’ll see these answers change when you conduct the survey next year?
I’d expect to see increasing divergence between companies that have been undertaking the necessary changes to integrate data science into their businesses and those that haven’t. In the event there is a significant economic downturn, this gap will widen even further as, for the lagging organizations, it becomes harder to fund transformational projects despite the fact that this is exactly the time when data science projects are particularly valuable. Data science can help organizations detect, understand, and adapt to changing market conditions in a way that intuition and traditional decision-making processes cannot.
What are the general data science trends enterprises should watch in the next 6 to 12 months?
Organizations are increasingly looking to develop data science as an organization-wide capability. Expect to see initiatives to create consolidated data science communities inside organizations and heavy investment in centralized platforms that provide shared infrastructure, collaboration, and governance capabilities while supporting an increasing range of data science tools. Expect the focus on MLOps to increase and efforts to streamline and automate end-to-end workflows to accelerate. As part of this, expect to see more firms implementing feature stores, model registries, experiment-tracking features, and pipeline tools.
Separately, synthetic data, which is valuable for accelerating training and improving robustness of data science model, appears to be finally taking off. There is now a critical mass of vendors and an increasing understanding by users of the value. A similar trend is happening in tinyML, with an increasing array of solutions helping to develop ML models that can be embedded in edge devices such as consumer electronics.
What’s next for MLOps-driven companies?
As companies expand and integrate their use of machine learning, there is the increasing realization that data scientists need to be able to access data, and develop and deploy models everywhere the enterprise operates, whether that be in multiple regions, on premises, in the cloud, or across multiple cloud platforms. In effect, they need hybrid cloud solutions for their data scientists. Unfortunately, no satisfactory ones exist and few vendors are making credible commitments to supporting this. Companies should implement a hybrid cloud offering to help organizations become data science–driven everywhere, and in the process help them solve issues of data sovereignty, cost, infrastructure availability, and vendor lock-in.
[Editor’s note: Kjell Carlsson is the head of data science strategy and evangelism at Domino Data Lab. Previously, he covered AI, ML, and data science as a principal analyst at Forrester Research where he wrote reports on AI topics ranging from computer vision, MLOps, AutoML, and conversation intelligence to augmented intelligence, next-generation AI technologies, and data science best practices. He has spoken in countless keynotes, panels, and webinars, and has been frequently quoted in the media. Carlsson received his Ph.D. in business economics from the Harvard Business School. You can contact Mr. Carlsson via LinkedIn.]