This article was written by Troy Gedlinske (CDI’s Expert in Residence for CS Ops + AI) along with colleagues at CDI Services, a global leader in Conversational AI Auditing, Training, and Consulting services.
This is the fifth article in a program designed to help Customer Service leaders develop a realistic, effective AI strategy.
The first four articles can be accessed at these links:
Article #1: Four tips to help CS leaders up their game
Article #2: Managing CS Quality in the new era of AI
Article #3: Improving CS Finance in the new era of AI
Article #4: Achieving CS Scale in the new era of AI
Each article is a summary of a more detailed, hands-on workshop designed exclusively for CS leaders at the VP level and above. To learn more about CDI’s suite of workshops offerings, you can contact Ben Taylor ([email protected]), CDI’s Global Head of Sales.
Executive Summary
- Flexibility is about planning for the moment when your strategy falls apart. And it will, because even the best, data-based strategies can’t predict the future.
- For CS Ops teams, CS Flexibility is about responding to unexpected variance in forecasted demand. Whether up or down, next week or two months from now, this type of unexpected demand (usually event driven) can’t be predicted, but it can be planned for.
- Historically, CS Ops teams have achieved CS Flexibility with a playbook mapping pre-defined response options to pre-defined scenarios. This is a complex but critical approach that continues to be important in the new era of AI.
- AI-powered assistants reduce the pain from unexpected variance by making it easier for you to introduce and expand self-service support options. Unlike human agents, self service options don’t have capacity constraints that can be overwhelmed by unexpected demand.
- AI-powered chatbots also allow for a new class of response options for your playbook, creating a reserve of support you can call upon to protect your human agent—and your SLAs.
CS Flexibility is about planning for the moment when your strategy falls apart
We all know that no strategy survives contact with reality. And yet will all still depend on strategy to help us prepare for the future—and manage our CS operations.
Why?
First, because the act of developing a strategy is valuable on its own. It forces us to ask and answer tricky questions, question our cognitive biases, and choose priorities.
Second, because it helps us learn. When our strategy runs afoul of reality, we can evaluate where we were right or wrong, and why, and then incorporate lessons learned into our next strategy.
As CS leaders, such real-world learnings help us achieve CS Scale, a critical long-term goal that we discussed in the previous article in this series. They also help us work towards another critical goal: achieving CS Flexibility.
Flexibility is one of the five pillars of CS. It’s a measure of your ability to weather periods when both your forecast for CS demand and the expected variance from it is off the mark—often due to impossible-to-predict events.
Achieving CS Flexibility—in effect, planning for the moment when your plan falls apart—is complex, but not impossible. And with the help of AI-powered assistants, you have new options for dealing with unexpected variance in demand when you experience it.
Reminder: scale and flexibility are not the same thing
CS Scale and Flexibility are similar in several key ways. Both are critical components of a successful CS operation. Both rely heavily on forecasting. Both are achieved using similar tactics, such as adopting new technology or recruiting new agents.
The big difference between CS Scale and CS Flexibility is duration. When planning for Scale, CS Ops teams are responding to longer, more predictable changes in demand. In contrast, with CS Flexibility, you’re responding to swift changes in demand, often unexpected.
Although it varies by company and industry, a trick for distinguishing between “short” and “long” duration is to look at your “street-to-seat” duration. This is a measure of how long it takes to train a new human agent to handle direct customer contacts without active supervision.
For many companies, the “street-to-seat” duration is 3-4 months. So according to this simple measure, any investment or program to improve demand response that takes longer than 3-4 months is CS Scale. Everything else is CS Flexibility.
Quantifying the unexpected, from forecast to variance
Like many things CS Ops teams do, achieving CS Flexibility begins with forecasting.
Whether forecasts are based on years of data or simply common sense, they’re all trying to predict expected demand for CS. CS Ops teams then build their strategies to match expected demand with the required supply of human, physical, and technical resources.
While the forecasts that CS Ops teams present to company leadership are often high level with smooth trend lines, the ones they rely on for day-to-day operations are granular, sometimes with increments as short as 15 to 30 minutes. Their forecasts also include expected variance from the forecast, that is: predictions for when the forecast could be wrong, and by how much.
The good news is that a lot of expected variance comes from known sources, such as holidays. Because these sources of variance are known in advance, CS Ops teams can plan for them.
The real trick is responding to unexpected variance. This could be actual demand greater than the expected variance, such as a new product introduction that went much better or worse than anticipated. This could also be actual demand above or below forecast in response to an event that wasn’t forecasted at all, such as an extreme weather event or product recall.
No matter the cause, this type of variance is both unexpected and impossible to predict. This is also why strategy often falls apart, and why CS Flexibility is so critical.
Achieving CS Flexibility, from variance to response
When developing a CS Flexibility strategy, CS Ops teams begin by 1) identifying the time horizons within which a change in demand may occur and 2) identifying the response options available for each time horizon.
For example, for very short time horizons of less than two weeks, your responses options range from the very bad, such as ditching SLAs, to the challenging and/or expensive, such as mandatory overtime. For time horizons between two weeks to four months, you have more options, and fewer tradeoffs.
Once you have your matrix of time horizons and available response options, you have to create a playbook that maps the optimal response option to the appropriate time horizon. This is an extremely important planning task, for two reasons:
- When faced with unexpected variance, your team has to be able to act fast. For extreme events, they may even have to act before leadership has the chance to provide guidance or official approval.
- Flexibility responses always involve some sort of tradeoff that must be informed by your company’s priorities. For example, if your company prioritizes CS Finance, then your playbook may call for ramping labor down in response to slower demand more readily than ramping it up in response to higher demand. Then again, if your company prioritizes CS Quality, the opposite may be true.
Obviously, there are many, many other factors involved in establishing a playbook. But the key point is that a playbook helps your team make smart, fast decisions when unexpected changes in demand pressure your plans—as they almost certainly will.
How AI can help you achieve CS Flexibility
As you start thinking about your AI strategy with respect to CS Flexibility, we strongly encourage you to remember that preventing a problem is often easier than fixing it. In this case, the prevention you should focus on first is self service via AI-powered assistant.
Of course, no type of self service option can help you prevent unexpected variance. But an AI-powered assistant can help you limit the pain by containing as many customer support interactions as possible. After all, an assistant doesn’t care if the number of customers reporting a service outage surges from 10 per hour to ten thousand (assuming, of course, that the technology can handle such a load spike). That’s certainly not the case for a team of human agents.
AI-powered assistants can also unlock new and powerful responses for your response playbook.
For example, the actions taken by a well-designed AI-powered assistant depends upon its confidence recognizing customer intentions. Above a pre-defined minimum confidence threshold, the assistant responds; below that threshold, the assistant kicks the interaction to a human (usually after a few attempts to increase its confidence). When facing an unexpected variance in demand, you can reduce the minimum confidence threshold for some (or even all) intents. This would immediately increase the number of interactions the assistant will attempt to manage on its own, relieving pressure on human agents.
Similarly, you can respond to unexpected demand by allowing an AI-powered assistant to handle interactions that would otherwise be routed to a human agent. This isn’t without cost, since you’d need to train the assistant on these otherwise human-only use cases. Nor is it without risk, since there are usually good reasons why some customer interactions are classified as human-only. But in an emergency, redirecting even a small percentage of demand away from overwhelmed human agents may justify both the cost and the risk.
If you would like to be notified of additional releases in the series along with white paper content being developed from the series in 2025, you can sign up for direct notifications here.
