At VantagePoint we look after networks for businesses that don’t have a big IT team. After twenty-odd years of it, the pattern is hard to miss. The genuinely hard problems are rare. What fills the week is the routine: patch nights, locked accounts, full disks, a certificate that expired on Saturday, the Wi-Fi that ran out of addresses, the alert nobody read until Monday.
All of it is work we already do by hand, for real clients, and know how to do well. The question was whether an AI could do it the same way: carefully, with evidence, and only within limits someone has signed off.
What we decided it has to be
- On site. The AI runs on a box in your building. Your configs, logs and credentials stay there.
- Under a manager. Your IT manager decides, job by job, what it may do alone and what it must ask about. No setting lets it raise its own permissions.
- Accountable. Every action in a tamper-evident log and a plain-English diary.
- Proven first. It earns each job by passing an exam of real faults, and we publish the results.
We’re not trying to replace your engineers. We’re trying to give them their week back.
Guardian is in development, and this blog is where we’ll share how it’s going: the wins, the setbacks and what we learn. If you’d like to shape it, become a design partner.