BNOW has no one dedicated to DevOps. Now it hands production work to an AI agent and fixes problems itself.
About BNOW
BNOW collects biometric data from animals through capsule sensors and gateways in the field. Its platform analyzes the readings and sends an alert when something looks wrong.
- Industry: Animal biometric data—field sensors, gateways, and analytics
- Team: No dedicated DevOps or infrastructure engineer
- Uses: Alpacon, with Claude Code as the AI agent
Highlights
- An hour at most from outage to fix with Alpacon and an AI agent, down from four hours or more at worst
- Under a week from the first working session to every server on Alpacon
- 0 → 2 people in-house who can diagnose a production outage
"For a company like ours, Alpacon is the security team."
—Elon Choo, CEO, BNOW
Challenge
A real-time service with no one on infrastructure full time
BNOW's platform takes in sensor data from the field in real time. When a server goes down, the service goes down, and it's obvious right away.
"Our service runs on real-time sensor data. The moment it stops coming in, it shows."
—Chang Choi, VP of Engineering, BNOW
Someone ran the servers, but not full time. Chang Choi, BNOW's VP of Engineering, has run its servers and databases all along, on top of leading engineering. What BNOW didn't have was a dedicated DevOps engineer.
So when something serious broke, BNOW called in outside help. And things rarely broke at a good time.
"It'd be Friday evening, or the morning of a day off. That kept happening."
—Elon Choo, CEO, BNOW
Alpacon was there. No one knew how to use it. Some of BNOW's servers were already connected to Alpacon. But no one in-house could use it to find out what had gone wrong. Checking whether a server was up was about all they did with it.
"Before that, we really only used it to check whether a server was up or down."
—Chang Choi, VP of Engineering, BNOW
"Nobody here knew how to use Alpacon."
—Elon Choo, CEO, BNOW
Everything got fixed. It just took too long. The data can't stop, so every problem got fixed eventually. But fixes that look simple now could take four hours or more at worst. Filing the request, finding the cause, and waiting for an answer each added time.
AI agents could help, but not on production
When AI agents arrived, BNOW wanted to use them for exactly this: diagnosing and fixing problems. But it couldn't connect one to production or let it do real work.
Give an agent your server credentials and it can do the job. It can also reach anything those credentials reach. If it runs something you can't undo, nothing stops it.
That risk is bigger for a team without a dedicated infrastructure engineer. There's no one standing by to roll it back.
Solution
Why Alpacon
- It was already in place. Some of BNOW's servers were on Alpacon before any of this started.
- Scoped access instead of shared credentials. An agent gets only the permissions a task needs, and Alpacon controls the commands it actually runs.
- The timing fit. BNOW was already looking at AI agents when it saw how Alpacon puts them to work on production. It started after one demo.
Only the permissions a task needs
Alpacon draws the line that was missing. Instead of handing an agent your credentials, you give it a controlled path: the permissions a task needs, and control over what it runs.
That was the first thing BNOW saw in practice. When an agent reached for something it shouldn't, Alpacon raised an alert and stopped it.
"The one thing we could actually see was Alpacon throwing an alert whenever the AI tried to connect to whatever it wanted. At first, I loved that: 'Alpacon's blocking it.'"
—Elon Choo, CEO, BNOW
Every server connected and documented
At first, only some servers were on Alpacon. Once BNOW saw it working, the team connected the rest, Windows servers included.
They also surveyed the infrastructure and wrote it down. In the first session, the agent analyzed the database structure and documented it. The team also wrote a rule into BNOW's own repository: back up the database and get user confirmation first.
With every server connected, one agent could see all of BNOW's infrastructure. Work that had been split server by server came together, and BNOW could start handing over real work.
Up and running in under a week
From the first working session to every server on Alpacon took less than a week. There was no separate rollout and no training program.
Results
| Before | Now | |
|---|---|---|
| Diagnosing a production outage | Couldn't be done in-house; needed outside help | An AI agent diagnoses the servers connected to Alpacon |
| Servers on Alpacon | Some | All of them |
| Outage to fix | Four hours or more, at worst | An hour at most |
The results come from using Alpacon and an AI agent together.
Diagnosing everything from one machine
Finding out why an alert never went out used to mean logging in to each server and checking them one by one. Every server needed its own agent, and each agent needed its own instructions.
Now it all runs from one computer.
"I can say, 'This alert didn't go out on this date. Why not? Check the three or so VMs, run a simulation on what they picked up, and tell me why.' And it does it."
—Chang Choi, VP of Engineering, BNOW
Finding the cause, fixing the algorithm, pushing it to GitHub, and deploying it to the servers is now one workflow. A fix that could take four hours or more now takes an hour at most.
Handling problems after hours
Outages still hit on Friday evenings and weekends. What changed is what BNOW can do about them.
Calls from the field now reach the CEO first, and he handles it himself. No more filing a request and waiting for an answer.
"Now I'm the one who goes into Alpacon and reboots it. It's completely flipped—before, I had no idea how."
—Elon Choo, CEO, BNOW
A security team without a DevOps hire
BNOW needed someone dedicated to infrastructure and never filled the role. It used to rely on outside help to find out why production broke. Now it diagnoses and fixes problems in-house.
"A big company running tons of servers has a server security team, and this is a tool that team uses. For a company like ours, Alpacon is the security team."
—Elon Choo, CEO, BNOW
