Answering and researching vendor calls can be a time-consuming and often difficult task for the AP department. Whether a vendor is asking ‘Did you receive my invoice?’ or ‘When can I expect payment?’ answering these questions is a necessary part of vendor relations, but it takes an AP employee away from processing payables.
There is an alternative. With Interactive Voice Response (IVR) and Interactive Web Response (IWR) systems, the above questions and many others can be answered without an AP staffer at all.
An IVR system can receive the call, accept the question, access the company's database and provide a caller with the information they are seeking. It can also take information from the caller, convert it to data and input that data to the database.
How IVR works
A typical IVR system uses a server that contains digital signal processing (DSP) hardware to analyze and reproduce voice patterns. The server can be separate from the main system, or a part of the company’s legacy system. For a small system, a desktop PC can be used.
The server interfaces with the phone system through a dedicated connection. The type of connection depends on the phone system’s physical connections. It can be through a key system, PBX, or any other type of phone switch through analog or digital ports, or even LAN or WAN connections.
Software controls the information flow between the caller and the appropriate server or system. Depending on the software, the interaction can be through touch-tone access or direct speech request. With the speech capability (see below), the vendor can simply say for instance “invoice receipt” (after having already identified themselves) and receive a voice answer as to whether the invoice has been received. The vendor can, in a similar way, find out when it is scheduled to be paid or if a check has already been cut. But a synthesized voice is not the only way in which the information can be provided. Many systems are capable of transmitting the information by fax, Web, or e-mail.
Not once in this process does an AP professional become involved, though IVR does have the capability to direct the caller to a person if their request is very complicated. The information that the caller provides can be sent directly to an agent screen, so the agent can see the data as they answer the call.
The key to this capability is the ability of an IVR system to access external databases. In most medium to large systems there are at least two servers—a telephony server and a host server, although, as noted above, a small system can run on a single server, including a PC.
Typically the telephony server actually functions as two servers—a call server and voice server, although both functions usually run on one CPU. The call server processes the incoming calls and directs them to the appropriate location or function. This may include Computer Telephony Integration (CTI) software, or may reside on a separate server. The voice server provides the voice processing and dual-tone multi-frequency (DTMF, i.e. “touch-tone”) input and output required. As noted before, stand-alone equipment is not necessarily required because both functions are software based.
For large systems however, stand-alone equipment that interfaces with the host system is often used.
A second server—the Host Server—generally provides connections to mainframe hosts where the databases reside.
Front and back
Many IVR systems utilize a Front End Application (FEA) and a Back End Application (BEA) in a client/server approach. The front end is the user interface—speech recognition (see below), IVR, IWR (see below) or faxback functions. The back end interacts with the host systems and databases and provides data to the front ends and manages the host connections. If there are multiple databases or hosts, there can be multiple back ends.
By the same token, where there are different ways of accessing the system, such as voice and Web, there can be multiple front ends.
In a simple example, assume there are just two applications—a voice front end and a single data back end. The front end would handle one phone call at a time, and the back end one host connection at a time. The FEA interacts with the BEA by making requests, which generally are specific to the application. In an AP application for example, the request might be “get balance owed” or “get last check sent.” The requests are not in the form of “go to screen XYZ and read lines n through n” or “go to ABC database and read PQR.” This insulates the FEA from the details of the BEA implementation, and the data sources used.
Turning speech into ones and zeros
Speech recognition is basic to any IVR system. This technology converts speech into a computer-recognizable format. The system then compares the digitized data with a speaker-dependent or speaker-independent database of expected values and returns a response.
The major advance that made IVR practical was development of speaker-independent capability. Ninety to ninety-nine percent of callers can be understood without having to repeat their requests.
But speaker-dependent functions still have a special use. Today’s speech recognition technology offers the capability of voice verification. This security feature makes it possible for a vendor to call in, their voice be recognized, and their transaction secured so only they can receive it.
Metamorphosis: IVR to IWR
IVR systems today are quickly morphing into IWR systems, particularly in large companies. The change sometimes is as simple as using the IVR system to access Web pages, as noted above. In such cases, the IVR system functions basically as a Web browser, with the Web page information being converted to speech. The caller asks questions and responds to prompts to receive the data.
More advanced IWR systems are designed to allow the vendor to access their data via Web pages with their computers. The vendor can search for an invoice receipt, payment information and other data just as it’s done on any Web page. This type of application actually integrates with Enterprise Resource Planning (ERP) systems (see the ERP information on the AP Network).
[Along similar lines but coming from a completely different technical direction are the newest integrated payables systems, like one by Xign, which are building in visibility for vendors via the Web. In Xign’s system, vendors can verify invoice receipt and see the scheduled payment date for an invoice via the Web.]
Representative IVR/IWR products
Representative of many IVR/IWR products is the Vista system offered by Syntellect (Phoenix, Ariz.). Vista is a Java-based application that runs in a client/server environment. It also features a VoiceXML browser that can obtain and interpret VoiceXML pages from a Web server. If the VoiceXML feature is not used, Vista accesses the data through a Host Server, as explained above. In VoiceXML applications, the Web server or an application server is utilized and the Host Server is not.
Vista provides two special features—text-to-speech and fax. The text-to-speech function permits the system to read text information to a caller. The text can be virtually anything, such as databases or host screen information.
The fax function enables vendors to receive a fax containing the desired information. Data accessed by the caller can be merged with forms, graphics or logos.
Another special feature is Voice Forms, which allows a caller to complete a form by speaking the answers to computer-generated questions.
Like Syntellect, Santa Clara, Calif.-based Avaya bases its Avaya Integrated Systems on a client/server environment. It too is Java-based with VoiceXML. It also offers text-to-speech capabilities, which are used to “read” e-mails from vendors. The system can then respond to the e-mails based on their content. Unique to this capability is the fact that the system has its own database and “remembers” other e-mails and transactions, so the response is based on the vendor history.
The Avaya system is a fully integrated, multimedia system that can access Web pages and use Internet protocol (IP) to function. Using IP, no in-house server is required. All functions are accessed via the Internet.
Also, some services like The Accounts Payable Network’s Supplier IQ are entirely web-based. Organizations send TAPN daily listings of all payables transactions, which are entered into a database that vendors can log in and search. Simple payment inquiries are handled on the internet with no need for the vendor to call AP.
One peripheral function of virtually all IVR systems is an outgoing function. This enables an AP department to automatically dial a vendor and give them a message that says their check has been mailed and when they should expect to receive it.
So what’s it going to cost?
The cost of an IVR system will depend on many variables, of course. But according to Syntellect, a system that would be adequate for very large companies, comprising four IVR ports, four IWR ports, two channels of TTS, with an ODBC back end database, would cost between $75,000 and $100,000. Such a system could handle 40 calls per hour. Another route would be to select a hosted system. In that case, coverage comparable to the above would require a set-up cost of approximately $30,000 and a monthly charge of $2,400. Thus the cost in year one is roughly equivalent to two FTEs, and ongoing cost under the cost of an average FTE associate level.
Avaya's costs are similar, though they express cost in terms of the number of calls to be handled. A relatively basic system handling 100 calls per day would run approximately $50,000; A large system capable of handling up to 1,000 calls per day would cost $150,000. Clearly there are economies of scale.
These cost figures coincide with the experience of Pfizer. For both IVR and IWR, Pfizer's cost was $150,000. But Pfizer found that systems can cost much more, particularly—in its experience—with the larger suppliers.
IVR and IWR systems provide an expeditious and valuable vendor relations tool.
