Assist-X
xop.ai, AI built for MSPs by an MSP: lower your cost to serve, see every account, and stop giving away services revenue.
Your engineer is on the phone.The ticket is already written.
The transcription agent rides along on the call and produces the notes, the subject, the type and subtype, the priority and an accurate time entry. Your engineer spends the call on the client instead of on the keyboard.
The number that matters
45–60
minutes back, per engineer, every day.
Under a fixed-fee agreement that time is not a service metric. It is your cost line, and it is the only thing your client can actually feel.
An engineer’s hour is rarely spent fixing. It is spent finding, asking, rediscovering and writing up. Assist-X takes those four apart:
- Searching
- Three systems, two of which the answer was never written into.
- Asking a colleague
- Which costs two engineers instead of one, and teaches nobody.
- Rediscovering
- Solving a problem the team already solved, because the fix was never captured.
- Reconstructing
- Notes and time entries rebuilt from memory at five o'clock, thinner every hour that passes.
Every vendor claims two hours a day and buyers discount it on sight. Those four lines are where ours comes from, and you can check each one against your own desk.
The transcription agent
Ask it. Speak it. The ticket writes itself.
An engineer starts the recording and then forgets about it. The call is transcribed as it happens, and when it ends the ticket is written: notes, subject, type, subtype, priority and the time entry. Then the engineer reads it, changes what they want, and posts it.
What comes back when the call ends
- The ticket notes
- Summary, the issue, what was discussed, the resolution and the next steps. Comprehensive in a way notes written at five o'clock never are.
- The subject and classification
- A suggested subject, ticket type, subtype and priority, drawn from what was actually said rather than from a dropdown habit.
- The time entry
- Captured against the work as it happened, not reconstructed at the end of the day.
- The client notification
- Drafted and ready to send, in the tone the engineer picks.
- The recording itself
- A copy of the call attached to the ticket in the PSA, so the record is the call and not somebody's memory of it.
Nothing posts until an engineer says so
The agent does not write to your PSA directly. It hands the engineer a draft to edit and make their own, and the ticket updates when they approve it. This is the first question every service manager asks, and the answer is the reason engineers trust it.
Not a phone system integration
It takes the microphone and the speaker from the browser, so it works on whatever the call is running on and there is nothing to change about your phones. One way for the engineer alone, two way for the engineer and the client. On the desktop and on a phone.
Engineers stop being willing to work without it, which is the most reliable adoption signal there is. It is also the fastest thing on this page to prove: there is no debate about whether it worked, because the ticket is either right or it is not.
Ask AI
One question, across everything you have ever written down.
The answer to most tickets already exists. It is in a resolution from eighteen months ago, or a page in IT Glue, or a document nobody has opened since it was written. The problem was never that the knowledge was missing. It was that finding it cost more than guessing.
Ask AI searches all of it at once and answers inside the ticket the engineer is already working. They can paste a screenshot in and keep going.
What one question returns
- The tickets that matter
- Previous relevant tickets, each summarized, with what actually resolved them.
- The documentation
- The pages that bear on this problem, pulled from wherever they live, with the source attached so the engineer can check it rather than trust it.
- Key observations
- What the engineer should know before they touch it. That the platform went out of support eighteen years ago, for instance, and is still carrying a critical application.
- Recommended next steps
- An ordered list, with the links to follow. Not a paragraph the engineer has to turn into a plan.
- The gap in your documentation
- When the answer would have been better with a page that does not exist, it says so, and offers to write it.
Ticketing
- ConnectWise
- Autotask
- HaloPSA
- ServiceNow
Documentation
- IT Glue
- Hudu
- SharePoint
- Confluence
Three sources, searched as one. Your tickets are the part no vendor can give you and no competitor can copy, because they are yours. How the data is handled.
The documentation crisis
Every MSP has tried to fix this with a project. Nobody has.
Documentation fails because it is nobody’s job and has no deadline, and because the only moment anyone knows a specific page is missing is the moment an engineer needs it and it is not there. Assist-X is in the ticket at that moment.
- 01
An engineer needs a page
and it is not there. That is the only moment anyone can see the gap.
- 02
The gap is flagged
against the client it belongs to, while the work is still happening.
- 03
A draft is written
from the resolution itself, and lands in IT Glue or Hudu, tracked in the PSA.
- 04
An engineer reviews it
and the next search finds it. Nothing was scheduled, and nobody ran a project.
The tracking ticket references the ticket that exposed the gap, so the reason the page was needed does not get lost between raising it and writing it.
Which makes this the only part of the platform that is worth more in month twelve than in month one.
It also answers the objection every owner raises about search: our documentation is a mess, so it will not work for us. That is not a prerequisite here. It is the thing being fixed.
The analysis
Every ticket is read, understood and scored.
Automatically, on every ticket, not when somebody remembers to ask for it. No survey goes out, no form is filled in, and no engineer is asked to rate their own work.
Twelve readings on every interaction. A manager spot-checks a handful of tickets a month; this reads all of them.
Disruption, for example
“I cannot print to the printer in my home office.”
A local printer, one user, low priority. Every system you have ever used scores that ticket down and moves on. Further into the body of the ticket, the person who cannot print is the CEO, and what they are trying to print is a tax return that is due in three days.
Disruption is not severity. It is how much of somebody’s day this is actually costing, which is a thing you can only find by reading the ticket.
Actions
The clerical half of the job, done in the ticket.
None of this is the work your client is paying for, and all of it is work your engineers are doing. It is also the part that gets skipped first when the queue is long, which is why time entries and notes are thin exactly when the week was busiest.
