Understanding API’s and AI — Adding an intelligent layer to the “super glue of the internet”

Alethea AI Official Announcements
Published in
7 min readMay 28, 2019


TL;DR — Adding a layer of Artificial Intelligence to API-sourced data layers can lead to explosive productivity growth

“Productivity is not everything, but in the long run, it’s almost everything” — Paul Krugman

In the digital world, API-first strategies are essential. With APIs one can exchange and share data seamlessly across different applications. For example, the weather app on your phone may be calling weather data from OpenWeatherMap, and this blog is shared on our Facebook, Twitter, LinkedIn, and Instagram via API-technologies as well.

NOTE: An API — which stands for Application Programing Interface — represents a set of defined methods of communication among various components that represent building blocks that can be used to build software. It simplifies programming by abstracting the underlying implementation and only exposing objects or actions the developer needs.

Essentially, an API is a business asset that creates value for a company by making carefully chosen data available under a specific access policy that is in line with the value offering to the user and value capturing of the provider. Companies such as Zapier have created tremendous value with these APIs, as they enable you to connect different popular apps with each other in order to automate routine tasks.

Automating routine tasks can give many companies a performance boost. But what about more strategic tasks? What could happen if we add an intelligent layer to routine data exchanges? How could we impact the workflows of the future?

Affectionately referred to as the “super-glue of the internet” — Zapier allows you to connect your apps to automate workflows in order to create a so-called “Zap”, which is a workflow recipe consisting of a specific predefined path. A Zap shares data from one app to another app, which helps to create faster processes and to automate the routine tasks. Anyone can build Zaps with just a few clicks, it is a DIY automation tool without the need for developers to hop in and help the end-user.

Zapier connects with 1500+ apps, such as Twitter, Slack, various Google products, WordPress, and has recently added integration with one of our favorite tools. As Quartz’ Khe Hy states: “The most important productivity app is the one that stitches them all together”. More than 3 million people are currently using Zapier to automate their routines.

Beyond Zapier — An AI-powered Zapier

When we think about the future of work, we usually think about automation, robotics, and artificial intelligence. Automation can happen in various ways, such as automating routine tasks with software tools such as Zapier, or using robotics to execute certain repetitive processes. Artificial Intelligence is usually seen as a tool that can provide “beyond-human” capabilities in decision-making and complement human labor in the workplace. Increasing productivity is a result of all these technologies, although it remains to be seen how effective these productivity gains will be.

Let’s take the following scenario as an example — Imagine a simple robotics system is trained perfectly to separate bad apples from good apples in a production line. (it looks at the color hue, “squishiness of the apple”, number of wormholes, etc as inputs). Let’s also assume that this imaginary robotics system can perform this task for all apples in the world in just one hour. This is great productivity, but perhaps it is not intelligent productivity.

Why not?

We would be optimizing for the wrong outcome if we don’t understand the problem-set effectively within the various constraints and contexts. For example, do we want the machine to run at certain times at night instead of during the day so we could get a discount on energy costs? Do we need a production line model or can we imagine a new system where robots are directed to the trees to pluck them freshly for the daily required supply? And do we then, instead of classifying all apples that are not ripe as “the bad apples”, should we aim to classify apples based on their degree of ripeness or other factors so the robot could make the necessary calculations to determine the optimum time of plucking the apple or to preserve the ones that are unripe and repurpose them for a later harvest ? Could the wasted apples be used in a regenerative way to fertilize the soil or contribute to the next harvest? What outputs are we maximizing for? Are we designing for intelligently productive systems or productive but net-negative outcomes?

Ibby our Developer Relations lead likes to call this: Productivity Intelligence (PI).

Productivity Intelligence is the study of how multi-factor inputs can influence the real-time output generation of a productivity-enhancing artificial intelligence algorithm. By looking at the contribution of its performance to overall business productivity, multi-factor inputs can generate real-time outputs that are more purposeful, adaptive, agile, and intelligent. Internal systems are organized in a modular way, allowing for restructuring and reassigning importance weights to the components involved, depending on the multi-factor data inputs and real-time overall business productivity.

Consider the 2018 AI Index report, which reported that VC funding for U.S. AI startups increased by 350% from 2013 to 2017. According to Forbes Insights, 80% of 350 interviewed U.S. executives believe that placing powerful AI-fueled applications in the hands of knowledge workers is critical to productivity and performance. With every year more patent applications and global funding than the previous year, one could reasonably expect global productivity to have skyrocketed.

Regrettably, this has not happened. Financial times argues that digitization has failed to lift global productivity and to improve economic performance around the world. This indicates that innovation, technological advancement and our endless mantras around of productivity have not yet translated or impacted our global economic performance. This is counter-intuitive given that automated systems are by design, more productive, precise, and have higher output.

What is going on here?

“You can see the computer age everywhere but in the productivity statistics.” — Robert Salow

In the 70s and 80s, the computing capacity increased a hundredfold, whilst the labor productivity growth slowed down by several percentage points. Various experts pontificated that a delayed productivity jump would come with a learning curve and an adjustment curve causing a mismeasurement of inputs and outputs.

On closer inspection, we note that purposefully implementing modular innovative tools and an intelligent, adaptive, agile approach to managing their efficiency and productivity optimization. Let’s unpack what this means for you and future developments within decentralized microservice architecture software such as SingularityNET.

The Modular Workflow — A case study of Zapier

Zapier allows you to create Zaps that are modular: you take a data output from App 1, and you input this into App 2 which will give you another output. An example of this is their most recent Medium integration.

You can continue adding Apps and data components to the Zap recipe as you’d prefer. For example, let’s say that for every new Google Doc in a folder named “SingularityNET Blogs” we automatically want to post a blog to Medium, and add our Medium posts to our social media, and from our social media to a slack channel called #social-updates-log so we can see all external communications collected in one channel, then this is possible. This is just one example of implementing a modular innovative tool with the help of Zapier.

We could even push this to the next level, by building on top of innovative tools created with Zapier by adding a purposeful and intelligent, adaptive, and agile approach to managing efficiency and optimizing performance. With SingularityNET’s decentralized microservice architecture, we are exposing APIs of various AI services that can be integrated in any software of your choice — including Zapier’s data flow automation tools.

Currently, Zapier does not have yet have multi-factor real-time data processing by AI algorithms that are geared towards Productive Intelligence. This is where SingularityNET can potentially be utilized. By using our decentralized microservice architecture with numerous AI-APIs, the automation of routine tasks as currently done by Zapier can enjoy an AI-layer, allowing for not only automating routine tasks, but also automating routine tasks in a more intelligent way — based on real-time data, with adaptive outputs, and optimizing for overall business performance and efficiency.

What’s Next

Together with Zapier, we are experimenting with developing a SingularityNET integration for Zapier that would add an AI layer to your Zaps — perhaps at a Meta-level your Zaps-can be Zapped by AI to create the optimum Ai-powered intelligent zaps.

This will make AI services accessible and usable to everyone — including those without a developer background. If you are a big fan of Zapier and have cool ideas for what you would like AI to do for your Zaps then Share it with us on the forum in our “Suggest a Project” category. We may be picking one of your ideas to work on for our first potential collaboration with SingularityNET x Zapier integration. We also invite developers from the community to express their interest under these topics if they would like to contribute to these projects, and we will make sure to contact you so we can collaborate together to make it happen.