AMPLIFY VOL. 37, NO. 5
You shall know a word by the company it keeps.
— John Rupert Firth1
Eighty percent of organizations say their primary technological goal is hyper-automation (the end-to-end automation of as many business processes as possible), which promises significant leaps in business value.2,3 It comes as no surprise, at least to us, that business analytics is considered a promising business function for generative artificial intelligence (GenAI) augmentation. Indeed, movement toward data democratization (empowering everyone within an organization to work with data) has been a long-standing goal for many organizations.4
Data democratization traces its roots to the early 2000s, marked by the advent of drag-and-drop data analytics tools like Tableau and Power BI. These tools simplify data access, enabling all members of an organization to engage with data without much reliance on data experts, with the ultimate goal of delivering higher business value.5,6
So if data democratization aimed at significant progress in business value is nothing new, why do some perceive GenAI as a threat to data-related jobs?7,8 And why does this necessitate a rethink of our conception of business value?
GenAI’s Role in Business Analytics
Done right, business analytics can significantly increase business value.9,10 But keep in mind that “value” has become a fashionable term, and such words are often repeated to the point of oversimplification. Consequently, value is frequently discussed without explicit definitions, the assumption being that everyone understands what it means. But as Firth rightly observed, you shall know a word by the company it keeps.
The metrics associated with value, including productivity and profitability, speak volumes about prevailing priorities among business leaders.11 The dominant focus on financial metrics often eclipses crucial humanistic outcomes (particularly for employees), which are increasingly vital with the advent of GenAI.
Although proponents of AI draw parallels between GenAI and past innovations like ATMs and autonomous delivery systems, those advances unfolded gradually and did not drastically reduce jobs in their respective sectors. GenAI is advancing at an unprecedented pace, with large language models (LLMs) becoming easier to use and being integrated into widely used software products, including Microsoft Office and dominant data analytics tools.
These integrations are set to alter not just workflows but entire work ecosystems, potentially impacting humanistic outcomes, especially for frontline workers like data experts.
Indeed, although data democratization is nothing new, GenAI has taken things up a notch. Drag-and-drop analytics simplify some tasks, but they still require specialized knowledge in statistical analysis and familiarity with the analytics tool.
English is now a programming language and data analytics tool, thanks to GenAI. The ability to converse with an analytics platform by asking plain-English questions about company data implies that all employees can extract insights and ideas from data.12 As mentioned, drag-and-drop tools such as Tableau and Power BI are now imbued with GenAI capabilities. Tableau Pulse enables the automation of data analysis and communicates insights concerning metrics of interest conversationally, aiming to let virtually everyone become data-driven. We are all programmers and data “experts” now.13
In that case, what becomes of data-related jobs? How should these GenAI teammates be added to the workflow? How would that impact workforce ecosystems and, in turn, employee-related outcomes? Although experts disagree about whether GenAI will wind up displacing more jobs than it will create over time,14 some job cuts in data analytics have been linked to changes resulting from AI.15
We tend toward a more ambivalent stance on the issue of displacement. From any reasonable person’s framework, imbuing data analytics tools with GenAI capabilities will undoubtedly enhance efficiency. Unless there is a lot of work to go around, some people may have to go. The issue with current discussions on job displacement is that it is viewed as a binary outcome — either GenAI replaces humans, or it doesn’t. In reality, the extent of displacement is likely to vary across different industries and over time. Additionally, those who argue that GenAI will create new jobs often cite prompt engineering as a prime example. However, as GenAI tools become increasingly sophisticated and user-friendly, prompt engineering jobs may not be sustainable. We therefore advise caution against definitive predictions and recommend focusing on how employees can be better prepared for the future.
In this article, we (two academics in information systems [IS] and analytics and a practitioner with decades of experience in digital transformation) explore how GenAI can be integrated into the analytics value chain. We examine real-world cases to understand how GenAI is changing this value chain in various business functions, the impact on humanistic outcomes, and what is being done to re-skill employees. In this way, we lay a stone on the path toward ethical digital transformation and AI for social good.16
Current Landscape
As mentioned, there are two prevailing perspectives about the threat GenAI poses to data analytics and related jobs. The proof of this divide is evident in industry actions. Some companies have committed to using GenAI as a tool to augment their workforce without replacing human workers; others have demonstrated its potential for job displacement by letting go of employees (though the total number of jobs directly lost to GenAI remains low).17,18
Some argue that we are not yet at the point of displacement, since GenAI tools still lack critical thinking, strategic planning, and complex problem-solving abilities.19 We agree that GenAI is less likely to cause immediate displacement. The layoff we cited earlier, at EXL Service Holdings, impacted only 2% of the workforce, and it was mainly targeted at junior staff in data analytics and digital operations.
