AI savings are double-counted when the same freed-up capacity or the same avoided work is presented as a saving in multiple places within a business case. This happens more often than you think, and it undermines the credibility of your business case as soon as financial managers or the board look critically at the figures. The questions below help you build a fair, internally defensible calculation.
Why are AI savings so often double-counted?
AI savings are double-counted because different departments or project teams independently present the same efficiency gain as a benefit. An automation project that saves ten percent of talk time is simultaneously claimed by Operations (less FTE deployment), IT (lower system costs), and Finance (shorter lead times). The same freed-up hours then count three times.
The core of the problem lies in the way business cases are constructed. Teams work in silos and each have their own KPIs. If an AI application in the contact center shortens talk time, the operational manager sees capacity gain, the HR manager sees a lower recruitment need, and the IT manager sees a lower load on systems. All legitimate observations, but if they are added up without checking if they describe the same underlying saving, an inflated picture is created.
In addition, AI projects are often presented with a combination of hard savings and soft benefits such as higher customer satisfaction or lower employee turnover. When those soft benefits are subsequently converted into a monetary amount and added to the hard savings, a double count is easily made.
Which cost items are most susceptible to double counting?
The cost items most susceptible to double counting in AI projects in customer contact are personnel costs, handling time, repeat contact, and system costs. These are exactly the items that are tracked and reported by multiple departments simultaneously.
- Personnel costs: Freed-up hours are presented as both FTE reduction and productivity gain, while in reality, they are the same hours.
- Handling time (AHT): A shorter average talk time leads to less capacity pressure, but that gain is sometimes also counted separately as a quality improvement or as a reduction in telephony costs.
- Repeat contact: Less repeat contact means fewer calls, less employee deployment, and higher customer satisfaction. If all these effects are quantified separately and added together, the original volume reduction counts multiple times.
- System costs: Consolidation of platforms yields license savings, but these are sometimes also presented as operational savings because employees spend less time switching between systems.
Recognizing this overlap is the first step. The second step is determining which saving is the primary one and which effects are derived from it.
How do you calculate the actual net savings of AI?
You calculate the actual net savings of AI by tracing all claimed benefits back to one common unit, usually hours or euros, and then checking if the same unit is not already included elsewhere in the calculation. Then, subtract the full implementation and management costs from the gross savings.
A practical approach consists of three steps:
- Inventory all claimed benefits from all involved departments. For each benefit, write down which input forms the basis: hours, calls, FTE, or license costs.
- Consolidate overlapping inputs. If multiple benefits are based on the same freed-up hours, count those hours only once. Then determine how those hours are utilized: are they actually converted into lower staffing levels, or are they reallocated to other tasks?
- Subtract all costs. Think of implementation costs, licenses, management, training, and the internal hours spent on guidance and change management. Savings that only materialize after two years must be discounted.
The result is a net saving that you can defend, even if someone requests the underlying assumptions. A good business analysis records those assumptions in advance, so you are not faced with surprises later.
What is the difference between avoided costs and realized savings?
Avoided costs are costs that you do not incur thanks to AI, but which are also not visible in the accounting. Realized savings are costs that have actually decreased and are measurably reflected in the figures. This distinction is crucial for the credibility of an AI business case.
An example: if AI ensures that you do not have to hire extra employees despite growing contact volume, those are avoided costs. Staffing levels do not rise, but they do not fall either. You save on future recruitment and training costs, but those were never actually spent. Finance rightly sees this as a softer benefit than a demonstrable decrease in wage costs.
Realized savings are more concrete: lower license costs because you consolidate platforms, lower staffing levels because AI fully handles part of the contact volume, or fewer overtime hours because peaks are better absorbed. These savings are directly visible in the budget.
In a strong business case, you name both categories separately and indicate which part of the total saving is hard and which part is soft. That prevents disappointment afterwards and increases credibility with the board and Finance.
How do you build an AI business case that is internally credible?
You build an internally credible AI business case by making the assumptions transparent, splitting the savings into hard and soft benefits, and fully including the costs, including implementation, management, and change management. Involve Finance and the board early in the process so they validate the methodology before the outcomes are presented.
Concrete elements of a credible business case:
- One owner of the calculation. Do not let every department add up its own saving. Appoint one person or team to manage the total business case and flag double counting.
