Suggestions
Henry Hund
COO @ Aptible | Finally, get the developer platform that gives you the control, reliability, and enterprise-grade infrastructure you need
I love to help early stage companies execute faster and more efficiently.
I develop strategies to help companies navigate challenges of product-market fit and beyond. I help companies define their goals and then do whatever I can to help them exceed them.
I specialize at being the generalist that propels companies forward. In a typical day, I will:
- Analyze some user cohort data (working in our data warehouse or in our BI tool)
- Update our marketing/demand gen campaigns
- Design new data enrichment, lead assignment, and scoring strategies with our RevOps team
- Coach our AEs on testing new sales or outbound strategies
- Collect pricing feedback from customers
- Help our designers and engineers refocus our roadmap
- Coordinate on back office: dunning, monthly close process, and regular reporting
- Explain performance (burn, runway, growth) and our strategy and OKRs to the board of directors
In the past seven years, I've held the following job titles (in reverse chronological order):
- Chief Operating Officer
- Interim Head of Product
- Interim Head of Marketing
- Interim Head of Sales
- Interim Head of CS
- VP Operations
- VP Growth
- Director Growth
- Marketing Operations Lead
Highlights
Engineers are starting to adopt AI for incident response. But we’re still seeing three main concerns:
𝗖𝗼𝗻𝗰𝗲𝗿𝗻 #1: “𝗜 𝗱𝗼𝗻’𝘁 𝘁𝗿𝘂𝘀𝘁 𝗮𝗻 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝘁𝗼 𝗿𝗲𝗽𝗹𝗮𝗰𝗲 𝗺𝘆 𝗵𝘂𝗺𝗮𝗻 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀” 🤨
Of course you don’t, and you shouldn’t. We’re on the same page. Instead, we think AI should be used to help humans communicate better and quickly access the information they need. This is especially important during times of high stress like incident response.
Aptible’s AI Agent is an assistant, not a replacement or teammate.
𝗖𝗼𝗻𝗰𝗲𝗿𝗻 #2: “𝗜’𝗺 𝘄𝗼𝗿𝗿𝗶𝗲𝗱 𝗔𝗜 𝘄𝗶𝗹𝗹 𝘁𝗮𝗸𝗲 𝗱𝗼𝘄𝗻 𝗺𝘆 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲” 💣
Aptible is designing our Agent to ensure that (at least for now) it cannot do anything other than act as an advisor, assistant, or copilot. It can make suggestions and fetch data about your infrastructure, but it doesn’t have the ability to change anything in your infrastructure.
𝗖𝗼𝗻𝗰𝗲𝗿𝗻 #3: “𝗜 𝗰𝗮𝗻 𝗵𝗮𝘃𝗲 𝗯𝗲𝘁𝘁𝗲𝗿 𝗶𝗻𝗰𝗶𝗱𝗲𝗻𝘁 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲 𝗮𝗻𝗱 𝗹𝗲𝘀𝘀 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝘀𝘁𝗿𝗲𝘀𝘀 𝗯𝘆 𝗶𝗺𝗽𝗿𝗼𝘃𝗶𝗻𝗴 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝗰𝘂𝗹𝘁𝘂𝗿𝗲 𝗿𝗮𝘁𝗵𝗲𝗿 𝘁𝗵𝗮𝗻 𝗮𝗱𝗼𝗽𝘁𝗶𝗻𝗴 𝗮 𝗻𝗲𝘄 𝗔𝗜 𝘁𝗼𝗼𝗹” 🤝
In a high growth startup like ours (and maybe like yours), it’s impossible for every engineer to have all the context they need to respond to every incident.
It’s impossible to train every engineer to calmly and rationally search the docs and fetch the relevant log lines or traces manually during an incident. And it takes a long time to onboard new engineers for on-call.
Aptible’s AI Agent is really good (and fast) at handling research during times of high stress. Your engineers (even the newest ones) can ask it to quickly comb through docs and logs to diagnose and summarize the problem.
__
We don’t think AI should replace humans or make any decisions that humans make.
We’re designing our AI to be your assistant, helping you handle incident response better and faster so that you can shave minutes of Time to Restore (TTR). Then you can stop relying entirely on the on-call engineers to handle every issue at every hour of the day.
If you’re interested in testing out the AI agent, check out the link in the comments for early access. You can also message me directly or comment below.
We haven't interrupted our lead engineer for an incident in 4 weeks. Here's how she replaced herself with a chatbot. 🤖
We have 10s of thousands of prod containers that we manage across hundreds of customers. Obviously, we take incidents very seriously.
Any time an unusual incident pops up, our VP Engineering Ashley takes over. She feels like a bottleneck. Her deep knowledge of our infra seems impossible to document and transfer. Her involvement always means faster resolution and less impact on customers.
So in this case, what is good for customers is bad for Ashley. It looks roughly like 6 years of:
- frequent context switching
- lugging her laptop around to events and nights out
- interrupted vacations
- waking up many times per night to investigate incidents
But finally...FINALLY... it feels like we've fixed this.
The solution is an incident response chatbot. An AI Agent. Here's how it works:
- The Agent investigates all incidents reported by PagerDuty, and remediates wherever possible
- It creates and updates runbooks as needed
- It helps all devs confidently resolve incidents, without bothering senior engineers
To do this, the Agent is trained on:
- Incident response and debugging best practices
- The AWS, Ruby, Python, and various database and library docs
- Our existing runbooks and internal docs
And it knows how to gather live information it needs (in milliseconds!) from sources like:
- Our AWS infrastructure
- Slack (for previous discussions on incidents)
- Sentry
- Datadog
- Papertrail
- Our internal APIs (that we've given it read-only access to...)
Needless to say, this has been transformational for our team... and especially Ashley.
Now, the team doesn't have to bother Ashley as often. Whether for incidents or bugs or keeping runbooks up to date. The AI Agent handles much of what Ashley used to do.
Want proof? Ashley is planning to disconnect for her first vacation in forever. 🌴
We’re considering releasing the agent as a product. We've already started a pilot with about a dozen other companies.
P.S. I'm pretty sure this can change a few more engineers' lives, so... want to help us test it out? If so, DM or comment.
Also: In the comments, I'll drop a link to a landing page where you can request early access and stay up to date with everything we're testing out with our AI Agent.