Nevertheless, we caution against predictions that displacement will not happen or that it is so far off that there’s no cause for concern. Remember that GenAI was released to the public in late 2022 — these technologies are less than five years old. ChatGPT’s advanced data analysis tool was introduced in June 2023.20 Yes, it still produces inaccurate results here and there, but it can perform complicated calculations, generate charts based on uploaded data, and generate code associated with produced analysis.
In fact, the original idea for this article came after one of the authors gave ChatGPT’s advanced analysis tool a tutorial question on optimization she had used in her business analytics course. Not only did the tool get the correct answer, it worked through it in less than 10 seconds. Again, these tools are still “infants.” We are only beginning to realize the impact LLMs are about to have on data analytics and cognate jobs, including software development.
Many other GenAI-powered analytics tools are available, including ChatGPT Enterprise, Tableau Pulse, and bespoke tools like the AI @ Morgan Stanley Assistant, which gives financial advisors speedy access to a database of about 100,000 research reports and documents.21
Given privacy and security concerns, we expect more organizations to develop proprietary AI-based analytical tools. For instance, Databricks, a company providing a unified platform for data, analytics, and AI, recently agreed to acquire GenAI start-up MosaicML in a deal valued at approximately US $1.3 billion, a strategic move targeting the rapidly expanding demand for businesses to develop or fine-tune their own ChatGPT-like tools.22 As people who deal with data in our research and work, we agree that these tools offer profound leaps in productivity.
However, little is known about how they are integrated into actual organizations and their impact on employees. This may be because organizations are starting to jealously safeguard their most lucrative GenAI use cases.23 We did find some real-world cases that can help us look at how GenAI can be integrated into analytics and the potential impact on data experts.
We scoured recent news articles and practitioner outlets, including the Wall Street Journal, the New York Times, MIT Sloan Review, and Harvard Business Review for informative cases. We identified three from which we draw insights on successfully integrating GenAI into analytics. Note that data experts are not just those with “data” in the title. We consider other professions that deal with data (e.g., portfolio managers, financial advisors, and wealth managers). Data expertise is a matter of skill, not job title.
Case 1: PwC Digital Accelerator Program
The digital accelerator program is not new, but we believe it offers some valuable insights. Launched in 2017, the program began with three components: data science, automation, and data management.24 At the heart of this program is what has been labeled “citizen development.” Citizen developers are recruited from among PwC’s employees and voluntarily take courses in relevant technologies, including data analytics and GenAI. Following certification from these courses, graduates take some time to develop applications relevant to the company.
Employees have complete autonomy regarding what automation products they pursue, and the most promising ones are promoted by the company’s centralized Products and Technology group.25 One digital accelerator created a program that automated a workflow for extracting data from various spreadsheets, saving 40 hours of work on audits; eventually, this program was integrated into the company’s internal library.26
We see two factors contributing to this program’s success. First, it is supported by Tim Ryan, PwC’s US chairman and senior partner.27 Second, employees who develop useful products are rewarded.
With PwC planning to invest $1 billion into GenAI, we foresee that in the next few years, the accelerator program will reach new heights.28 It is also promising to read that Mohamed Kande, PwC’s vice chair, leader of US consulting solutions, and global advisory leader, shared that PwC aims not to replace workers but to optimize their jobs. “We are not going to leave anybody behind. It’s going to be a team sport,” he said.29
Case 2: Morgan Stanley
Morgan Stanley is leading the charge as one of the first major Wall Street firms to develop a proprietary solution anchored on GPT-4. It designed a tool that gives financial advisors quick access to a database of about 100,000 research documents and reports.30 As we know, financial advisors deal with both structured and unstructured data. Structured data could be in the form of stock performance; unstructured data includes business reports, client meeting notes, portfolios, and more. Before implementing AI @ Morgan Stanley Assistant, financial advisors had to scour the unstructured data using manual search. The AI assistant lets them search workplace documents using queries akin to a Google search.
Interestingly, this tool does not tell financial advisors what to do;31 it simply searches the data and spits out what is required, giving the human advisor autonomy to serve the client. Morgan Stanley’s co-president shared in a memo that financial advisors will always be the center of the company and that GenAI will only offer advisors some leaps in efficiency.32 In a continued effort to move AI innovation forward, Morgan Stanley recently announced its first company-wide head of AI, Jeff McMillan, who previously headed the analytics, data, and innovation unit within the wealth management division and was instrumental in the development and implementation of AI @ Morgan Stanley Assistant.33
Case 3: Ørsted
At Danish multinational energy organization Ørsted, the data experts are the executive management team, which needs to stay abreast of industry trends. With north of 300 relevant news articles each day, it was almost impossible for each executive to read them all. Attempts at having individual folks create summaries resulted in a flow of information that wasn’t fully aligned.
