In this article 6 sections
Search on steroids
Darren Strong, founder of Scalable, and Mit Patel, founder and former owner of London MSP Netstar, co-host this scaleUP episode. Their guest is Morgan Aspinall, Head of Product at HaloPSA. The topic is AI versus automation: how to tell them apart and where an MSP should start.
Darren compares AI to the cloud, with vendors rebranding what they already did. Morgan agrees that vendors, Halo included, have jumped on the bandwagon. His view is that AI is “what searching should have been” 5 or 10 years ago. It organises the data in your PSA, RMM and other tools and presents it clearly. The conversation turns to asking AI for restaurants nearby with a high score and hundreds of reviews, then asking it to explain its picks. The label that sticks is “search on steroids”.
Darren’s concern is that AI does not give the same answer every time, which makes it hard to build into a process. Morgan draws the line there. Automation has fixed parameters and a fixed outcome, such as SLA reminders. AI is fuzzier, matching on meaning rather than keywords.
Vet tools before you buy them
On vetting the flood of tools, Mit wants a grown-up R&D process. He admits that 15 years ago he came back from vendor halls with new shiny tools, and he sees many MSPs using about 10% of what they buy. First make what you already have better. For anything new, check commercial sense, client fit and technical fit, score the vendors, and confirm you can support it.
Morgan tells of MSPs asking him to implement AI who cannot say what for. His question is what problem AI solves, and how many of those problems automation could solve instead.
Data, privacy and where to start
Darren raises staff pasting HR contracts and personal data into ChatGPT. Morgan says to read the retention terms. Mit mentions private models on Microsoft Azure that some MSPs use to keep data contained.
Morgan’s advice is to start with automation. Find the mundane, repetitive work and automate it as far as possible, then ask how AI could take it further. His quick wins are integrating platforms and automating billing reconciliation from the RMM, CSP and distributor into Halo. He has heard of MSPs saving a full day’s work a month. Darren adds that change requests now outnumber incidents for most MSPs, and standard ones can run through a self-service portal. Automation fails, he says, without a standard blueprint for every customer. Morgan describes Halo’s AI triage, which matches new tickets to similar past ones and suggests or applies the agent and category.
Jobs, data quality and training
Darren sees AI creating better jobs and reusing talent. When Morgan looked at ticket data for AI, many resolutions just said closed or problem fixed. Mit’s view is that AI is at level 1 heading for level 10, so MSPs should get their house in order and clean their data now. On training, Mit says teams that are not taught to use AI will fall behind, as engineers once had to be taught to use Google.
Where it goes next
Morgan expects AI to sit above the tools and see the whole stack. He also argues the MSP differentiator will be picking up the phone to a human. Mit expects a practice to grow around AI security, with frameworks and compliance to follow. He cites a claim that endpoints per engineer could rise from around 200 to 240 today to 600. Darren puts today’s figure nearer 260 to 280 and expects it to rise, but will not commit to 600. Mit thinks it is not possible without automation and AI.
The closing points. Darren: be intentional and plan before you buy. Morgan: AI is a tool, not a fix for everything. Mit: “people that use AI are going to replace the people that don’t use AI”, so train your people.
Common questions
What is the difference between AI and automation?
Morgan’s distinction: automation follows fixed parameters to a fixed outcome every time. AI is fuzzier and can give different answers, which helps search but makes it harder to build into a process.
Should an MSP start with AI or automation?
Automation. Morgan’s advice is to automate mundane, repetitive work first, then ask where AI could take it further.
Where are the quick automation wins?
Integrating platforms and automating billing reconciliation, SLA timers and follow-ups, per Morgan. Darren adds standard change requests through a portal.
How should an MSP vet a new tool?
Mit’s process: improve what you have first, then assess commercial, client and technical fit, score vendors and confirm you can support it.
What is the Assurix Trustmark?
The live trustmark for MSPs. Assurix independently assesses 59 controls: 40 security controls aligned to NCSC CAF 4.0 and 19 operational controls built for MSPs. Live-measured controls are checked every six hours and the rest on a set cycle. An Assurix assessor signs off. A failure left unfixed for 30 days suspends the Trustmark publicly.
