Managing speech projects for maximum efficiency.

So you're ready to use speech to enhance and personalize the customer experience--now what? To get started, it is critical to understand the elements of a speech project and how best to manage them.

[ILLUSTRATION OMITTED]

Upfront Considerations

As a starting point for the project, there are a several decisions to be made. Speech projects require a diverse set of skills including project management, voice user interface (VUI) design, application logic development, grammar development, telephony programming, system integration, usability testing and more. You'll first need to analyze your in-house skills and determine whether you need to bring in additional talent to augment your team, or if you want to have your solution developed by a company that specializes in speech solution development. If you decide to work with a development partner, look for an organization with a strong track record in your industry and experience deploying the latest open, standards-based speech solutions. If you choose to develop the solution in-house, consider using specialized development tools to augment your capabilities and lower overall development time.

Next, you should decide if a premise-based or hosted solution is right for you. Many contact centers today are choosing hosted solutions to lower upfront costs, gain greater flexibility and shorten deployment time.

Getting Started

Create a cross-functional team with representatives from the management team, business unit, contact center management, agent pool, sales, marketing, IT, telephony and development. Obviously, the business and contact center managers have a great deal of knowledge. Technical people will have good input on what type of data is available and how the information can best be used. Agents provide important information on customer attitudes and needs. A management team representative understands how the project can help meet corporate goals and objectives. With a cross-functional team where all involved parties have a voice, you're one step closer to success. It ensures that everyone is invested in the project from the start, making it more likely that each group will meet their individual deadlines and review materials in a timely fashion.

Upfront research is extremely important--conduct a thorough review of both your business processes and customer interactions. Listening to actual customer conversations can give you a fresh perspective on their needs. Look at existing IVR call data to better understand how customers are using your automated services and where they could be enhanced. Look for new automation opportunities and ways to streamline menus. Consider increasing contact center efficiency with speech-enabled call steering. Matching customer needs with a business process flow chart will further illuminate speech opportunities and how your solutions can be structured to best serve everyone.

This may seem like a daunting project, but the effort will pay off down the line with higher automation rates and superior customer experience. If you are working with a professional speech solution development organization, get them involved from the start and be forthcoming with information. They can show you how to conduct the discovery process and provide important insight into how speech has been successfully used on other projects.

With your discovery complete and your team assembled, you are ready to host the project's kickoff meeting. It is extremely important that all team members and selected vendors attend. During the meeting you will be defining the scope of the project, determining what services to offer, working out technical details and deciding on the persona of the solution. It is extremely important that the entire team be educated on all aspects of the project. Business executives do not always understand all the technical details and developers can create more effective solutions when they truly understand the organization's business model and objectives. Set clear goals at the beginning of the project (offering new services, offloading x percent of calls, lowering costs by y percent, etc.) and make sure they are agreed to by all. Create and communicate realistic timelines and ensure that everyone understands what they need to deliver and when.

[ILLUSTRATION OMITTED]

Many organizations fail to engage the right people and have the right conversations at the kickoff meeting. This results in project managers being shocked later in the project: management doesn't like the chosen persona, data cannot be accessed as specified, etc. To ensure success, continually check that everyone's expectations are in sync, provide major stakeholders with regular updates and get buy-in from all parties as you go. While this may seem obvious, it is easy to skip these steps while in the middle of a project.

Defining The Project

Now you are ready to create the functional specification for your speech solution. These specifications are a formal description of the application's intended capabilities. They provide a blueprint from which the developers will work. Address as many details as possible in this document to eliminate surprises along the way and ensure that your final solution functions as you envisioned it. A strong functional specification should also include:

Voice user interface specification. The voice user interface (VUI) specification defines how the caller will interact with the speech solution. It should contain a call flow diagram with as much detail as possible on prompts, grammars and error handling. It should define how and when live assistance will be introduced. It should also specify the level of speech needed for the project and include details on the solution's persona.

System requirements specification. This is a technical document that defines the network layout of speech servers and details how the speech solution will integrate with data sources including the input/output parameters.

