The implementation of CTI requires three components-telecommunications equipment, data processing equipment, and software resident on each that can allow for communication-one to the other. In most cases, the switch functions as the telecommunications equipment. Part of the hardware configuration is an input/output port that can support RS232 communications (like a modem), or LAN communications. This is not native to telecommunications switches and on some switches is not available. In some specialized applications, the telephone itself has an RS232 communication port that can be directly connected to a desktop PC. Either way, the telecommunications equipment must be configured to speak and listen like a computer.

The data processing equipment is typically a server, but can be a mainframe style host, or a single PC. What is unique and different for the telecommunications switch, RS232 and LAN connectivity is native to most data processing systems. You might notice that it is the telecommunications equipment that has been forced to evolve.

The last component is software. In most cases, software is the most costly and difficult-to-integrate component. The switch is limited in its overall computing power because its primary function is to manage call processing/dial tone/rings, accept digits, and provide some reporting. Most manufacturers have published very low level command sets that send status information regarding calls and accept commands like transfer, hang-up, and dial. These commands are unique to each switch and offer no standardization. The data processing equipment on the other end can not execute these commands directly, so customized software must translate the switch commands into executable instructions. This process is known as middleware. How does middleware work? Let's use the example of a contact center that must automatically display the record for a calling customer on the agent's screen when the call is first presented to the agent. We can assume for purposes of this example that there is technology in place to prompt for the account number and the caller provides it.

The switch sends to the middleware information regarding a new incoming call. It includes the time, date, line number, and possibly a unique call identifier. (Most switches use the line number as the unique call identifier, and in our example we will do the same). The middleware reads the message, opens a record in its memory, and waits for the next message with this line number. The switch finishes gathering the account information by prompting the caller and capturing the digits. The switch sends a second message that contains the time, date, line number and now the account number. The middleware recognizes this message as additional information and appends the account number information to the active record.

The switch puts the caller on hold as it waits for an agent to become available. When an agent becomes available, the switch transfers the call. Simultaneously, the switch outputs to the middleware another message that identifies the time, date, line number, and the agent position. The middleware maintains a lookup table that matches every agent telephone position to every computer port. Once the transfer message is received, the middleware looks up the agent position and matches it to the appropriate computer port. It issues a command to the computer system supporting the contact center to access the customer record for this account number and place this information on this agent's computer port. In most cases the record flashes on the agent's screen before the agent's initial opening is completed.