The 3 waves of AI and the significance of AI as a service
Artur Grishkevich, Head of Growth6 minute read
You can’t go a day without hearing of another SaaS raising millions. Everything as a service is a hot commodity right now. At the same time, the barrier to entry to data science, AI and ML - both personal, at the developer level, and at the business level is high. High enough to make Forbes write about it in 2019… and to then write about AI as a Service in 2020.
Read on for the impact of artificial intelligence as a service on businesses, opportunities for individual developers’ growth and, our favorite, conversational AI as a service.
What is the significance of artificial intelligence for business and where does AI as a service fit in?
For better or for worse, AI has been on the tip of every business leader’s and software engineer’s tongue for the last couple of years. With reason.
I think a great illustration of the spectrum of views on AI is the viewpoint of a good friend of mine who leads one of Europe’s most successful self-driving car initiatives. While he works with neural networks on a regular basis, he likes to say that the term AI is a little too much for a little too little. “It’s just algorithms,” he says. I agree with him to an extent. At the same time, decomposed to their lowest levels, most of our own behaviors come down to pre-programmed algorithms which adjust for situations and self-learn from experience. In that sense, the software algorithm imitates intelligence and can indeed be called an artificial intelligence.
After all, Alan Turing (whom we dearly love) believed that any machine capable of exquisitely complex calculations can be said to be a thinking machine.
What makes today’s AI different from the algorithms that came before is the impact the AI applications can actually have on the business. Processing data with a model versus processing data with a human should significantly reduce the company’s costs.
Should but does not always. As a matter of fact, last year Boston Consulting Group found that only 10% of enterprise companies saw a significant return on their investment in AI. The biggest blocker it identified to success was the dichotomy between engineering and operations. 70% of a project’s success is in getting the processes and training the people using the tools the right way. 30% is designing and engineering the algorithms.
This is where AI as a Service can make the most waves. Instruments letting any developer (and with time a line of business citizen developer) use artificial intelligence without having to know actual data science and machine learning will be empowering to the nth degree to solve business problems right where they reside.
The three waves of business AI
Shoot me if I’m taking on too much here but the way I see it, and in the context of artificial intelligence usage in organizations, there are three waves of AI.
In-house or third party consulting data-science teams using low level AI and ML instruments
This is the approach that the majority of enterprises working with AI are using today. It has its benefits in terms of extreme ownership of the IP within the organization. It has its significant downsides as well - hiring, high cost, long cycles. When it comes to consulting, all standard outsourcing risks apply.
AI as a Product
Some of the most successful SaaS startups today are using AI as a part of their product offering. For example, Gong.io or Chorus.ai offer their customers a load of features around analyzing sales conversations to glean valuable insights and make their sales floor more successful. That is AI as a Product, not AI as a Service. While offering lots of benefits, its main cons are the narrow scope of impact. I do expect that with time more and more products will roll out offering AI benefits in a simple to transact software product fit for a specific use case. In theory, a small business will be able to build a stack of such products to solve for whenever it can use AI. The downside is that, according to BCG, due to the differing nature of organization structures, AI projects generally would not replicate well from company to company.
AI as a Service and artificial intelligence platform as a service
AI as a Service refers to instruments for developers or citizen developers that put all the power of AI in the hand of the user without requiring them to know anything of or have any experience with actual machine learning and data science. Some companies in this space are Siemens with their Mindsphere and UIPath with their AI-enabled self serve RPA. There are some very exciting AI as a service startups building under the radar. Dasha AI is also a Conversational AI Platform as a Service.
This approach offers the most flexibility and, with no-code or low-code interfaces, enables small business owners or line of business leaders to solve for their problems directly. AI as a Service and data science as a service will be enabling new generations of business leaders to become citizen developers in their own right.
How to make Dasha’s AI as a Service platform a part of your developer toolbox
Let’s go over the Dasha architecture first.
Dasha Studio is the extension to Visual Studio Code. You use the studio to create your conversational script. Aside from the code interface, there is a visual graph editor which you will find handy. For a quick getting started guide - take a look at this post for a tutorial on building a conversational AI app from scratch and fill out the form to get your Dasha API key.
Dasha SDK lets you make Dasha a native part of any applications you are building. It runs on your side and enables you to launch conversations, process data and store the results of the conversation. If you are security-conscious, there is a parameter that, when checked, will keep the Dasha Cloud Platform from saving any of your conversation data. To use this you have to set up your own storage for conversation results.
Dasha Cloud Platform is, in essence, where you get the AI-as-a-Service functionality. You connect through the SDK to its API to make the AI have the conversation. It provides your automated conversations with such basic conversational features as Natural Language Processing, Speech to Text, Text to Speech and advanced features like interruptions, turn taking, sentiment analysis, named entity recognition, text classification, active hearing, slot-filling, dictation and common sense logic.
To put it in a nutshell - in your AI conversational script you program the flow of the dialogue using Dasha script. As part of this, you set up custom intents, which are used to train neural networks (AI as a service). You also specify when you want interruptions, when you do not (AI as a service). You provide the data that is used to train the neural networks to understand the specific named entities you expect to encounter in the course of the conversation (AI as a service). But you, personally, don’t have to train the neural networks, you don’t have to maintain the cloud instances of NVIDIA V100s running and you don’t have to agonize over the confidence calibration problem in machine learning, like our data scientists do. We’ve got a whole blog section dedicated to our R&D (and it’s the most read section of our blog).
What you have to do is boot up your Visual Studio Code and start creating apps and making calls. You can then set up the SDK, connect it to our conversational AI API to make Dasha an integral part of your application. How you use it is up to you. You can build a brand new product from scratch that automatically qualifies inbound leads. Or you can give your iOS app a human-like conversational interface. Or you can teach your SaaS to take customer service calls on its own.
It’s a pretty cool time to be in tech.