Password resets and the same ten requests bury my techs every week

It is 7:40 on a Wednesday morning and the ticket queue already has thirty one items before anyone has had coffee: four password resets, two printer mapping tickets at the law office that always has printer trouble, an MFA re-enrollment, and a monitoring alert nobody looked at since Friday. None of it is hard. All of it is yours, or your two best technicians’, because that is who is left. Password resets and the same ten requests bury your techs every week, not for lack of thinking about it, but because there has never been a day free to fix it.

When MSP owners talk to me, this is where it starts. Not the AI pitch for next quarter’s client meeting, but how to get repeat tickets off two overworked technicians, how to stop the after-hours line from being their own cell phone, how to make the monitoring tool, ticketing system and documentation agree, and how to turn the client asking about AI into a line item instead of a favor.

I am Isaac Krupp. I spend one full working day inside a business, watch how the work actually happens, and write down where AI pays off and where it would just be a new thing to maintain. I sell no software, build nothing custom, and take no referral fee. You run technology for a living; this page skips the primer.

What AI for MSPs can actually do right now

Ordered by how much of your week each one gives back.

Repetitive tickets eat the week

Your senior people spend real hours every week closing tickets a script could close. LTVplus found 55% of MSPs had automated no more than a quarter of their ticket volume, and a separate study found 58% of IT teams lose over five hours a week to repetitive requests. The fix sits inside the ticketing tool you already run: an AI layer tags each ticket, sets priority, drafts the first reply, and for a printer reinstall or another fully routine case, runs an approved script and closes the loop. A password reset or MFA re-enrollment still needs a technician to verify the requester’s identity against your standard process before it goes through, since those are exactly the requests an attacker impersonates to take over an account; the AI drafts the steps, a person confirms who is asking and approves the change. The weekend queue is half cleared by Monday. What it never touches: a server, a firewall, or backups and permissions, which wait for a technician.

Owner’s cell phone is the night shift

At a shop with two to eight people, you are the overnight helpdesk, and Konnect calls that pattern the core driver of MSP owner burnout. The fix is an AI answering service on the after-hours line. It answers in your company’s name, checks the caller against your client list, asks the questions you would ask, logs a ticket with a transcript, and wakes a person only for your written emergency definition, server down, ransomware, whole office offline. What it never does: promise a fix time, walk a caller through remote access, or reset a credential by phone. You write and review the escalation rules monthly.

Tools that do not talk to each other

Your monitoring tool, ticketing system, documentation and invoicing each hold a piece of the truth, and someone re-keys between them all day. Kaseya found 95% of MSPs treat connecting those tools as essential. The fix is a no-code automation platform, the kind people search for as workflow automation software, wired into what you run. An alert in monitoring becomes a ticket with the right client and site filled in, and a closed ticket updates the time entry and the docs. What it never does: push a change into a production system or write to an invoice unreviewed.

Client documentation lives in heads

Half your network details live in one technician’s memory, and when that person leaves, so does the knowledge. MSPs reporting trouble keeping documentation consistent rose to 17% from 10% in a year. The fix is an AI assistant on your documentation and ticketing tools. It drafts an entry from every resolved ticket, what changed, on which device, flags entries that contradict each other or sit untouched a year, and answers a technician’s question in plain language. What it never does: store a credential outside the vault, publish anything unread, or become the only place a fact lives.

Winning new clients is the hardest part

Every provider offers roughly the same thing, and most of your growth means taking a client from someone else. Kaseya found 71% of MSPs name new customer acquisition as their top challenge, and LTVplus found only 12% of new clients are outsourcing IT for the first time, while 33% are switching providers. The fix is two tools: a proposal generator turning a discovery call and pricing sheet into a first draft you edit, and a reporting tool turning ticket, uptime and patch data into a plain-English quarterly review. What it never does: set prices, promise a service level you cannot deliver, or send anything unread.

Cannot find or keep technicians

Hiring takes months, you cannot match enterprise pay, and when a technician leaves you are back on tickets yourself. MSPs reporting trouble hiring skilled technicians rose from 9% to 16% in a year, and Konnect found helpdesk roles see roughly 40% annual turnover at around $12,000 per departure. AI does not fill the seat, it raises what the seat can handle. A copilot inside the ticketing tool suggests the next step, surfaces past tickets, and drafts the client update, so a junior technician resolves what used to get escalated. What it never does: evaluate, hire or fire anyone, or replace training. Suggestions are labeled suggestions.

Clients ask about AI, nobody sells it

Clients keep asking what you are doing about AI, and you have not turned the question into an invoice line. Kaseya found 48% of MSPs rank AI and automation as the top client need for 2026, while only 13% earn meaningful revenue from it. The fix is a packaged offer, not a purchase: a usage policy and approved-tools list, business-tier assistant seats under the client’s identity provider, a short training on meeting notes and email drafting, and a quarterly check on what staff use. What it never does: become a custom model, connect to a client’s email or CRM unreviewed, or sell into regulated data unchecked.

