intwin tech
Automation, plainly described

AI does the repetitive half. An engineer owns the rest.

Plenty of providers now put AI on the homepage. Fewer will tell you which tasks it touches, where it is not allowed, and who is accountable when it is wrong. Here is our version, in full.

Why it matters to you

Automation is a cost decision, not a feature

A managed service is priced on how many engineer-hours your account consumes. Most of those hours go to work that is repetitive and low-judgement: the same five requests, the same noisy alerts, the same notes nobody wrote down. Take that work out and two things change: answers come back faster, and the rate per user stops climbing every time you hire.

That is the whole argument. We are not claiming a smarter product; we are claiming a cheaper way to run the same promise, and we would rather show you the mechanics than the word.

What runs automatically

Four areas. Each one is a place where speed matters more than judgement.

Alert triage

Monitoring & noise

A network of a hundred devices produces thousands of alerts a week, and the overwhelming majority are duplicates, flapping sensors or known-benign events. Automation correlates, de-duplicates and ranks alerts before an engineer sees anything, so attention goes to the handful that represent a real fault. We keep the raw stream rather than discarding it, so we can always show you what was suppressed and why.

Routine service requests

Help desk

Password resets, account lockouts, licence assignment, distribution-list changes, new-starter provisioning from an approved template, software installs from an approved catalogue. These follow a fixed, auditable path and complete without waiting in a queue. Anything that deviates from the template stops and goes to a person.

Documentation

IT Glue

The reason MSP documentation is always out of date is that writing it is a separate chore from doing the work. We generate the record from the work itself (the change made, by whom, on what, and why), and an engineer approves it. When your account manager changes, or when you leave us, the documentation is current rather than archaeological.

Reporting and review prep

Quarterly business review

Automation assembles ticket patterns, recurring faults, licence drift, ageing hardware and security posture into your quarterly review, so the meeting is spent on decisions instead of on someone assembling slides the night before.

What we deliberately do not automate

This list matters more than the one above. These are the places where being fast and wrong is far more expensive than being slow and right.

Security incident response

Detection is automated; the response is not. Isolating a machine, disabling an account or declaring an incident is a human decision with a name attached to it.

Anything that deletes or restores data

Backups run on a schedule automatically. An engineer performs every restore and verifies it against what you expected to get back.

Permissions and privileged access

Granting admin rights, changing security groups or opening firewall rules goes through a person and is recorded, whoever asked and however urgently.

Advice about your business

Roadmaps, budgets, vendor choices and the answer to “should we replace this now or next year” come from an engineer who knows your environment.

Guardrails

How we keep it accountable

Automation without an audit trail is just a faster way to lose track of your environment. Four rules, and we will hold ourselves to them in writing.

  • Every automated action is logged: what ran, when, on what, and what changed
  • If an action cannot be undone, a person does it rather than the automation
  • A named engineer owns your account, so escalation reaches them and not a rota
  • We do not feed your environment into shared models, so your data is never training data
Ask for this in writing
Honest limits

What we are not claiming

We are a Delaware C Corporation incorporated in March 2023, running managed services with a small senior team. We are not going to show you a chart claiming AI cut your costs by a percentage we made up, and we are not going to pretend an automation layer replaces engineers, because ours needs them to work. What we will do is tell you exactly which of your recurring tickets can leave the queue, and show you the log afterwards. Nicolas Bobroff, CEO

Questions we get asked

Does our data get used to train an AI model?
No. Your environment, tickets and documentation stay inside your tenant and our tooling, and are not contributed to shared or public model training. If a tool in our stack changes its terms in a way that affects this, we will tell you before it takes effect, not after.
What happens when the automation gets it wrong?
The same thing that happens when an engineer gets it wrong: it is logged, it is reversed, and a person calls you. Because every automated action is recorded and reversible by design, the blast radius is a single change rather than a mystery. We review misfires monthly and retire rules that keep producing them.
Will we ever be forced to talk to a bot?
Never on the phone. Call the number during business hours and a person answers. That is the point of the arrangement. Some routine requests do resolve automatically through the portal or email because that is faster for you, but there is always a path to a human, and it is one step away.
Do you charge extra for the automation?
No. It is how we deliver the service, not a tier. The benefit shows up as response times and as a rate that does not need to climb every time you add headcount.
Can we see what has run on our environment?
Yes. Ask at any time and we will pull the log for any window you name. It is also summarised in the quarterly review by default.

Find out which of your tickets can leave the queue

Bring us a month of your current tickets, from a helpdesk tool, a shared mailbox, or just memory. We will tell you which categories are automatable and which are not, and what that does to the monthly number.