Forget low-code. AI does it better.

In this article you will learn how Generative AI can replace or augment low-code

• Enable non-technical users to live out their API development dreams
• Customize Generative AI to your business API's

The eternal Product struggle

If I have learned one thing as a product leader in the world of #startups:

You can't please everyone.

Customers come to you with a myriad of requests ranging from features, problems and desires. Often times the best answer you can give is simply:

You could use the API

People hate this answer - because it means it's a low priority request and they have to figure it out themselves.

APIs are unfortunately shrouded in technical complexity. For non-technical users, this directive is a roadblock, slowing processes, creating inefficiencies and increasing end-user complexity.

APIs are annoying. So low-code was born.

Users without coding experience rely on developers to make API integrations.

In response to user desires to interact with APIs with limited technical skills, the low-code movement emerged like Zapier , and Spot 's Fylamynt.

Low-code aims to democratize software development by reducing the technical skills required. While this is a step in the right direction, low-code still requires users to learn a new solution, and to stitch together chains of events in a workflow (lets not even talk about Data Security concerns).

Let loose Generative AI + APIs

Instead of learning a new low-code product, what if your end-users could simply talk to your API documentation and build custom scripts?

Employing large language models (LLMs), we can tap into API documentation and the practical use context of APIs to dynamically generate integration code.

This approach is turbocharged with the OpenAPI specification, allowing users to simply describe what they need, and the system generates a custom API script for them.

Non-technical users will simply be able to...

1. Ask a question "I want to bulk upload a CSV containing <data>"
2. Generative AI scans your API documentation for this use case
3. AI creates a custom script for the user to execute, solving the problem.

How to: Integrate Generative AI + Custom APIs

Here's the simplified process

1. ETL your OpenAPI spec
2. Ask an LLM (in my case, Anthropic ) to create a business use case that an end user would ask, for each API (IMPORTANT!)
3. Append the use case to your API Schema
4. Upload your Schema to a Vector database (in my case, Weaviate )

Generative AI Results

In my solution, I'm using Spot 's OpenAPI spec (found here)

Their product, Elastigroup, allows you to save money on Amazon Web Services (AWS) EC2 instances, by leveraging Spot Instances.

Prompt: "list and delete elastigroups"

In seconds a custom script is created for the end-user. Now they can live out their dreams of building custom solutions and features on top of the Spot API.

Practical applications of this technology

• A sales manager seeking specific data from a CRM platform.
• A marketing executive wants to extract data from their email campaign software for an upcoming presentation.
• An HR manager needs to consolidate employee data from various HR tools for reporting.
• A business analyst wants to pull financial data from a SaaS platform for a quarterly review.

Remove bottlenecks and improve low-code

This novel approach allows business's to dramatically boost customer satisfaction and reduce the technical burden on your end-users.

Additionally, this same solution can bring substantial cost savings to development teams. Giving them an interface to quickly write integrations between systems.

We should aim not just to simplify the user experience, but to truly revolutionize it. Forget about low-code platforms; with the speed and intuitiveness of #generativeai and #openapi, businesses can overcome the API barrier.