Ørsted integrated an LLM from Azure OpenAI into the Dataiku platform, enabling an automated daily news digest. The company’s head of digital strategy and innovation, Heidi Østergaard, shared, “For us, democratization of AI is about increasing AI literacy in the organization, giving people the tools and platforms they need.”34
Recommendations
Case studies drawn from news articles may not fully represent the empirical reality of how GenAI is integrated into these firms. However, they help us think about what seems to work and how it might happen. Drawing from these cases, we propose the following five recommendations (summarized in Figure 1).
1. Align GenAI Tools with Data Expert Needs
Senior management has a tendency to pursue digital transformation goals without much input or feedback from members of the organization who use and are impacted by GenAI integrations. Combining top-down and bottom-up approaches is helpful to ensure that GenAI tools align with data experts’ actual workflow and needs. For instance, at Ørsted, GenAI analytics were implemented to solve a data expert problem. The literature on IS has consistently shown that individuals are inclined to continue using a technology they find beneficial. Thus, tailoring these tools to meet user needs will enhance their perceived usefulness, which is needed because the success of digital transformation depends on the active use of these technologies.
2. GenAI Tools Must Lend a Hand, Not Control
There is a delicate balance between helping data experts reap the benefits of GenAI integration and ensuring that GenAI does not assume supervisory control over human experts. In the three cases we discussed, GenAI’s main task was data analytics, leaving the human experts to make the final call. Indeed, GenAI should not dictate employees’ actions. This is not only because GenAI is not yet at satisfactory levels of emotional intelligence, but also because giving GenAI some sort of supervisory role would exacerbate fears of displacement. Additionally, it complicates compensation and reward mechanisms. Who takes the credit in the event of a big win for the company, and who takes the blame in the event of a mistake? Let data experts be experts.
There are cases, however, where GenAI may have to step in. For example, Voya employs GenAI to assist humans in selecting stocks because, unlike humans who may become emotionally attached to a stock, GenAI remains impartial.35 However, it is critical to clarify who holds the veto power in case of a disagreement between GenAI and the data expert. One way is to set up a “devil’s advocate” team comprising humans and GenAI. The capabilities of GenAI are so vast and profound that it is feasible to create a GenAI that functions as a devil’s advocate, critiquing suggestions from GenAI stock advisors.
3. Re-Skill & Upskill Data Experts
As organizations move toward democratizing data analytics, it is crucial to ensure that data experts maintain their expertise in data analytics, mitigating any risk of redundancy or loss of specialized skills. Data experts can also re-skill and upskill. We have seen that layoffs in data analytics are affecting employees at the junior level. Data from LinkedIn shows that jobs requiring AI skills are on the rise.36 As non-data experts are introduced to surface-level analytics (no code/low code), there is an opportunity for junior data analytics employees to enhance their capabilities by advancing to more sophisticated analytics and AI skills.
4. Practice Inclusive GenAI Integration
Top management must dispel any notion that data experts using GenAI for domain-specific knowledge are at risk of being replaced by the technology. Leadership at all three organizations we studied publicly acknowledged that these tools are intended to support, not threaten, data experts. Such assurances from leadership are vital as they foster a sense of psychological safety, which is essential for effective collaboration between GenAI and data experts.
Let’s put ourselves in the employees’ shoes. Imagine having a colleague who can crunch numbers and produce thousands of lines of code in seconds. Wouldn’t you feel threatened by them? We should not rob employees of their “somebodiness” by even implying that they are easily replaceable. One way to circumvent such thoughts is to take an employee-inclusive GenAI adoption approach. For instance, PwC makes the tools available to company citizens and allows them the autonomy to automate, ipso facto including them in GenAI initiatives. Indeed, as mentioned, products deemed promising from these employee projects are adopted firmwide, and those employees are compensated accordingly.
5. Senior Leadership Must Be Knowledgeable in AI & Analytics
We’ve heard many times that 90% of digital transformation projects fail. One of the critical antecedents to successful digital transformation is a realistic appreciation of what technology can and cannot do. To fully benefit from integrating GenAI into the analytics value chain, leadership must understand — at least to a basic extent — the capabilities and limitations of these LLMs.
This understanding helps prevent digital transformation strategies from becoming mere wishful thinking. For example, Morgan Stanley appointed the previous wealth management chief analytics and data officer as the companywide head of AI who reports directly to the co-presidents. This decision flattens the organizational hierarchy by positioning the head of AI at a senior level rather than under the CTO or CFO, signaling a long-term commitment to AI investment. Additionally, this positioning ensures that the firm’s AI initiatives are aligned with broader strategic goals.
