AI Won’t Fix a Broken MSP (Do This First)

Most MSPs are talking about AI, but most haven’t got the basics right.

What we’re talking about here is, number one, we’re screaming after AI like it’s the new shiny toy, but we haven’t actually decided what we’re going to fix. Why are we going after AI to fix things in our business?

Vendors will want you to drink that Kool‑Aid and buy tools with “AI” in the name. But have we actually defined what we’re trying to start with? Are we trying to improve resolutions, create more profitability, make our teams more performance‑related?

We need to think about why we’re going into AI. Like any other product, we should take it through an R&D process: we’re trying to fix this with this outcome for this cost.

You also need to be careful about systems and vendors talking about AI. A lot of what we’re seeing in the market today is actually automation or small AI trying to fix a rhythm. They’re trying to keep it very contained.

The problem with AI to date is: we ask it a question, it gives us 10 different answers. So they’ve made LLMs and the input very structured to make sure we get the right output every time.

The other thing is, if we’re trying to look at our own data, that data is shocking. So vendors have really tried to control it. In most cases, just like the cloud boom, they’ve rebranded the workflow and automation that was already in the system and called it AI.

That’s one of the basics we need to nail down.

Maximise PSA Automation Before You Chase New AI

In your current systems, are you using all the workflows and automation to improve your business? Typically, we’re seeing 5–10% utilisation of automation and workflows inside a PSA.

We want to really maximise that to drive the efficiency that’s there today, and make sure we’re not changing to a product that talks about AI when really it’s talking about something we already have.

The next thing we want to focus on is processes that output data. A good example is the service desk. We often see resolution notes like “fixed”, “rebooted and it’s fine”. There’s no detail about what we actually did to resolve this type of category and this type of fault.

Once we turn AI on, that kind of “it’s fixed” resolution is not going to help AI at all.

We don’t know exactly where AI will land in that transition path because there’s a lot of speculation, but we do know it will use data. That data is your crown jewels, like it’s always been.

So we need to fix processes and resolution codes, sales opportunity data, executive summaries and so on. Those will be the real focus when AI is ready.

My big push today is:

• Get your data correct
• Maximise automation in the system
• When you look at AI as a new tool, make sure you’ve got a full R&D process

The danger of jumping in too early is: we don’t know what we’re trying to fix, what our data looks like, or where we’re trying to take the business.

AI and the “People Will Disappear” Myth

Another big example people are talking about is people. Most MSPs talk all about people: people want people. You’ll hear “tier one service desk is going to disappear very quickly” and “if you’re not on the AI journey your MSP will be dead.”

We don’t think that’s true. AI is way off that path. And your customers are way off that journey.

Your customers already complain about AI chatbots on websites, chatbots to log a ticket, voice chatbots when they ring up. They want to speak to people. They’re not ready for a fully‑automated AI journey.

You’re also in a very different place to big corporates. They can force automation on customers; you’ve probably got 20–40 competitors locally. You can’t play that game and just force it onto your audience.

So some MSPs are pivoting. Just like some now say “we will come on‑site” when others stopped. They are saying: when you ring us, you will always get a person.

What those businesses are doing is looking at AI, data and structure and asking: how do we make that person more efficient?

They want the tier one person, who doesn’t have all the skills, to still appear to know everything about the customer: history of data, history of resolutions, top priorities. AI can feed them that information so the customer receives excellent service at low cost, while still getting the human touch.

That’s the first step of AI in an MSP: making individuals more efficient and better informed.

Practical AI Uses on the Service Desk

On the service desk today, there are some clear AI use cases.

Ticket handling and summarisation:
You might have a ticket with 20 activities and a long error at the beginning. A new tech picks it up as an escalation. AI can summarise:

• Core pain points
• What we’ve done
• What the next agreed action is

That’s a great efficiency: using data in the ticket to help the technician move it to closure quickly.

Response drafting and tone:
Spelling and grammar matter. End users make decisions about whether you provide good IT based on how those emails look and whether they’re too technical.

AI can take rough notes and turn them into a well‑written, well‑structured, non‑technical response in your tone. That means every customer gets the same quality of communication, regardless of which tech wrote the note.

