Key Takeaways
- Darren Taillie eliminated a 40-hour weekly queue of manual data requests at Aerocore Technologies by connecting an LLM to database MCP servers inside Microsoft Teams.
- Operational managers now pull aircraft tail numbers and historical wash records in 3 seconds using natural language, down from turnaround times that previously took between an hour and a full week.
- Bypassing custom dashboard portals and meeting non-technical staff directly in Microsoft Teams drove immediate internal adoption across operations.
- Replacing ad-hoc Jupyter Notebook scripts and messy
pandas.read_csvworkflows with direct database agents removed what Taillie terms "work-around sprawl."
The Method
Aerocore Technologies cleans jet engines in the field. As Taillie put it: “We are an incredibly analog company in a digital world. There are no digital components on the wash trucks whatsoever. So, my job is to bring some of the digital world to our company.” (play from 31:49)
Because Taillie was neither a dedicated database engineer nor a formal developer, ad-hoc business intelligence requests created an operational clog. Taillie found himself writing repetitive Python scripts just to answer routine business questions. He described this cycle directly: “And then that also leads to another lovely coined term, which is work-around sprawl, which essentially is if I, cuz again, as you didn't hear in my title, not developer, not SQL person, so all of this turns into pandas.read_csv with groupbys and location filters, right?” (play from 35:11)
That routine consumed an entire work week. “The request comes to me, I have to go in, pull the data, merge the data, analyze the data, send the results,” (play from 34:50) Taillie explained. “And we do that for 40 hours a week and we've done no long-term growth, we've done no extra tasks, we've done no research.” (play from 34:57)
To break the loop, Taillie connected an LLM agent directly to the database via Model Context Protocol (MCP) servers, surfacing the bot inside Microsoft Teams. Instead of waiting for an analyst to parse CSVs, field managers type questions about specific tail numbers and past engine wash cycles into a standard chat channel.
Adoption followed distribution. Field operators ignore third-party analytics dashboards, but they check chat applications all day. “Once that got moved into Teams, very high adoption, everybody's using it all the way through, and it's getting, it's getting a lot, a lot of hits and seeing that in the analytics,” (play from 37:33) Taillie noted. Operational questions that once stalled in a ticketing queue now return answers in seconds: “Things that used to take 30 minutes, an hour, a week for complex stuff, the system now does in a couple of minutes if you gave it a really complicated one. You know, this came back in about 3 seconds, so it does a really, really, really good job of getting the decision-makers the information they need.” (play from 37:53)
Where This Breaks Down
Giving an autonomous agent query access over operational data solves backlog bottlenecks, but it carries strict limits. If your underlying schema contains ambiguous column headers, messy tail-number formats, or unindexed tables, natural language translation will generate flawed joins. A bot that queries an uncleaned database can easily misinterpret null wash logs as incomplete jobs.
Text-to-SQL agents also require read-only permissions and strict row caps. If an operational lead asks an open-ended question that triggers an unconstrained join across millions of historical service rows, an unmetered agent can degrade production database performance. The interface works only when the database schema is clean and permissions are fenced.
What to Do With This
Audit your last two weeks of internal Slack or Teams requests. Find the three repetitive data queries your operators ask you to pull via CSV or admin panels. Connect a read-only database MCP server to an internal chat bot targeting only those specific tables, then deploy it directly to the channel where your team already works.