All case studies Download PDF

A Perth MSP cut its L1 desk from six people to three

How AI workflows took the routine load off a short staffed service desk, and saved around $200,000.

Client20 person MSP, Perth

PlatformHalo PSA and n8n

Result$200,000 saved

$200,000saved, with the L1 desk down from six people to three

Client: 20 person MSP, Perth
Challenge: L1 desk short staffed after turnover, with no good candidates to hire
Solution: A suite of AI workflows that carries the routine L1 load inside Halo PSA
Tools: Halo PSA, n8n, Claude Sonnet 5.5, Claude Opus 5.5, GPT-6.1 Sol

The situation

Staff turnover had left holes in the service desk, and the hiring market was not filling them. Good L1 people were hard to find, and every new hire still needed months before they pulled their weight.

The team that remained covered the empty seats. Queues grew, tickets sat unassigned, and customers waited longer for answers. Client issues started to fall through the cracks, and the owner knew that churn was the next step.

We did not try to hire faster. We took the repetitive work off the desk.

What we built

We did not try to hire faster. We took the repetitive work off the desk. Each workflow handles one job a level 1 tech used to do by hand:

Categorise and prioritise
Every new ticket is read, categorised and given a priority against the MSP's own rules and client agreements.
Assign
The ticket goes straight to the right tech, so nothing waits in an unassigned queue.
Look up knowledge
Past fixes and knowledge articles that match the problem are attached to the ticket before a tech opens it.
Recommend the next step
Stuck tickets get a suggested next action, so they keep moving.
Create knowledge
Resolved tickets become new knowledge articles, so the knowledge base grows every day.
Chase customers
Tickets waiting on a customer reply get followed up on a schedule, with no tech time spent.

Halo PSAn8nClaude Sonnet 5.5Claude Opus 5.5GPT-6.1 Sol

Halo PSA triggers each workflow, and n8n runs it. Each step uses the AI model that suits the job: fast models for high volume sorting, stronger models where the step needs judgement.

The result

$200ksaved on desk staffing

6 to 3people on the L1 desk

6AI workflows running inside Halo PSA

The L1 desk now runs with three people instead of six, saving around $200,000. Every ticket arrives sorted, assigned and with a likely fix attached. The hiring problem that started it all no longer exists.

Why it worked

Most L1 time goes on sorting, searching and chasing, not on fixing. AI does those parts well and never gets tired of them. The team kept the work that needs a person: the judgement calls and the customer conversations.