The Low-Code/No-Code Movement Builds in Speech Technology
Writing code is a tedious process with a number of time-consuming manual steps. Beyond the basic authoring tasks, the process involves moving code from one environment to another, testing, checking for bugs, and more. This means a lengthy time between when the concept is created and the actual application is ready for deployment, which no longer works in today’s fast-moving environment.
Today, much of the testing, checking for bugs, and moving to other platforms and environments can be automated, but even more significant in the world of application development is the meteoric rise of low-code/no-code platforms. Low-code/no-code platforms are suddenly showing up in a variety of industries, including speech technology, enabling much faster development of new solutions or changes to existing ones with easy-to-use building blocks. This has democratized application development, taking it out of the exclusive domain of the IT department.
"Low-code technology is geared toward citizen developers who do not have good knowledge of software development but understand what business logic they need," says Preetpal Singh, senior vice president of intelligent automation at Persistent Systems, a company that has offered no-code/low-code solutions for 15 years.
Beyond the ease of use, another reason for the growth of no-code/low-code solutions in speech and other technologies is improved acceptance by corporate executives, according to Singh.
“They used to have some serious concerns on the security side,” he explains. “They were also very worried if these platforms would have the capability to address all of their customer experience needs.”
Scalability was another initial concern.
But all of those issues have been addressed in the past couple of years, Singh says. “I think this market is going to grow upward in the upcoming year because all these no-code/low-code platforms are investing very heavily to get adopted from a large-scale perspective among large enterprises.”
From an efficiency and cost containment perspective, these platforms are popular because they allow service managers or frontline workers, not necessarily the technology experts, to quickly build, deploy, and alter business technologies, including speech, says Rebecca Wettemann, CEO and founder of Valoir. “For voice interactions, this is really important.”
“Developing CX or conversational AI solutions used to represent extremely costly, lengthy, and human-resource-intensive projects,” adds Fredrik Larsson, chief technology officer of text-to-speech systems provider ReadSpeaker. “Today, when companies need to be in a position to pivot quickly and cost-effectively to meet the changing needs of their customers, fortunately low-code/no-code software development is now an option, and one that is rapidly becoming adopted.”
Low-code/no-code development platforms now allow ReadSpeaker customer companies to pick and choose from pre-assembled application components, which have ReadSpeaker’s Neural TTS voices integrated within them, giving them the flexibility and ease of use they need to create mobile or web apps without having to create ad hoc lines of code, Larsson says.
Low-Code More Common for Speech Technology
Though no-code is the ultimate ideal, for speech-related applications, low-code is more typically the norm. No-code platforms are purely driven by visual point-and-click or drag-and-drop interfaces, with the user needing to do nothing more than name components, according to Josef Novak, cofounder and chief solutions architect of Spitch, a Swiss vendor of conversational artificial intelligence, automatic speech recognition, voice user interface, and natural language processing solutions. With low-code solutions, users might need to configure some more complex pieces of the infrastructure.
“The world of low-code/no-code is designed to make everything fairly simple and easy for a nonprofessional coder to manage,” according to Andrew Davis, senior director of methodology at Copado, provider of a development platform built for Salesforce. “The problem is, as soon as you start working through a development life cycle like this, it gets a lot more complicated; a lot more things can go wrong. You’ve got to begin to think about other technologies to use, like version control, to ease that whole process.”
Low-code arose as companies realized that Salesforce and other out-of-the box solutions with many drag-and-drop elements and a user interface for easy customization couldn’t accommodate all situations without some coding, according to Davis.
“Low-code means that you can do a whole lot of stuff without any code. But you can also supplement or augment that with code, if need be,” Davis adds. The more augmenting with code, the more a user might need the involvement of DevOps, which is becoming more common in low-code development.
“Every single customization you make to Salesforce [or similar solutions] adds to the complexity and adds to the number of things that can break as you’re making changes,” Davis says. “As you add customizations, it’s easy to break things. That’s where DevOps comes in.”
Salesforce and other technology solution providers offer no-code/low-code solutions because it increases the size of their potential customer base, Novak explains. “In the context of natural language understanding applications, what we’re seeing is the opportunity for people who are closer to the business, those who understand the language of the business, being able to configure these applications. That’s good for the customers and good for the business.”
Larsson agrees, pointing out that the availability of no-code/low-code means that companies such as ReadSpeaker can more efficiently offer their products to a wider user base by easily enabling language customization and other capabilities.
Less Data, More Power
Low-code/no-code also enables users to work with much smaller data files to build speech applications, according to Larsson. “If you require less data to build voices, then more voices can be built.”
Bringing low-code to the voice world lets users work quickly at scale on speech technology and adapt to changes in customer expectations about interactions, whether they’re event-driven, competition-driven, or trend-driven, Wettemann says.
“It’s an area that has evolved rapidly. What we’ve seen in particular is the capabilities—what you’re able to accomplish with no-code/low-code for text and voice interactions—has really improved,” she says.
For speech technologies, low-code/no-code primarily uses third-party libraries to leverage capabilities like speech-to-text for the developer community and the user community, Singh says.
Some providers of speech-to-text, conversational AI, and other voice-related technologies provide capabilities that can be leveraged by end users to manipulate the applications to meet their individual needs, he explains.
“For ease of use, no-code/low-code applications should be visually driven,” Novak says. “It may require a little bit of configuration, like setting parameters, IP addresses, names, variables, and identifiers, but there shouldn’t be anything complex. It shouldn’t be something that requires the application developer to be writing curly brackets or worrying about the spacing and large blocks of code.”