- Baseline scenario without AI. Explicitly describe what happens if you do nothing: more FTE with growth, higher wait times, declining customer satisfaction. That makes the value of AI more tangible.
- Phased realization. Indicate per phase which saving is expected when. A business case that plans all benefits in year one is less credible than one that takes a ramp-up period into account.
- Distinction between capacity gain and cost reduction. Freed-up hours are only a saving if they lead to lower staffing levels or reallocation to higher value. Explicitly state which choice you make.
- Risk scenarios. Show what happens if AI adoption goes slower than expected or if the volume turns out differently. That shows you are being realistic.
The business case process is at least as important as the outcome. A well-founded calculation that has been internally validated reaches the finish line much more often than an optimistic estimate that is later shot down.
How Pegamento helps with a fair AI business case for customer contact
We help organizations build an AI business case that holds up internally, even if Finance or the board critically scrutinizes the assumptions. We do this not with separate advisory documents, but as a partner that guides the entire trajectory: from analysis to implementation and ongoing optimization. Everything under one roof, without silos or complex supplier constructions.
Specifically, we help you with:
- Mapping the actual cost structure of your customer contact environment, including hidden costs and repeat contact
- Identifying AI applications with a demonstrable return, such as smart call routing, automated handling of frequently asked questions, or intelligent assistance for employees
- Splitting hard and soft savings and monitoring consistency in the calculation, so that double counting is prevented
- Guiding the implementation with proven modules that align with your processes and systems—no costly custom work, but a smart combination of standard building blocks
- Process automation, which we position as Agentic AI: an evolution from executive bots to self-thinking assistants that do not just follow instructions, but independently take initiative and act
Do you want to know what AI realistically delivers for your contact center? Contact us and we will look together at what a fair, defensible business case looks like for your situation.
Frequently Asked Questions
How do I know if my current business case already contains double counting?
Check if multiple departments have presented benefits based on the same freed-up hours, the same volume difference, or the same cost item. A practical test: trace each claimed benefit back to the underlying input (hours, calls, or euros) and see if that input already appears elsewhere in the calculation. If so, you have a double count. When in doubt, have an independent party or Finance scrutinize the calculation before presenting it to the board.
What do I do with freed-up hours if staffing levels do not actually decrease?
Freed-up hours that do not lead to lower staffing levels are not a direct cost saving but a capacity gain. You can still present those hours as a benefit, but you must explicitly state where they are going: higher quality, handling more complex customer questions, or internal projects. Clearly indicate in the business case which choice you make and avoid presenting the same hours as both cost reduction and productivity gain.
How do I handle soft benefits such as higher customer satisfaction in my calculation?
Always include soft benefits separately in the business case and do not mix them with hard savings in the total amount. You can quantify soft benefits, for example through lower churn or higher NPS, but be transparent about the assumptions you use for that. A credible approach presents hard and soft benefits in separate columns, so that Finance and the board can judge for themselves how much weight to assign to each part.
Which costs do organizations most often forget to include in an AI business case?
The most forgotten cost items are change management, internal guidance hours, and the time employees spend on adoption and training. In addition, ongoing management costs, such as model maintenance, data quality monitoring, and periodic recalibration of AI models, are structurally underestimated. Also count the costs of integrations with existing systems, as these often run higher in practice than previously estimated.
How quickly can I expect realistic AI savings in a contact center environment?
For the first measurable results, count on a ramp-up period of three to six months, depending on the complexity of the implementation and the degree of adoption among employees. Full realization of the planned savings usually lies between twelve and twenty-four months. A business case that plans all benefits in the first quarter is difficult to defend; phase the expected benefits per quarter and build in a realistic adoption path.
Do I need to make a separate business case for every AI application?
Not necessarily, but it is wise to make the costs and benefits per application visible separately, even if you combine them into one overarching business case. This prevents a high-yielding application from masking weaker components, and allows you to set priorities if the budget comes under pressure. One consolidated business case with a clear breakdown per application provides both overview and detail.
How do I involve Finance early enough in the business case process without it slowing down speed?
Involve Finance already in the phase where you establish the assumptions and calculation methodology, not just when the calculation is finished. Ask them to validate the definitions used, such as what counts as a realized saving versus an avoided cost, and record those agreements in writing. This prevents discussions afterwards and actually speeds up approval, because Finance recognizes and has co-signed the methodology.
