A no-code platform is defined as a platform that allows users without a developer background to connect various tools and build apps through its visual interface. It’s a “shift in how users interface with software tools,” per TechCrunch. It’s a means of accessing GUI-based workflows that sit atop back end functionalities. Why do we care about “no-code” or “low-code”? Because it democratizes software creation by abstracting away the complexity of technical requirements. It gives us access to agile, low-risk innovation while removing the dependency on engineering teams that are already over-tasked.
On the software side, you’ve got everything from Unqork, Bubble and Webflow, to Zapier and AirTable, all of which a user can cobble together to build full-fledged apps or manage the software development cycle without knowing any programming languages. Airtable recently raised a $185M series D funding, Unqork secured a $207M in Series C, and Webflow nabbed a $140M Series B just a week ago, putting each of these companies at or over $2B in valuation. Forrester expects the market to surpass $21B in spending by 2022 and Gartner forecasts that low-code app platforms will make up more than 65% of all app development by 2024. This positions “no-code” and “low-code” to be more than fleeting buzzwords.
No-code machine learning as a key strategy
2020 changed a whole lot of things. Not the least of which is the organizational shift from marching toward a neatly planned roadmap and high-stakes horizon work to reliable bets and accelerated transformations that can prove results today, not a year from today. The question isn’t what new innovations organizations want to nurture along, but a question of what they can’t afford NOT to do. This new sense of urgency started to introduce the move to “dual-track transformation,” or organization’s ability to concurrently execute long-term transitions and real-time process innovations, which we wrote about at the end of last year.
As a result, no-code platforms are emerging as a key strategy for intelligence-powered automations. Just as no-code software allowed apps to be built without programming knowledge, no-code ML is opening up doors for machine learning integrations without data science domain expertise. With the no-code route, product and engineering teams can compress 10-12 months of ML vendor evaluation and development down to mere hours or minutes. In our case, a GUI-based point-and-click workflow (or a drag-and-drop, if you prefer) enables the following without a single line of code:
- Data prep and ingestion
- Queue management and error handling
- ML analysis on pre-trained APIs
- Indexing, storing, and search of the ML predictions
- Visualization of the data output without external tools
Why data science isn’t software development
The thing about AI and machine learning is that it seems like everyone is doing it. We talk a great deal about Skynet and singularity. But the reality of the matter is that AI/ML adoption is still reserved only for the most well-funded initiatives within enterprises that can afford to support years-long, high-risk horizon work that may or may not ever come to fruition. What’s more, 87% of data science projects fail.
Data science is just that—science.
Ironically—data science in of itself is not designed to be a highly predictable process. Traditional data science and machine learning development take a long time and require domain experts that are in short supply and are expensive to hire. Data science is not agile like software development, even when using off-the-shelf pre-trained models. That’s where no-code ML comes in. While it may not be a solution for highly custom, complex use cases, it goes a long way in solving existing challenges for over 85% of organizations who’ve yet to deploy AI/ML into production.
What challenges can no-code AI/ML solve?
- High AI/ML adoption barriers
- Single track innovations
- Tech stack complexity
- Talent shortage
- Risk and unpredictable nature of data science
How can organizations benefit from no-code AI/ML?
- It lowers the barrier to AI/ML adoption.
By reducing the time, resources, cost, and domain expert requirements, it gives projects that wouldn’t have otherwise gotten off the ground a chance to thrive.
- It enables AI/ML experimentation and rapid cycle development.
It gives teams the agility of software development traditionally not available in data science workflows, and the ability to better collaborate with their stakeholders.
- It enables interoperability of cloud tools and workflows.
An average org uses 1,200 cloud-based tools. No-code platforms provide a way to integrate ML automations without breaking existing systems and workflows.
- It allows teams to very quickly stand up ML projects and show results.
It’s a fine alternative to waiting months or a year to find out, “does it actually solve a real business problem?”
- It allows organizations to run multi-track transformations.
Due to its low cost and accessibility, it enable teams to parallel track multiple innovation projects instead of putting all their eggs in one, very expensive basket.
Where do we see no-code AI/ML headed?
For starters, no-code ML will help quickly uncover value in unstructured data and validate ML use cases for the business. We’ll see new ML pipelines created, broader ML applications, and experimentation, thanks to a lower barrier of entry. In the same way that we employ unsupervised learning to show us what we don’t yet know about a dataset, no-code solutions will point us to opportunities for workflow automations and otherwise untapped patterns that have the potential to generate new revenue streams.
Over time, we expect to see no-code solutions to be the answer to pre-defined business questions and deliver pointed outcomes with increasingly higher levels of confidence—not unlike supervised learning. And as the ecosystem of market-ready ML APIs continue to grow across computer vision, natural language processing, and predictive analysis, so will the adoption and role of no-code platforms within organizations of all sizes.