Blog//

AI & Automation

,

Service Desk

,

The Business Case That Keeps You Up at Night

August 3, 2026

July 31, 2026

Some call it "labor efficiency" or "FTE optimization." Some don't acknowledge it at all, at least not out loud. It's that value in the business case deck created to present projected savings for the new AI program. Somewhere near the top of the Future State column, it's there: reduced headcount. The number of people on the team who may no longer have jobs.

The CFO's lens is rational, even if it's incomplete

IT labor is one of the largest controllable cost lines in a mid-market or enterprise IT budget. Service desk functions are notoriously labor-intensive and difficult to scale without also increasing headcount.

So it makes sense that AI vendors, analyst reports, and case studies have all led with the same headline ROI of labor displacement. Labor costs are relatively easy to measure, so it should be easy to measure AI’s value through its impact on labor costs. If AI deflects 50% of ticket volume, it should reduce demand on the service desk by 50%, meaning a 20-person team could be reduced to 10. 

Rational, but incomplete.

Most service desks are not running at perfect efficiency before AI arrives. They are either understaffed — carrying backlog, triaging what gets answered, asking agents to absorb more than the role was designed for — or overstaffed relative to current demand, usually as a hangover from over-hiring during a growth period. A small number are genuinely well-matched to their volume.

Each situation points to a different outcome when AI enters the picture. 

  • For an understaffed team, deflection doesn't free up headcount. Rather, it finally gives the team the support to do the job properly. Cutting staff in that scenario doesn't save money; it recreates the same problem at lower volume. 
  • For an overstaffed team, deflection forces a conversation that was probably already overdue. But that conversation should be grounded in actual staffing data, not a deflection model. 
  • For a well-matched team, deflection creates genuine capacity, and what to do with it is worth thinking about carefully.

The business case that skips that diagnostic and goes straight to the reduction model is the reason Gartner predicts that half of companies that cut customer service staff due to AI will rehire for the same functions by 2027, often under different job titles. It’s the same reason Nvidia CEO, Jensen Huang, said in a recent interview with CNA, “It's more likely that the companies with ambition will be more productive, they will do things faster, their company will increase in velocity. As a result, they become larger, more profitable. When they become larger, more profitable, they'll end up hiring more people."

Reframing the conversation without abandoning the business case

When AI deflects a ticket, what happens to the human labor that would have handled it? 

The default answer is cost savings — reduced headcount, improved deflection rate, a number on a slide.

Another option is to reinvest it, assign those analysts to other tasks that will bring in more value for the company:

  • Digital experience engineering: Beyond solving tickets, someone has to become the ambassador for AI integration and adoption throughout the workforce. Analysts are uniquely skilled to help users modernize and adopt an AI-fueled productivity toolkit.
  • Proactive problem management: Analysts reviewing recurring incident patterns to prevent tickets from generating in the first place

  • Knowledge engineering: Building, curating, and continuously improving the knowledge base that AI deflection runs on

  • Onboarding and change management support: Structured programs that absorb demand spikes during growth periods or technology transitions

  • L2 and L3 escalation ownership: With lower-tier volume deflected, agents can own more complex work more systematically

None of this happens automatically. It requires a deliberate decision to treat efficiency gains as reinvestment capacity rather than pure cost reduction.

The reinvestment case is harder to make than the reduction case, but it's more defensible over time. Headcount reduction is a one-time saving that shows up cleanly in Year 1 and plateaus. Reinvestment compounds: a better-maintained knowledge base improves deflection rates, better deflection rates reduce incident volume, lower incident volume reduces support burden during periods of growth. The CFO who understands that trajectory will respond to it.

Making that case requires a few things that most business cases skip. The reinvestment options need to be defined specifically — not "agents will handle higher-value work" but a named function, a named owner, and a named budget. The transition plan needs to be real enough to execute, not just optimistic enough to present. And the reinvestment scenario needs to sit alongside the reduction scenario in the deck, so the CFO is choosing between two modeled paths rather than approving the only one on the table.

One cost that almost never appears in either scenario: knowledge engineering. The deflection rate a vendor quotes assumes that a knowledge base already exists, is well-organized, and is actively maintained. Building it from scratch — or bringing a neglected one up to standard — takes time and people. That expense belongs in the model. Leaving it out makes the reinvestment case look more expensive than it is relative to a reduction model that's also undercounting its true costs.

How Reinvestment Compounds

A service desk of 20 agents handling 10,000 tickets per month. AI deflection reaches 50% over 18 months, reducing volume to 5,000 tickets.

Reduction Model

Reduce staff to 10 agents.

Reinvestment Model

Retain 15 agents and reallocate the equivalent of 5 FTE in capacity:

  • 2 FTE equivalent toward proactive problem management (target: 15% further ticket reduction in 12 months)
  • 2 FTE equivalent toward knowledge engineering (target: improve deflection rate from 50% to 65%)
  • 1 FTE equivalent toward onboarding support during a planned systems migration

The Math

If knowledge engineering lifts deflection from 50% to 65%, that generates an additional 1,500 tickets deflected per month. Proactive problem management targeting a 15% reduction adds another 750.

Total ticket volume drops to approximately 3,750, handled by 15 agents at higher quality than 10 agents could have managed.

How your MSP can help

A good MSP partner treats workforce modeling as a shared exercise — making the relationship between ticket volume, staffing requirements, and service outcomes visible to the client from the start.

That means knowledge transfer and cross-training built in as an ongoing practice, and a business case built for two audiences: the CFO who needs to see the numbers, and the team that needs to see that someone is thinking about what happens to them.

Some roles will change and some will disappear in a serious AI deployment. A partner worth working with will say so, model it honestly, and bring a reinvestment argument specific enough to execute. 

The better way forward for everyone

There's a version of the future of work with AI that doesn't require choosing between the CFO and the team. That case is harder to make, but it's also the one most likely to hold up to scrutiny, time, and to the people in the room who will remember how this decision was made.

If that number in the Future State column still keeps you up at night, it might just mean you're building the wrong case.

Ready to change course?

Contact us

What to Ask Before Finalizing Your AI Business Case

  • Where is the higher-value work, specifically? Which roles, which functions, which skills, and who’s responsible for the transition?
  • What’s the timeline and budget for upskilling the people whose roles change? Is it realistic? Is anyone acting on it?
  • What happens to service quality during the deflection ramp? Who absorbs the gap between projected and actual AI performance?
  • What is the knowledge engineering cost? Who builds and maintains the content that makes AI deflection work, and is that factored in?
  • What does the reinvestment scenario look like alongside the reduction scenario? Has the CFO seen both?
  • If deflection underperforms by 20%, what does the model look like, and who’s accountable?

About the author

Josh Spring

Practice Head, Service Desk

With 4 years at Astreya, Josh brings over 15 years of experience in a wide variety of End User Services environments and sectors. Key focus areas have been developing our AI-First Service Desk GTM strategy, our comprehensive library of Astreya Support Methodology, defining our Service Desk implementation & training strategy to sustain high-caliber service delivery, and deploying our Service Desk maturity framework for program evaluation & transformation roadmapping.

No items found.
AI & Automation
Service Desk