AI impact MSP marges
Is AI squeezing your billable hour?
AI presses on the margin for standard work: a client now gets a six out of a tool as well. MSPs that want to protect the billable hour have to perform above that level consistently, otherwise the margin quietly walks away.
MSP Late Night · 8 October 2026
Where do MSPs really lose margin to AI?
The first productivity gain is visible: in their own customer service, 50% of cases are resolved autonomously by AI. At the same time, loose AI ideas keep popping up that are task driven, with no steering on output. Without focus on what the role actually has to deliver, the time saved evaporates into scattered experiments and the pressure on billable hours stays exactly where it was.
A six is something an AI tool can do as well. So you will have to score higher than that.
Frequently asked questions
How much service desk work can AI genuinely handle on its own?
In practice, customer service teams already see around 50% of cases resolved autonomously by AI, provided the knowledge base and handling flows are structured. That mainly covers repetitive tier 1 questions: password resets, standard configurations, status updates. More complex incidents and conversations that need trust or nuance stay human work, and autonomous handling makes them more visible in the workload.
Why does AI hit the time and materials model in particular?
The classic MSP model charges for time, not for outcome. The moment an AI tool delivers the same result at an adequate level, the hour loses its price anchor. The client is no longer comparing you with another engineer, but with a tool that is available around the clock at marginal cost. MSPs that keep billing hours for work that can be automated at a six will see their margin shrink.
What does output driven work mean in practice for an MSP team?
Output driven means steering on what the role has to deliver, not on which tasks someone performs. An engineer is judged on solved client problems and stability, not on tickets closed. That asks for clear definitions of what good looks like per role, and for cutting loose AI experiments that have no measurable effect on client outcome or margin.
Does AI really save time inside an MSP?
The assumption that AI saves time usually holds at task level, but not always across the day. Freed up time tends to fill with new tools, more messages and faster replies, without any gain in depth or client trust. Without a deliberate choice about where the freed hours go, the gain disappears into extra availability instead of better output.
The full conversation, transcribed
En we zien nu wel de eerste productiviteit. Ik zie het bij onze eigen klant en bij onze eigen customer service. 50% van de cases kunnen wij nu gewoon autonoom oplossen. Maar focus. Ik mis vaak focus. Want overal poppen losse AI-ideeën op die vooral heel taakgedreven zijn. Alleen output is uiteindelijk wat je levert en dat is het belangrijkste. Dat is uiteindelijk waarvoor je rol bestaat. Ik denk dat dat een beetje de trend is waar we nu zitten. We zien onze kaas wordt er een beetje aan gesnoept. Onze uurtje factuurtjes staan onder druk. Een 6 kan je met een AI-tool ook doen. Dus je zult als persoon minimaal een hoger score moeten gaan halen dan dat. Want iedereen kan er voldoende halen. We bereiken elkaar heel makkelijk, maar bereiken we elkaar echt nog? Met andere woorden, die diepgang, die verbinding, dat vertrouwen creëren. Echt het gesprek aangaan met elkaar. We zeggen allemaal we besparen tijd met AI. Het maakt ons persoonlijk effectiever. Maar besparen we nou echt tijd? Muziek