Technology often follows a familiar progression. First, it’s used by a small core of scientists, then the user base expands to engineers who can navigate technical nuance and jargon until finally it’s made user-friendly enough that almost anyone can use it.

Right now, the process for building software is making that final leap. Just as the clickable icons of Windows and Mac OS replaced obscure DOS commands, new “no-code” platforms are replacing programming languages with simple drag and drop interfaces. The implications are huge: Where it used to require a team of engineers to build a piece of software, now users with a web browser and an idea have the power to bring that idea to life themselves. This means that powerful tech, which only large, well-resourced businesses have been able to afford, is suddenly within the reach of even small companies.

Perhaps most significantly, it’s making it possible to deploy artificial intelligence — one of the most transformative technologies in a generation — without hiring an army of expensive developers and data scientists. That means that smaller businesses, which often have huge amounts of data, can employ the benefits of AI such as powering new kinds of customer experiences (like a self-driving Tesla), growing companies’ top line (like P&G’s AI-driven advertising spend), and optimizing operations for maximum efficiency (like Walmart’s supply chain).

For smaller businesses, knowing where and how to deploy this tech can be daunting. Like larger companies, who may have already gone through the process of figuring out how data science might work for them, it makes sense to begin by deploying no-code AI on bite-sized tasks vs. ocean-boiling mega-projects. Ideally, you want to:

  • Work with the data you already have. There is often more value to be captured there than you may initially think.
  • Pick high-value tasks where being more efficient will drive growth.
  • Get quick wins in common areas, sales funnel optimization or churn reduction, so your team can learn how AI applies to a wide range of use cases
  • Don’t be afraid to move on quickly if you cannot achieve a 10x ROI from any AI project. There are plenty of high-return applications to get value.

No-code tools empower employees to think about creative ways to use data to drive or optimize their work — and consequently, the business.

Consider an example like intelligent lead scoring. Sales teams collect leads from all kinds of places – web scraping, cold calling, online forms, business cards dropped in a bowl at a trade show. But once a team has thousands of leads, the problem is deciding which ones to chase down. By spotting patterns in user behavior, demographics, and firmographics, a simple no-code classification model, for example, can rank leads according to their probability of turning into sales — a task many large firms use AI.

Using a no-code AI platform, a user can drag and drop a spreadsheet of data about sales prospects into the interface, make a few selections from a drop-down menu, click on a couple of buttons and the platforms will build a model and return a spreadsheet with leads sorted, from the hottest to the coldest, enabling salespeople to maximize revenue by focusing on the prospects that are most likely to buy.

The potential of AI is everywhere in the enterprise, and the advantage of no-code platforms is that they are not restricted to any particular use case. These tools can be used to detect machine maintenance patterns and predict which machines need attention before they fail, used by marketing teams to spot dissatisfaction and reduce churn, or by operations teams to reduce employee attrition. They can spot patterns in text, not just numbers, and be used to analyze sales notes and transcripts alongside sales history and marketing data, allowing companies to automate complex processes.

For many companies, working with no-code platforms will come down to simply finding the right project — and the right platform.

Where to Start with No-Code

A competent no-code platform needs three critical features.

First, it needs a simple interface that makes it easy to get data into the model training process. That means integrating with today’s popular business systems, such as customer relationship management systems like Salesforce, and spreadsheet software, such as Excel. If relevant data lives in multiple places, the platform should be able to merge it.

Once the data is uploaded, the platform needs to be able to automatically classify and correctly encode the data for the model training process — all with minimal input from the user. For example, the platform might identify columns in the data as categories, dates, or numbers and the user should check to see that the columns are labeled correctly.

Second, the platform needs to automate model selection and training — tasks that would normally be performed by data scientists. There are many machine-learning approaches and each works best on a specific type of problem. The platform should have a search mechanism to find the best model based on the data and the prediction required. The user should not need to know their way around regression or k-nearest neighbor algorithms. The platform should just deliver what works best.

Finally, it needs to be simple and easy to deploy with existing processes. A platform should be able to monitor model performance over time and retrain as the business environment shifts and new data becomes available.

How to Pick the Right No-Code Platform

Not all no-code AI platforms are made the same, and the right tool depends on a company’s business needs. Solutions range from just a few dollars a month to enterprise platforms that cost six figures a year.

Finding the right one for a particular company may require some trial and error. The good news is that the best platforms are open, which means that anyone can try them to see how they work. In other words, users can take the platforms for test drives on relevant tasks and see how they perform.

For example, users can compare the accuracy of various platforms based on their relative performance on public datasets, such as the Australian credit approval dataset where the goal is to classify credit card eligibility. With minimal effort, users can see how often each no-code AI platform is correct when it predicts an outcome in the validation set — a random selection of training data, typically 20%, that is held back and run against the model to measure performance.

But accuracy can sometimes be misleading. It’s also important to consider the number of false positives and false negatives in prediction results. This is particularly important for “imbalanced” datasets, where only a small number of cases, like credit card fraud or cancer, need to be detected within large amounts of data.

For example, if a model to predict credit card fraud said “no fraud” every time, it would have very high accuracy, but would be useless. A good no-code platform will score false positives and false negatives.

Users should also consider the time it takes to use these no-code platforms. One key metric is the time it takes the platforms to train their models. That can vary from minutes to hours, and if it takes hours, it won’t fit easily into a busy person’s day.

Training is not the only time consideration. For these platforms to be truly transformative in an organization, they must be so simple to use that non-technical people will adopt them into their workflows. Check the onboarding processes of various platforms. If it takes help from the IT department or even significant effort, the people in sales or accounting aren’t likely to bother.

For more companies to wield the power of AI in more applications across their business, the answer can’t be “create and hire more data scientists.” As little as one-quarter of 1% of the world knows how to code. Yet, as tech investor Marc Andreessen wrote presciently a decade ago, software is eating the world. There’s no doubt that no-code is the future.

Someday every part of every business will be AI optimized. The data is there today. The rate of progress and maturation of the platforms that let more and more people turn that data into AI-driven prediction and optimization machines will determine the speed at which it happens.

Removing friction from adoption will help unleash the power of AI across all industries and allow non-specialists to literally predict the future. In time, no-code AI platforms will be as ubiquitous as word-processing or spreadsheet software is today.