Deployment plan. The deployment plan is extremely important in detailing how the solution will be tested, tuned and deployed.

Once the project has been carefully specified, the development work can begin. The elements that you will need to manage during the process include VUI design; call flow and application development; systems integration; prompt recording and audio production; application tuning and testing. Of all these elements, the voice user interface design is the most critical. Many speech solutions fail to make the grade with customers because best-practice principles were not followed in the design process. Make sure your solution is fast, friendly and easy to use. Look to see that the grammars carefully match prompts to generate higher recognition levels and create a better customer experience.

Be sure to use open, standards-based development methods and deployment platforms. This ensures portability and a smooth integration with existing contact center, IT infrastructure and business applications. Check for pre-built grammars and packaged applications that match your solution needs to speed development. Where possible, leverage existing application development efforts. Today's speech solutions are more flexible than older, legacy IVR systems. It is much easier to create a speech interface for existing Web and business applications than to completely re-write a solution. This approach also simplifies the work needed to upgrade and enhance your solutions.

Testing

Everyone understands the need for testing once the application is developed, but there are several things you can do before the developers get started to ensure success. For new self-service solutions, consider providing agents with the dialog scripts to determine caller acceptance of the "prompts" and gauge their responses. For complex, "say anything" speech solutions, Wizard of Oz (WoZ) testing is a great way to monitor responses to a variety of prompts with subtle differences. Here the VUI designer plays chosen prompts to the caller to simulate an automated call. By hearing the actual responses, the designer gains a better understanding of which wording and prompt combinations will work best.

Once the application is developed, be sure to run it through an extensive QA program. Assemble an internal test group with multiple dialects and accents to ensure that your solution accurately understands a wide range of input. Then create a pilot program that allows you to divert a small portion of calls to the test site to better understand the "tuning" required to perfect each solution. Be ready to adjust call flows, prompts, speech grammars, thresholds and other parameters as testing continues. Be ready to repeat this process even after your solution is launched to ensure the greatest usability over time. Many organizations try to make up for project delays by cutting the testing and tuning time, only to end up with more problems down the road.

Pre-launch training is important as well. Make sure that everyone is aware of the scope and goals of your speech project before it goes live and be certain to educate contact center personnel on how to handle confused or irate customers.

As always, plan now to ensure future success. Create and analyze customer satisfaction surveys to better understand what your customers need from your self-service offerings. Tweak prompts to shorten navigation as users become more sophisticated, enable barge-in to allow users to more quickly get to where they want to go. Continue to analyze the solution from an IT perspective to reduce latency and update grammars as needed. Look for new ways to expand your self-service offerings to give your customers more of what they want.

For years, call centers have heard about the benefits of using speech technologies to improve the customer experience and increase contact center efficiency. Implementing speech in the contact center is complex, yet far from impossible. Working with an experienced partner will take much of the uncertainty out of the process.

As with all new initiatives, planning is the key to success. Upfront research and a productive kick-off meeting with a strong, diverse team enable you to fully flesh out how a speech solution can best serve your customers. Many companies are quick to jump into development and start working on design elements before the scope of the application is fully defined. Avoid this temptation, it will save you a lot of time and rework headaches in the long run. Another temptation to avoid is skimping on testing. Organizations that try to make up time by shortening the testing and/or tuning phase will end up with an inferior solution. The tuning phase is critical to creating speech solutions that are both engaging and have high recognition rates. As with all aspects of your customer care operation, ongoing evaluation and evolution of your speech solutions is necessary and will ensure superior customer interactions.

A great speech solution tells customers that you value their business and helps to reinforce your corporate image and branding efforts. By following these guidelines, your resulting speech solution will boost customer loyalty, increase contact center productivity and lower operating costs.

Christoph Mosing is Vice President of Professional Services at Envox Worldwide (www.envox.com), a provider of voice solutions. The company's software products and related services, including Envox OnDemand, reduce the time, cost and complexity of creating voice solutions and empower enterprises and service providers to reduce operating costs, improve customer satisfaction and retention and generate new revenue streams.

By Christoph Mosing

Envox Worldwide