Robots work hard, not humans! Really, why pay people if ASR speech recognition and TTS text-to-speech systems can do at least part of their work? You can’t erase words from a song about progress, but in real life, when replacing call center operators with a voice robot, there are important nuances associated with its own cost. Read the details below.
Do you need a voice robot? Visual calculation
First, let's do a theoretical checkerboard experiment. The input is like this:
- There is a call center that serves the information line without sales.
- The call first enters the interactive voice menu and, if it does not end there, is transferred to the waiting queue for operators.
- There is no 8-800 traffic.
- All calls are the same length, no after call processing.
- FCR=100%, subscribers receive the correct answer to all questions on the first attempt and do not call back again (in practice, of course, this does not happen, but we have a theoretical experiment)
Schematically: 
The company's management is going to replace part of the call center operators with a smart robot with speech recognition according to the SaaS scheme. It should turn out like this: 
General description of savings
The description of the savings at first glance looks simpler than a cucumber (for simplicity, we do not single out the installation and operating costs for IVR and ASR separately): 
If the left side of the inequality is greater than the right side, then it is profitable. Profit, victory in the hearts of millions. This is exactly what system vendors offer. But the devil, as usual, is in the details. Let's take any arbitrary period of time (numbers are conditional): IVR load, traffic hours – 1,000 (we know that in science the load is measured in special units – Erlangs, abbreviated as Erl, but for clarity we will use hours). Load on operators before system upgrade, hours – 3.000. Let's take IVR expenses =0. We will assume that the system has been in the company for a long time – completely depreciated, and the costs of its operation are negligible. Then, from the point of view of economics, only the gain compared to the cost of operators BEFORE the introduction of ASR will be of interest. Suppose that the processing of 1 hour of traffic by the operators costs 300 rubles. Expenses before modernization will amount to 3.000×300=900.000 rubles. Let only 300 hours of traffic reach the operators after the upgrade, the remaining 2,700 hours will be served by the robot: 
For SaaS systems
For SaaS systems, as in the condition of the problem, in 99% of cases, suppliers use the charging model for minutes or for requests to the platform, which is about the same, because each request takes some time. If an hour of traffic processed by ASR is cheaper than an hour of an operator, then the inequality seems to be true, and all is well. But the problem is that it is not known in advance how long the dialogues between the machine and subscribers will actually last: there is no certainty whether 2,700 hours will turn into 4,500 or 10,000 hours, which will block the expected benefit in price. Example: The price of processing an hour of traffic in ASR is 200 rubles per hour. The system talks with subscribers longer than a live operator, 1.5 times longer. 
Border situation, benefit is lost: 
Of course, this will not always be the case, we exaggerate a little, but we lead to the fact that this moment must be taken into account. The platform provider and integrator must clearly explain to the buyer how to protect themselves from such a risk, this is one of the criteria for their professionalism.
If you plan to buy a boxed version of ASR
If you plan to buy a boxed version of ASR with billing not for minutes, but for ports (simultaneous connections of subscribers to the system), then nothing fundamentally changes. You need to understand how much traffic will pass through it per unit of time, and how much this traffic will cost. The supplier will have to provide the buyer with a clear and transparent calculation. By the way, you can look at the situation from a different angle. In the example above, it was assumed that all calls last the same time. However, you need to understand what exactly is planned to be fed to the robot, because automation for the sake of automation is not cool. Example: 
What else needs to be considered? Back to the scheme
It is recommended to be very careful in calculating the economic effect of implementing ASR, especially when operating in highly competitive markets. Let's say that before the introduction of the robot, operators answered calls quickly: 
The supplier in the feasibility study of the project indicated that it would be possible to reduce 6 operators while maintaining the Service Level: 
At the same time, the new average response speed may turn out to be higher than the tolerance threshold of customers, that is, the time that they are ready to wait for a person's response. Let's say: Customers are willing to wait an average of 15 seconds. It was – 7.7 seconds. This suited them. It became – 26.1 sec. They are not prepared to wait that long. The company risks losing orders or reducing audience loyalty. On paper, there is optimization, but in fact it is imaginary or even harmful. It will not be possible to reduce all six operators, therefore, the assessment of the effect of the introduction of ASR must be approached carefully. By the way, if you add the presence of traffic 8-800 to the initial input, then with an increase in the average response speed, the cost of traffic will also increase. This must be taken into account when calculating the actual savings, because the cost of 8-800 is proportional to the square of the length of the waiting queue. The benefit condition becomes: 
In reality, the situation is even more complicated, because the "live" scheme looks like this: 
The FCR of the robot is less than 100%, while some of the subscribers will not be immediately redirected to the operators. They will call back. Even if ASR is smart and will send repeat customers directly to operators, this will reduce the expected economic effect.
Obtain a feasibility study from the supplier
When buying a robot, you need to obtain a detailed feasibility study from the supplier, which covers, at a minimum, the following issues:
- Will (and why) override the benefits of implementing a voice robot by the cost of processing contacts, due to the fact that calls can become longer?
- What share of traffic for which specific client issues is planned to be transferred to the robot?
- If the average response speed of operators behind the voice robot increases due to staff reductions, will this lead to a loss of customers if their tolerance is low and the market of the ASR buyer company is highly competitive?
- Does the project justification take into account the change in the cost of traffic 8-800?
- Are repeated calls of subscribers taken into account if the robot did not solve their issue on the first try, and they call back? How is the volume of such repeat traffic estimated?
Transparent professional answers to these questions will allow the buyer to make an informed decision on the purchase of a robot to replace call center operators and reasonably choose a competent supplier.
Dmitry Galkin, independent contact center management consultant