These solutions will often offer the user interaction via a web browser, Novak adds. “That’s the case for us.”
Novak adds: “In the context of what we build at Spitch, what focuses on [low-code] development is primarily in our dialogue composer application. This application is similar to what you would build with Google DialogFlow. It builds call flows.”
Nonprogrammers can interact with Spitch’s call flow building application via web page to configure call flows for a wide variety of use cases, Novak explains.
The application is focused on a combination of nodes, each with its own specific function, Novak continues. “Each of those nodes is described graphically, and with some kind of documentation, they are designed to perform actions within the call flow, like playback for an audio file or asking the user for voice input, maybe to perform a selection or branch in the dialogue based on a specific input that’s provided back by the user.”
A recent trend in low-code application development, according to Novak, is enabling them to integrate more complex machine learning models. This enables the user to define intent targets and then ask users how they feel (“great”), with the machine learning associating the intent with the feeling.
In the past, you’d have to train models on several examples that would need to be defined. Machine learning makes this process much more efficient.
Wettemann likened the evolution of low-code/no-code in speech technology to the evolution of speech technology itself over the past few years. With voice recognition, for example, the ability to manage more complex tasks and understand more context has dramatically improved, as have the capabilities of what can be done with low-code platforms for chatbot development.
Now that an increasing amount of speech technology is being delivered in the cloud, there is much more ongoing, incremental improvement to technology that can go to all customers than when on-premises installation was more common, Wettemann adds. “Low-code is super-important when we think about the chatbot space, because the last thing I want with a chatbot is rigidity. I want to be able to spin up new responses, new workflows, new ways to interact with customers, based on what my competition is doing.”
Low-code technology enables users to quickly respond to market needs by easily tweaking an existing chatbot or building a new one, according to Wettemann. “If I have a critical issue with customers, and I’m looking to a developer to create a new chatbot response, I have already lost the war. I want to be able to drag and drop, quickly build and deploy for business users who actually understand what the customer issues are and can do that in a timely manner.”
For example, a new issue could arise resulting in 90 percent of certain calls going to an agent because existing chatbots aren’t equipped to handle them, frustrating both customers and agents, Wettemann explains. “If I have a service manager, business leader, or business analyst who can identify a trend or something that’s happening and modify or spin up a new chatbot or IVR queue [via low-code technology] to handle a specific issue, then not only have I reduced my incoming call volume, I’ve increased my customer satisfaction.”
There are many tools that enable nontechnicians to build chatbots in a low-code environment, but the main limitation is the level of customization, particularly when it comes to more complex workflows and integrations to more data sources, Wettemann maintains. The more customization, the less that low-code technology will fit the bill and the more that developers will be needed, she says. “I can get pretty sophisticated today with low-code tools that reduce development time by 60 percent to 80 percent. I’m seeing business users who have some tech savviness but are not coders by any means being able to do 80 percent to 90 percent of what they want to do with the low-code tools.”
However, Singh says the technology still needs to evolve further to provide the support much of the developer community needs.
There are increasing numbers of companies offering low-code/no-code, from major, well-known firms to little-known and low-capital startups and everything in between.
“When we talk about low-code, Salesforce was doing that from the beginning, and that’s true with their voicebot and chatbot technology,” Wettemann says. “A major marketing point that Salesforce used was that customers could deploy the software out of the box.”
“Salesforce is obviously an extremely versatile platform, and they’re trying to become the central hub for all things business,” Davis agrees. “They’ve got the ability to integrate with partners like speech technology partners, and then they have the ability to allow you to build your own customizations on the Salesforce platform. The same is true for other out-of-the-box solutions.”
ServiceNow is another large player in this market, and like the other companies, it keeps expanding the ability of its customers to customize its solutions through low-code/no-code, according to Davis.
Microsoft has made huge investments in this area as well, focusing on NLP-related capabilities, Wettemann adds. “And there are tons of interesting, smaller niche players, like Solvvy, which was bought by Zoom.”
There are also companies building speech solutions for specific verticals, such as financial services and healthcare, that enable users or third-party developers to customize their applications using low-code tools, according to Singh.
As low-code/no-code in speech technology continues to evolve, it won’t be thought of as a separate category within a year or so, Wettemann predicts. “Expectations have been raised. In the old days, I expected that I wasn’t going to be able to get the answer from an integrated, interactive voice but that eventually it was going to get me, best case, to the right person who could answer my question or solve my problem. Today, I’m expecting to be able to interact via chat, to be able to interact via voice if I wish, and to get that resolution.”
Companies offering solutions designed to incorporate low-code solutions have already started using artificial intelligence and machine learning to help their developers define best practices for building the desired capabilities, according to Dadiala. AI/ML can review programming to inform developers where they should change code to make end solutions better.
Novak expects the trend toward better and simpler interactions with high-level machine learning components to continue. Companies will also continue to add more features that require less technological knowledge and more capabilities, like automatic language detection.
Davis expects the largest companies, like Salesforce, SAP, Microsoft, and Amazon, to continue to grow in terms of their low-code/no-code offerings in speech and other technologies. “They don’t want to give any ground, but it’s an expanding universe. There are new companies emerging, so Salesforce [and other large tech firms] have a lot of competitors offering low-code platforms.”
Wettemann expects the number of low-code/no-code platforms available to expand quickly and such technologies to become commonplace soon. “It will just be expected that [coding] is something that service managers and business leaders can do. So we won’t be talking about the concept of building a chatbot with code anymore. It will just be expected that people are building these and spinning them up in a low-code environment.”
Phillip Britt is a freelance writer based in the Chicago area. He can be reached at firstname.lastname@example.org.