The Future of Optimization in Contact Centers: How Much is Enough?

Two weeks ago, I shared a free tool for "on the spot capacity planning" - for when the C-Suite needs answers fast. This week, I've built another free tool for the community to use, inspired in part when I ran a quantum circuit on IBM's real quantum hardware:

This exercise got me thinking about the limits of optimization in contact centers. What could this quantum technology mean for workforce optimization, specifically for scheduling?

For decades, we've been chasing better forecasting models, improved scheduling techniques, and more efficient ways to match supply with demand. Now, with AI advancements accelerating and quantum computing on the horizon, the big question isn't just about what's possible—but about what's truly necessary:

• At what point does additional optimization yield diminishing returns?
• If we could build a "perfect" forecast and scheduling model, would it even matter in the real world?

Below, I'll explore why the real key to contact center success is not be pursuing mathematical perfection—but building systems with intelligent real-time adaptability. And I'll introduce you to the Minimal Interval Variance Simulator, a free tool to help you understand and model unavoidable variance - even if you think you've got nailed your demand forecast and fully optimized your schedules!

The Quest for the Perfect Model: Are We Chasing the Wrong Goal?

In workforce management (WFM), there's an almost relentless pursuit of precision—better forecasts, optimized schedules, and models that promise to eliminate inefficiencies. The thinking goes: If we can just refine the model enough, we can remove uncertainty and create the perfect staffing plan.

But the reality is far more complex. Even if we could achieve 100% forecasting accuracy and flawless schedule alignment, fundamental challenges remain:

• Agent behavior variability – Human adherence is never perfect. Breaks run over, unexpected absences occur, and agents make real-time decisions that shift availability.
• Unpredictable customer behavior – Even with robust forecasting, customer demand never arrives in a smooth, predictable pattern. There will always be unexpected spikes and dips.
• Real-world disruptions – External factors like system outages, weather events, and last-minute marketing campaigns can instantly invalidate a carefully planned schedule.
• The mathematical limits of variance – Even in a perfectly modeled system, variance in both supply (agent availability) and demand (call arrivals) cannot be eliminated. Demand variance is where I demonstrate limits at the lowest level, examining minimal interval variance.

This leads to a critical realization: The problem isn't just forecasting accuracy and perfectly optimized schedules—it's the inescapable reality of variance. The question isn't how we eliminate uncertainty, but how we design systems that can intelligently respond to it in real time.

AI vs. Quantum Computing: Which Path Forward?

AI and quantum computing are often framed as competing solutions to complex optimization problems. But which approach actually makes more sense for contact centers?

Google DeepMind's CEO Demis Hassabis, recent Nobel Prize winner in Chemistry, has suggested that AI may eliminate the need for quantum computing in many real-world applications. Why? Because AI's ability to approximate solutions efficiently by learning underlying patterns often removes the need for brute-force quantum solutions.

In a fascinating interview at the AI for Science forum, Hassabis elaborates on this "controversial take" about classical versus quantum computing. He suggests that our conventional understanding of what classical computers can achieve might be too limited. Watch as he explains how techniques used in AlphaGo and AlphaFold demonstrate that classical computers, when used correctly, can solve problems previously thought to require quantum computing:

This brings us to a key insight: Perfect planning might be overrated when real-time flexibility and adaptation could deliver greater business value.

The Quantum Appeal: When Scheduling Becomes Mathematically Intractable

Before we dismiss quantum computing's appeal, let's understand what we mean by "mathematically intractable." Imagine trying to find the best route for a delivery truck that has to visit hundreds of addresses. Seems simple enough, right? But the number of possible routes grows incredibly fast as you add more stops — so fast that even the fastest computers can't check them all in a reasonable amount of time to guarantee they've found the absolute best route. That's the kind of challenge we face with agent scheduling in contact centers. It's like our delivery truck problem, but on a massive scale. In computational theory, these types of problems are classified as "NP-hard," meaning that finding the absolute optimal solution requires examining an astronomical number of possibilities.

Agent scheduling in a contact center exemplifies this intractability challenge:

Every agent has:

• Different skill sets (Lets assume 5+ options)
• Different shift preferences (3+ options)
• Different availability constraints (4+ options)
• Different historical performance patterns (varying by channel, time of day, etc.)

When you calculate the possibilities, each agent could have 180+ possible configurations. Scale this across 500 agents, and you're looking at approximately 10^1127 possible schedule combinations—a number that vastly exceeds the atoms in the observable universe (estimated at 10^80).

To put this in perspective: if we built a classical supercomputer that could check one trillion schedule combinations per second, it would still need approximately 10^1107 years to evaluate all possibilities – when the universe is only about 10^10 years old. This is what computer scientists mean when they say a problem is "intractable" – it's not just difficult; it's fundamentally beyond the capabilities of classical computing approaches that rely on checking each possibility.