- Write the response
- A drafted reply to the client, in a professional, friendly or technical tone, at the length the engineer wants. They send it, so it is still their reply.
- Speak the notes
- Voice to text for the engineer who would rather talk than type, on desktop or on a phone.
- Enter the time
- The time entry, from the work. The single most reliably skipped task on a service desk.
- Log the expenses
- In the same place, at the same moment, rather than in a spreadsheet at month end.
What it tells you
The things nobody had time to notice.
Raised in the Assist-X panel inside the PSA, where your engineers are already working. Not another inbox, not another notification app, not somewhere they have to remember to look.
A VIP is on this ticket
The managing partner, not a graduate. Handled accordingly, before it becomes a phone call to you.
This is a multi-user event
Three tickets that look unrelated are one outage. Seen as one thing rather than three queues.
This is chronic
The same fault, the same site, the fourth time. A project conversation, not another repair.
This should be escalated
Scored one to ten with the reasons in plain language, including promises made and missed.
An executive needs to know
Raised so somebody senior has awareness before the client tells them.
This is outside the agreement
Possible excluded service, with the reason, before an engineer spends the afternoon on it.
There is an opportunity here
The ancient platform, the unsupported firewall, the project worth quoting. Written back to the PSA for an account manager.
The response quality is low
Not a judgment on the engineer, a flag on the interaction, while it can still be fixed.
Documentation is missing
Raised at the point of need, with a draft attached.
The agreement
You are giving work away. This catches it as it happens.
Under a fixed fee, out-of-scope work is not lost revenue on an invoice. It is margin leaving without anyone seeing it go.
Every request is checked against that client’s agreement in real time. Work that falls outside it is flagged before an engineer spends the afternoon on it, with the reason, so it can be billed, quoted or declined rather than quietly absorbed.
The same tab shows the engineer what that client actually bought. Which licenses, which backup, whether after-hours is in scope at all, with the financials stripped out. An engineer who can see the agreement stops guessing at it.
It runs the other way too. The unsupported platform and the project worth quoting are written back to the PSA as an opportunity assigned to an account manager, so spotting it and acting on it are the same event.
And then it compounds
The tool your engineers adopted is also the instrument.
Because Assist-X is present in every ticket, those twelve readings become something else entirely: client health by account, coaching by engineer, escalation risk by threshold, and the skills that decide who gets dispatched.
That is XOP Command, and it exists because engineers used a tool that saved them time. Adoption is not a nice-to-have. It is the whole supply chain.
Assist-X Mobile
The same assistant on a phone, for engineers on site, transcription included. It ships with Assist-X, because it is the same product rather than another purchase.
Onboarding
4–6
hours of your team’s time. Total.
Not a twelve-week implementation, because a twelve-week implementation is a twelve-week delay on the return. We spent a quarter automating this, and what is left is the part only you can do:
- API members and permissions
- In your PSA, scoped to what the agent actually reads and writes.
- Documentation and SharePoint access
- So search can reach the places your answers already live.
- Your pilot group
- The engineers who go first. Usually a subset, not the whole desk.
Most MSPs start with a pilot group rather than the whole desk, which is also how the commercial terms work. Engineers are the unit, so you add the rest when the first group has stopped arguing about it.
What it means to each of you
Four people have to agree, and the engineer is the one who decides.
The owner, and leadership
Your cost to serve comes down without your agreements changing.
Forty-five to sixty minutes back per engineer per day, against $55 a month. Under a fixed fee every minute of that is margin, and the same team carries more clients without hiring. That is the whole business case, and it does not depend on selling your clients anything new.
The strategic part is the second-order effect. Because Assist-X is present in every ticket, your business becomes measurable: how accounts actually feel, which issues are chronic, which promises were missed, what you are giving away. You cannot buy that separately, because it is a by-product of your engineers using something that saves them time.
The service manager
Nothing new to learn, and nothing new to remember to open.
It sits inside the PSA your engineers already have open all day. Every tool an MSP has abandoned was abandoned because it was one more place to go, so adoption here is not a feature bullet, it is the load-bearing assumption of the whole platform.
Every ticket gets read and scored without a survey going out or a form being filled in. The documentation holes you have been meaning to fix get flagged at the only moment anyone can see them, which is when an engineer needs a page that is not there.
Your engineers
The clerical half of the job stops being their job.
Notes, subject, classification and the time entry are written while the call is happening rather than reconstructed at five in the afternoon. Ask AI searches every resolution the team has ever written and all of your documentation at once, and answers inside the ticket they are already in.
They also stop accidentally giving work away, because each request is checked against the agreement in real time and an engineer can finally see what that client actually bought.
Your clients
Faster resolutions, and an engineer who already knows the history.
Time to resolution is the one thing a client genuinely feels, and it is what search and hands-free capture both attack. Nobody rediscovers a fix your team already found eighteen months ago.
Ticket notes stop being thin. When a client asks what happened on a call last March, the answer is a complete record rather than somebody's recollection of it.
The arithmetic
A 10x return on it, at minimum.
Forty-five to sixty minutes a day is 180 to 240 hours a year per engineer. At a fully loaded cost of $55 an hour that is $9,900 to $13,200 of recovered capacity against $660 of platform. We publish 10x rather than what that divides to, because a number you have to defend to your own board should be the conservative one. Most of your payroll is engineers, so across fifteen of them or a hundred it compounds.
The break-even is fifteen minutes a week. Everything past that is margin, and the same team carries more clients without hiring. $55 per engineer, per month.
Point it at tickets you already know.
The fastest way to judge this is on your own history, where you know what actually happened, and can see whether it tells you something you did not already know.