Low-code AI: Mission ready?

Low-code solutions offer a potentially faster and cheaper path to AI deployment without sacrificing effectiveness

By Bob Grabowski, Brandon Vaughn, Sean Ryan, Javier Becerril, Alan Litzel, Brian Diederich, Jennifer Burson, and Barak Martone

While applications for artificial intelligence (AI) exist in nearly every sector, AI technologies can be complicated and expensive and require highly skilled talent to develop and maintain. In the past, creating an AI solution required a team of experienced data scientists and IT infrastructure engineers to develop solutions. Today, many technology vendors offer low-code AI software suitable for a wide variety of mission and business applications. Low-code development offers a potentially faster and cheaper path to AI deployment without sacrificing effectiveness.

Low-code AI software can be configured and deployed with minimal coding and often employs drag-and-drop interfaces, prebuilt components, and other tools that make it easier for people who are not data scientists to implement.

While there are some limitations to low-code approaches—decreased flexibility in AI techniques available, for one—the trade-off may be worth it when tackling common use cases or when available pretrained models can rapidly accelerate the speed of deployment. Some advantages of using low-code AI:

  • Using a more straightforward low-code AI approach helps to leverage the skills of the current workforce.
  • The vendor’s prebuilt models may pair well with an existing automated process (with Robotic Process Automation) or workflow (with Customer Relationship Management or Enterprise Resource Planning systems).
  • Leveraging low-code AI modules nested within approved vendor platforms should mean a reduced IT governance approval burden.

There are countless use cases suitable for low-code AI, so it often takes a structured analysis to determine if a use case is appropriate. A good rule of thumb is that if multiple vendors offer prebuilt AI models for your use case, you can go with a low-code solution. While vendor performance claims should be taken with a grain of salt, use cases such as invoice processing, receipt processing, and text translation have been successfully solved numerous times by many vendors. So low-code development and prebuilt models may be the fastest path to capturing value when available.

At a high-level, low-code AI may be best suited for common use cases (such as sentiment analysis, image recognition, and text classification), when there is a lack of training data (prebuilt models are often trained on large, diverse datasets), or when you are trying to rapidly prototype or test new ideas. It is also important to note that while low-code AI can be a good choice in many situations, it may not offer the flexibility needed to tackle highly complex or unique use cases.

Is low-code AI “mission ready”? In a word, yes. In near real time, low-code vendor AI models are getting better and better, meaning passable results today will transform into high-performing results in just a few vendor release cycles. This means public-sector clients should find a business process that could benefit from a low-code AI model today, choose a vendor that complements its enterprise architecture, and go for it.

Low-code AI: Mission ready?

Get in touch

Bob Grabowski
Managing Director
Deloitte Consulting LLP

Brandon Vaughan
Managing Director
Deloitte Consulting LLP


Sean Ryan
Senior Manager
Deloitte Consulting LLP

Javier Becerril
Specialist Leader
Deloitte Consulting LLP



Did you find this useful?