Innersourcing an Enterprise Slack Bot to Automate Team Support; Public Cloud vs Cloud Repatriation: How the Financial Agendas of Early Proponents Muddy the Debate

“How can smart people look at the same set of facts yet arrive at different conclusions?”

I ran into that exact question while researching my initial (and later scrapped) technical topic around cloud computing and big data. I was enrolled in a cloud computing course whose lectures framed the public cloud as next-generation technology to replace private infrastructure, so I was shocked to discover a growing movement around cloud repatriation, the process where a company moves their software out of the public cloud and back onto private infrastructure. Even more surprising was how the proponents of this movement pitched its merits near identically to the case for public cloud: reduced cost, increased reliability, and improved security. That contradiction drove me to dig deeper: for my STS topic, I investigated the agendas underlying the cloud repatriation. Outside circumstances forced me to change my technical topic; instead, I wrote about implementing a frequently asked questions (FAQ) Slack chatbot run on natural language processing for my internship with Capital One. Despite the disconnect between these two topics, I found a way in both papers to analyze a sociotechnical system to better understand how engineering practice is defined by the circumstances around a solution.

The technical portion of my thesis produced a paper describing how my intern team designed the FAQ chatbot for use by internal clients of Capital One’s document database Gallery. The Gallery team was overwhelmed with support requests from newer associates, finding their time wasted by meticulous tasks such as answering simple questions or performing simple file lookups. My team alleviated their workload by extending an existing enterprise solution for answering frequently asked questions in Slack channels by compiling the dataset of common concerns and their solutions. Additionally, we provided a new Slack-accessible method to check the status of a document upload, removing a monotonous request that Gallery associates took time away from their important work to fulfill every day. In the development process, I learned the difference between the development-level work done in UVA classes and the production-level work required in business. Every aspect of the functionality and every line of code required thorough testing and end-user validation to ensure the final product was best suited to the existing system. Specifically, we worked to ensure our Slack bot instance fit seamlessly into the existing support channel already in use by Gallery users. Instead of a neat gimmick, we solved a real problem for the team and created business value for the company by saving their time.

In my STS research, I uncovered the marketing strategies that drove the growth of cloud repatriation as an alternative to the public cloud. Rather than answer the specific technical question of which approach is superior, I evaluated the sources that attempted to answer that question. My research revealed two categories of sources: those that offer a specific alternative sold by the article publisher, and those that offer fair analysis with a flexible recommendation. Many of the early proponents of cloud repatriation fell into the sales-pitch category, constraining the discourse to discrete either-or solutions. On the other hand, the more recent analytical sources provide a balanced perspective that focuses on solving the original computing problem, offering creative solutions that best match a company’s needs. I used the sociotechnical framework of the discourse of inevitability to make the distinction, looking for language around trends and progress to suggest marketing, versus innovations and originality to suggest design-focused discourse. The sources that used more design-focused language typically did not sell a product, but rather contributed to the possible set of solutions to solve mass computing.

In writing both documents simultaneously, I appreciated how the learnings from my technical paper influenced the writing of my STS report. The focus on end-users and fitting a solution into an existing sociotechnical system like a support channel illuminated the value of the creative cloud computing solutions. For instance, repatriating an entire cloud workload may fit a company’s culture better than using employee bonuses to incentivize cost cutting, even if the arguments in favor comes from a marketing standpoint. I better understand how a perspective, even if coming from an agenda, can still be valuable in the right context. From the whole experience, I am better prepared for my future work in software engineering because I can leverage the sociotechnical systems I interact with to craft the best solution for my end-users.

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Rosanne Vrugtman

STS Advisor: Kathryn Neeley