This astronomical complexity is precisely why quantum computing seems appealing. Quantum computers use quantum bits (qubits) that can exist in multiple states simultaneously, theoretically allowing them to explore vast solution spaces in parallel — like checking all those delivery routes at the same time. This could potentially revolutionize how we solve optimization problems. However, quantum computing is still in its early stages, and it's unclear when or if it will be practical for contact center scheduling. In the meantime, as DeepMind CEO Demis Hassabis points out, we have powerful tools at our disposal right now. Instead of waiting for quantum breakthroughs, our approach should operate on two parallel fronts:

Approach #1: AI-Powered Approximation Over Quantum Brute Force

As Hassabis notes, "Classical computers are capable of a lot more than we previously thought." His core insight, backed by AlphaFold and AlphaGo's success, is that AI can find near-optimal solutions by learning the underlying structure of a problem space rather than brute-forcing through all possibilities.

For contact centers, this translates to modern constraint satisfaction solvers and optimization algorithms that can deliver 95-98% optimal solutions in seconds rather than waiting years for a theoretically perfect solution.

Approach #2: Real-Time Adaptive Systems

Even the best AI-optimized schedule is only perfect for a moment. Rather than a single perfect plan, we need systems that can continuously reoptimize based on real-time conditions and respond to emerging patterns.

This approach recognizes that adaptability to variance is more valuable than trying to eliminate it entirely.

Understanding Minimal Interval Variance with Our Free Simulator

To help contact center leaders understand the unavoidable mathematical variance in their operations, I've built a free Minimal Interval Variance Simulator that visualizes this concept in action:

As shown in the interactive dashboard pictured above, Minimal Interval Variance (MV) isn't a forecasting error or planning failure - it's a mathematical certainty derived from queueing theory:

• For this example center with 1,900 total daily calls: The peak interval shows 189 forecasted calls Even with perfect forecasting, actual volumes will naturally vary by ±5.8% This means the interval could receive as few as 178 calls or as many as 200 calls Lower volume periods experience even higher variance (±7.0% on average)

The Math Behind the Variance

The simulator applies the formula:

MIV = √(2/(π × forecast volume))

This equation demonstrates why:

• Higher call volumes have lower percentage variance (peak periods show ±5.8%)
• Lower call volumes have higher percentage variance (making precise staffing more challenging)

What Makes This Simulator Valuable

What makes this simulation tool particularly valuable is that it isolates pure mathematical variance from external volatility factors. Even with:

• No marketing campaigns
• No weather events
• No technical issues
• Perfect agent adherence

... our volume that arrives on an interval level still experiences this unavoidable statistical variance. Try the simulator yourself to understand exactly how much unavoidable variance exists in your operation: Minimal Interval Variance Simulator

The Two-Pronged Approach to Modern Contact Center Management

Given these mathematical realities, contact centers should adopt a two-pronged strategy:

1. Optimize the starting point - Use AI to create the best possible initial forecast and schedule, accounting for known minimal variance
2. Build adaptive systems - Develop real-time response mechanisms that can adjust to both predictable minimal variance and unpredictable volatility

Lessons from Eisenhower: Planning vs. Plans

"Plans are nothing, but planning is everything." – Dwight D. Eisenhower

This quote encapsulates a critical truth about workforce management. Over-investing in a perfect plan leads to diminishing returns. Think of it this way: instead of spending weeks meticulously crafting a rigid schedule that will likely be outdated by the first coffee break, what if you focused on building tools that empower your managers to adapt in real-time? Imagine them effortlessly reallocating agents based on live queue data, dynamically adjusting breaks and training sessions as needed, and using AI-driven predictive models to anticipate upcoming demand. This shift in focus — from static plans to dynamic adaptability — is at the heart of modern contact center success.

Contact centers have historically spent months refining forecasts, optimizing schedules, and adjusting staffing models—only to see their plans collapse under the weight of real-world volatility.

Why Traditional Workforce Planning Fails:

• The more complex a plan, the more fragile it becomes. Small deviations can render a carefully optimized schedule useless.
• Execution matters more than prediction. A well-run operation isn't the one with the best forecast—it's the one that responds fastest to unexpected changes.
• Planning for flexibility beats planning for precision. The organizations that succeed create systems that adjust in real time.

Final Takeaway: Real-Time Adaptability Beats Perfect Planning

Instead of spending more time refining forecasts, leading contact centers are investing in live optimization tools that allow them to:

• Reallocate agents based on live queue health
• Adjust shrinkage dynamically instead of pre-scheduling every break & training session
• Use AI-driven predictive response modeling instead of relying on static schedules

Understanding your operation's minimal interval variance through my simulator is the first step toward building a more resilient, adaptive contact center operation.

About the Author

Ted Lango is a seasoned operations executive and founder of Kyōdō Solutions. With over two decades of experience in contact center transformation, Ted focuses on balancing technology, workforce optimization, and human-centered design. Whether you're just getting started with AI or refining complex service models, Ted's expertise helps organizations adapt, innovate, and excel in today's ever-changing landscape.