Triage:
Smaller MSPs often don’t do triage correctly. They don’t have a dispatcher, don’t want the overhead, and don’t understand the value.

AI‑assisted triage can:

• Summarise the top line of the ticket
• Prioritise the ticket
• Categorise it
• Match it to the right resource based on skill and past success

For example, if you’ve had this ticket 50 times and Bill always fixes it, Bill gets the ticket. Or it routes to the right group (tier two vs tier one) so you don’t waste time.

AI in RMM and PSA Together

RMM is another area people think is “dying”, but it’s often just underused. The hardest part with RMM is setting up monitors and scripts. AI is being built in to help leverage best practices on monitoring and scripting.

AI will also let you use big data between RMM and PSA. For example:

• “Why is this one machine running slower than the other 19?”
• “What’s different: applications, size, age, performance?”

AI can summarise that for you. This has a massive impact on mean time to resolution, because techs get the information without digging.

AI Beyond Service – Marketing, Proposals and Exec Summaries

Marketing is a huge area where MSPs struggle: how to do marketing and sales efficiently, with limited resources, time and money.

AI can help you:

• Turn bullet‑point notes into a draft blog (that you then improve)
• Draft marketing copy, exec summaries and proposals
• Fix grammar and clarity so prospects don’t reject you over spelling

You don’t want to copy‑paste AI output blindly. Think of it as getting you 70% of the way there, quickly, then you rewrite and personalise so it doesn’t read like generic AI content.

The same applies to proposals and executive summaries. Use AI to make them clearer, better written and more consistent, then edit for accuracy and voice.

Automation vs AI – Foundation and Layer

Let’s be clear on terms:

• Automation / workflows / runbooks: fixed input, fixed output. Data point changes, trigger fires, action happens.
• AI: learns from what it sees in the data and suggests or generates different outputs based on patterns.

Example of classic automation:

• Ticket status changes on a certain date
• Date passes
• Workflow changes the status, sends an email, flags something

Example of AI layered onto that:

• A new ticket email comes in describing a blue screen that reoccurs every 30 minutes during patching
• Workflow still sends an automated acknowledgement
• AI reads the description and personalises the email: references the blue screen, the patch, the 30‑minute cycle, asks for specific screenshots

The customer feels like a tech wrote it, but it’s still automated and efficient. You get better data back because they reply with useful, relevant information.

That’s automation as foundation and AI as a smarter layer on top.

Designing Better Workflows in Your PSA

Common pitfalls in PSA workflow design:

• No structure
• Inconsistent boards, statuses and categories
• Services with multiple conflicting journeys and SLAs

You want:

• Consistent status structures across boards (incident, tier 1, tier 2, etc.)
• Common categories
• Repeatable journeys tied to services

Once you have that structure, you can build useful automation:

• Compliance checks (e.g. scheduled tickets that never got assigned)
• Status‑based triggers that change states or schedule work

Everyone in the business should know: if I change status to X, automation will do Y, no matter which board or tier I’m in.

Avoiding Email‑Based Noise in Automation

A big mistake with automation: everything outputs as an email to engineers.

The intention is good (alert them when customers respond or something goes out of compliance) but the outcome is: engineers get hundreds of emails and create rules to hide them.

Better: keep automation outputs inside the system.

Example:

• Change status to “Action Required” for stale or non‑compliant tickets
• Train engineers and coordinators to clear that pot three times a day
• Build dashboards and KPIs around that status

Now you’re tracking, not spamming inboxes. And later, AI can also work off that clean, structured data.

Automation as the Foundation, AI as the Layer on Top

Ultimately:

• Automation is the foundation: workflows, runbooks, status changes, structured journeys
• AI is the layer on top: using data to personalise, summarise, suggest and enhance

If you fix the basics – automation, data and structure – AI will become useful. If you don’t, it will create chaos in your business.

If you don’t want to make the common mistakes MSPs are making with AI, start by fixing your PSA workflows, your service desk data, and your core processes – then look at AI as an enhancement, not a rescue plan.