Smaller contracts and slower invoicing

Deals are getting smaller, so it takes more accounts for the same revenue: more invoices, more disputes, more chasing than a small shop can staff. The share of MSPs whose typical customer spends over $25,000 a year fell to 41% from 75%, and LTVplus found unprofitable MSPs doubled from 5% to 10%. The fix is invoice automation tied to your ticketing and accounting tools: time entries roll into a draft invoice on the first of the month, the AI flags entries that look off, a forty hour ticket, a jumped seat count, and sends reminders. What it never does: change contract terms, discount an invoice, or chase a client you handle yourself.

Security expectations outrun the team

Clients expect you to be their security expert, the products are complicated, and you cannot hire anyone who knows them. Kaseya found 61% of MSPs say clients depend on them for cybersecurity guidance and 50% cite product complexity as a barrier, and 76% say security is what clients worry about most. The narrowest fix here is AI alert triage inside the tools you already run: grouping related alerts, writing a plain-English summary, and ranking what a person looks at first, plus drafted awareness emails and client summaries. What it never does: close an incident, isolate or wipe a device, or talk to a client about a breach. A named human owns every incident decision.

A day in a MSP with AI turned on

Meet Renata, who runs an eight person MSP out of Anaheim. Here is a Monday, before and after.

Before, her phone rang twice over the weekend, once for a real outage at a dental office, once for a user locked out after a new laptop. She opens the queue at eight to forty four tickets, a dozen unassigned, and spends the first half hour finding the one that matters. Her senior technician spends the morning on printer trouble and password resets, at a fully loaded cost of, say, $70 an hour, while a junior technician escalates a ticket already fixed back in April and never written down. A monitoring alert sits unread since Friday, a proposal promised Thursday is still blank, two invoices are disputed, and a client emails again about AI.

After, both weekend calls went to the after-hours agent first. The lockout became a ticket with a transcript and waited for morning. The real outage woke Renata, correctly, within two minutes. At eight the queue is sorted, the routine resets closed overnight, and the manufacturing alert is already a ticket with the client and site filled in. The junior technician’s ticket is solved in five minutes because the copilot surfaces the April fix from a documentation entry the assistant wrote the day it happened. The proposal is a draft on Renata’s screen by nine, twenty minutes to edit instead of two hours to write. The invoices went out flagged and reviewed, with, say, one dispute instead of three, and the AI question gets a reply with a one page offer attached. Renata is home for dinner. Same eight people, same tools, same clients.

What it costs, honestly

Almost everything above adds onto a tool you already pay for. Ticket triage, copilots and documentation assistants usually run tens of dollars per technician monthly. An after-hours AI phone agent is usually around $100 to $400 a month depending on call volume, well under one on-call shift. A no-code automation platform is often free to a few tens of dollars monthly at small volumes, rising toward a couple hundred for heavy use. Proposal and accounts receivable tools each land between roughly $50 and a couple hundred a month, and resold assistant seats usually run $20 to $30 per user monthly. Across a whole small MSP, plan on a few hundred dollars a month, not a five figure project.

The expensive mistake, and MSP owners are more prone to it than most because you can build it yourself, is commissioning custom software before you have used the off-the-shelf version for six months. Wiring a model to your PSA’s API over a weekend is easy; maintaining it while you carry forty tickets and an on-call phone is not. Use the cheap version first, and decide from there.

What NOT to hand to AI in a MSP

Production infrastructure. Servers, firewalls, backups, permission changes, account deletion. A script gets staged and suggested; a technician confirms it.

Credentials. No AI tool generates, stores or repeats a password outside the vault, on a call, or in a document.

Incident decisions. Whether an alert is a real breach, whether a device gets isolated, what a client is told and when. AI ranks and summarizes; a named human decides.

Money and the relationship. Pricing, contract terms, service level promises, discounts, write-offs, and any past-due conversation you would rather handle yourself.

Regulated client data. Health, legal and financial clients get no AI rollout until you have checked the vendor’s compliance documentation and the client’s own obligations.

People. Hiring, evaluation and security training stay human. A copilot is a suggestion, never a substitute for a junior technician learning the trade.

Anything a client has not seen. Proposals, reviews, documentation entries and security summaries stay drafts until read.

How the free AI day works for a MSP owner

I spend one full working day inside your shop. On site in Los Angeles or Southern California, by video anywhere else in the country. No card, no contract, no software pitch.

In the morning I watch. I sit with whoever dispatches tickets and see how the phones get answered. I pull the last few months of closed tickets by category to find the types that make up most of the volume, and ask for your call log from the last two weeks of nights and weekends to sort those calls into real emergencies and things that could have waited. I map where a technician copies information between screens, and ask a technician who did not onboard a client to work through that network from documentation alone.

In the afternoon we talk, walking through the pains above in the order they cost you most, and if a client has been asking about AI, we draft the one page offer together before I leave. The how it works page covers the rest of the day.

By the time I leave, you have a written plan: three to five AI moves, in order, with honest monthly costs and a note on what each must never touch. If the honest answer is to change nothing yet, the plan says that.

I also spend days like this with businesses that are often your own clients, so AI for accountants and AI for architects read like a preview of your next quarterly review.

If you want your evenings back and a real answer the next time a client asks about AI, book a free AI day and I will come see how your shop actually runs — the goal is a weekend where your phone isn’t the escalation path, and the ticket still gets handled with the same care.