 Organizations adopting generative AI solutions like ChatGPT adds an interesting twist to the ongoing debate among data and analytics (D&A) leaders about how best to organize and allocate data science, business analyst, and technical talent to scale advanced analytics in the enterprise. Do you centralize these resources and capabilities or distribute (Federate) them out to the business units?
The underlying assumption that AI starts to challenge is that capabilities across these groups are static, and it’s a matter of properly allocating them, like players on a football team, based on experience, size, skills, and ability. But what if your left tackle could suddenly run a 4.4-second 40-yard dash? Or your running back could throw a 60-yard missile into tight coverage? Does that change what you do, how you position the players, and the plays you run? Building on what exists now, generative AI solutions will significantly impact your ability to enhance customer interactions, with more sophisticated chatbots changing the online experience. But what about the folks building and supporting these tools? What might be the impact on D&A capabilities internally if you imagine significant productivity improvements and capabilities brought on by new AI solutions like Microsoft’s “Co-pilot” enhancements for Office 365 and GitHub’s existing Co-pilot product? The ability to query massive stores of unstructured data to extract insights or build applications with just natural language prompts changes what’s possible for individual output and should change our expectations about what people can accomplish and, potentially, their roles within the company. One possible outcome is that the D&A organizational function morphs from a builder of analytics tools to an accelerator of AI-assisted knowledge work. In this scenario, knowledge worker productivity goes beyond software developers (GitHub Co-pilot) to include non-technical business analysts and decision-makers, where there will be greater capabilities in the hands of individuals across the organization (see Microsoft 365 Copilot). As more knowledge is accessible to the individual in the future, less coordination will be required between employees to get answers (the elusive self-service model). This will result in individuals interfacing more with the “big brain” than each other by training, maintaining, monitoring, and contributing to the large language models that will ultimately include the entire corpus of unstructured data inside an organization. Given that they are on the front lines for new technologies, the D&A teams and their internal customers will likely be the early adopters of these new AI capabilities. So, what are some tasks that could be automated or augmented by generative AI solutions now (or soon to be) commercially available? To answer this, let’s break down our analysis into two types of knowledge work: 1. organizational work and 2. project-based work. Organizational work involves strategy and operational decision-making, setting direction and priorities, and allocating people and capital. Project-based knowledge work involves building things like analytics solutions that enable better, faster decisions across the organization. There is an obvious symbiotic relationship between the two groups as the analytics solutions built by the project teams enable better organizational decisions that lead to better allocation of resources and better prioritization. This is similar to what Jim Collins calls “Turning the Flywheel.” Organizational Knowledge Work Wharton professor, Ethan Mollick, has done extensive research into the capabilities of ChatGPT and, subsequently, BingGPT and came away with some compelling use cases. He has also mastered the art of prompt writing to produce extraordinary results. Initially a skeptic of AI, his opinion changed with the release of ChatGPT in November 2022. After a deep dive into the capabilities of ChatGPT and the eye-opening rate of adoption (the fastest technology to reach 100 million users), he noted that: “The current situation is truly unprecedented. We are seeing widespread adoption of a technology that has the potential to significantly boost individual productivity, but which is not yet being fully utilized by organizations. Recent research is starting to show us just how big a deal this is.” Below are excerpts from his blog on this topic covering tasks that knowledge workers engage in, emphasizing writing and coming up with new ideas. So, why is writing important in a business context and how will it produce better thinking and decision-making? Because writing forces you to think. As Wired founder Kevin Kelly notes, “I write primarily to find out what I’ve been thinking, and I don’t know until I write it.” Amazon’s now famous “six-pager” memo requirement for meetings uses narrative description in place of PowerPoint to produce clearer thinking and better decision-making. Bezos explained the move to a narrative form of communication and away from presentations by noting that “…writing a good 4-page memo is harder than ‘writing’ a 20-page PowerPoint is because the narrative structure of a good memo forces better thought and a better understanding of what’s more important than what, and how things are related."(1) According to Mollick, generative AI will help knowledge workers with the following:
Project-based Knowledge Work Let’s move on to project-based work and look at the latest AI-powered enhancements to one of the most popular technical project management tools and the leading code repository. Jira and Confluence, products from Atlassian, are two ubiquitous solutions for technical project management. Jira is an agile project management tool teams use to plan, track, release, and support software. Confluence is a team collaboration and knowledge management tool to help drive productivity in the software dev process. These tools facilitate constant feedback on the software development process and allow teams to adjust based on this feedback, such as slower-than-anticipated progress on a particular feature or the emergence of an issue related to data access. Atlassian just announced the release of Atlassian Intelligence, which will use Open AI’s GPT-4 to create a “virtual teammate” capability embedded in their products. In Confluence, the company claims that “workers will be able to click on terms they don’t recognize in documents and find automatically generated explanations and links to relevant documents. They will also be able to type in questions and receive automated answers based on information stored in documents.” The latter capability would minimize the back-and-forth between team members. For Jira, the new AI features will allow users to take natural language queries and turn them into JQL and SQL. Github’s Copilot product (using OpenAI’s Codex model) helps developers increase productivity by suggesting code changes or generating code from a user prompt. Curious about actual productivity gains from Copilot, the company surveyed 2,000 software developers and uncovered some interesting findings. They found that 88% of developers felt more productive using Copilot, but perhaps more interesting is the drill down into what contributed to this perception. Ninety-six percent (96%) noted that they were “Faster with repetitive tasks,” and 87% reported they exerted “Less mental energy on repetitive tasks.” To assess whether perception was reality, Github conducted an experiment with 90 developers, separating them into two groups and assigning the same task. One group used Copilot to complete the task (writing an HTTP server in JavaScript) the other did not. The Copilot group completed the task roughly half the time (55% less) than the other group. An interesting takeaway is not only the improvements in speed (productivity) but what contributed to these gains: the reduction in grunt work and the repetitive tasks that drain cognitive energy that could be applied to higher-level problem-solving. Similarly, eliminating the constant interruptions (often the norm for organizational knowledge work) by querying your “copilot” rather than your colleague might result in similar productivity gains seen with software developers. But the real gains won’t be just producing more output but allowing knowledge workers to stay focused on difficult tasks, getting into a “flow” state that will increase the quality and creativity of the output. From the GitHub research: “Developers reported that GitHub Copilot helped them stay in the flow (73%) and preserve mental effort during repetitive tasks (87%). That’s developer happiness right there, since we know from previous research that context switches and interruptions can ruin a developer’s day, and that certain types of work are draining.” Conclusion As generative AI solutions roll out in organizations, barriers to sharing information will recede as unstructured data is available org-wide and accessible through these large language models (LLMs). As a result, we will move from the current state of desktop search to find documents to desktop queries to find answers, and ultimately to desktop tasking – assigning tasks your co-pilot executes. The result will be that technical and business staff will see significant gains in capabilities. And by loading these models with organization-specific data and continuously updating them with new learnings, the continuous interaction will help train and improve models. The staff gets smarter, and the Big Brain gets bigger. To realize the full benefits of generative AI, organizational change will be required in two phases. The first phase involves helping automate and augment knowledge worker tasks to free up their time for higher-level problem-solving. The second phase maximizes knowledge workers’ newfound time to think and create. This will require mindset and skill set changes (see Quality Thinking). The challenge for D&A leadership is prioritizing and aligning resources in ways that accelerate this knowledge worker revolution brought on by AI to set the learning, continuous improvement flywheel (smarter people < > smarter models) in motion. ------------------------ (1) Chapter 4 of Working Backwards: Insights, Stories, and Secrets from Inside Amazon. Comments are closed.
|
Archives
June 2023
AuthorRick Hinton is the founder of Valerius, a consulting firm that helps organizations prepare for the age of analytics using adaptive change management practices. |
RSS Feed