Conclusion
The incorporation of GenAI in business analytics offers the potential for significant advancements in business value. Our article contends that prevailing views of business value tend to emphasize financial outcomes. With the apprehensions surrounding potential GenAI-induced displacement in white-collar professions (justified or not), it is crucial to thoughtfully integrate GenAI into data analytics processes to facilitate successful digital transformation.
References
1 Firth, J.R. Papers in Linguistics 1934–1951. Oxford University Press, 1964.
2 De Cremer, David. “For Success with AI, Bring Everyone on Board.” Harvard Business Review, May–June 2024.
3 Piskorski, Misiek, and Amit Joshi. “What Roles Could Generative AI Play on Your Team?” Harvard Business Review, 22 June 2023.
4 Lefebvre, Hippolyte, Christine Legner, and Elizabeth A. Teracino. “5 Pillars for Democratizing Data at Your Organization.” Harvard Business Review, 24 November 2023.
5 Petrova, Bilyana. “The Importance of Drag and Drop Analytics.” Reveal, 28 June 2023.
6 Ruzgas, Tomas, and Jurgita Bagdonavičienė. “Business Intelligence for Big Data Analytics.” International Journal of Computer Applications Technology and Research, Vol. 6, No. 1, January 2017.
7 Lin, Belle. “EXL Service Cutting 800 Jobs as It Shifts Focus to AI.” The Wall Street Journal, 4 April 2024.
8 Lin, Belle. “IT Unemployment Soars to 4.3% Amid Overall Jobs Growth.” The Wall Street Journal, 6 October 2023.
9 Cazier, Joseph A. Leading in Analytics: The Seven Critical Tasks for Executives to Master in the Age of Big Data. Wiley, 2023.
10 Koch, Hope, Wallace Chipidza, and Timothy R. Kayworth. “Realizing Value from Shadow Analytics: A Case Study.” The Journal of Strategic Information Systems, Vol. 30, No. 2, June 2021.
11 Koch et al. (see 10).
12 Lin, Belle. “Businesses Seek Out ChatGPT-Tech for Searching and Analyzing Their Own Data.” The Wall Street Journal, 12 May 2023.
13 Davenport, Thomas H., Ian Barkin, and Kerem Tomak. “We’re All Programmers Now.” Harvard Business Review, September–October 2023.
14 Delaney, Kevin. “Bringing AI Tools to the Workplace Requires a Delicate Balance.” The New York Times, 8 May 2023.
15 Lin (see 7).
16 Dennehy, Denis. “Ireland Post-Pandemic: Utilizing AI to Kick-Start Economic Recovery.” Amplify, Vol. 33, No. 11, 2020.
17 Lin (see 7).
18 Smith, Ray A. “AI Is Starting to Threaten White-Collar Jobs. Few Industries Are Immune.” The Wall Street Journal, 12 February 2024.
19 Marr, Bernard. “Will ChatGPT Put Data Analysts Out of Work?” Forbes, 7 February 2023.
20 Lu, Yiwen. “What to Know About ChatGPT’s New Code Interpreter Feature.” The New York Times, 11 July 2023.
21 Son, Hugh. “Morgan Stanley Kicks Off Generative AI Era on Wall Street with Assistant for Financial Advisors.” CNBC, 18 September 2023.
22 Loten, Angus, and Belle Lin. “Databricks Strikes $1.3 Billion Deal for Generative AI Startup MosaicML.” The Wall Street Journal, 26 June 2023.
23 Brans, Pat. “Will Enterprises Soon Keep Their Best Gen AI Use Cases Under Wraps?” CIO, 17 April 2024.
24 Davenport et al. (see 13).
25 Barkin, Ian, and Thomas H. Davenport. “Harnessing Grassroots Automation.” MIT Sloan Management Review, 11 September 2023.
26 Davenport et al. (see 13).
27 Barkin and Davenport (see 25).
28 Loten, Angus. “PricewaterhouseCoopers to Pour $1 Billion Into Generative AI.” The Wall Street Journal, 26 April 2023.
29 Loten (see 28).
30 Son (see 21).
31 Welsch, Andrew. “How AI Will Change Wealth Management.” Barron’s, 5 July 2023.
32 Son (see 21).
33 Corbin, Kenneth. “Morgan Stanley Names Jeff McMillan First Firmwide Head of AI.” Financial News, 18 March 2024.
34 “Ørsted: Monitoring Market Dynamics With LLM-Driven News Digest.” Dataiku, accessed May 2024.
35 Welsch, Andrew. “How Voya Is Blending AI with Humans to Pick Stocks.” Barron’s, 31 July 2023.
36 Laker, Benjamin. “LinkedIn Data Predicts 65% Shift in Job Skills by 2030 Due to AI.” Forbes, 3 October 2023.
