How coding became unnecessary for agent building

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

What if “trying a bunch of tools and going in random directions until something works” is completing a task in “a very human way?” Asking for a friend.

In today’s issue…

• How no-code conquered agent building

• 🎙️ An exclusive interview by Building AI Agents

• Build AI apps using AI with Claude

• The reality of being a freelance agent builder

• Salesforce’s race to one billion agents

…and more

🔍 SPOTLIGHT

In 2025, AI agents aren’t just be created by coders—they’re being created by everyone.

In the early days of AI agents in 2023 and 2024, knowing how to program was essential to working in the field. The first agents like AutoGPT were complex tangles of Python code—non-technical workers could run them with a short series of pre-written command line steps, but if you didn’t know how to code, modifying them or building your own was a non-starter.

Few companies have enough software engineers to work hand-in-hand with every employee across sales, marketing, data analytics, legal, and all the innumerable other functions that would benefit from automation. Even when software engineers are available, we usually lack the subject matter expertise to tell whether the agents are performing well. I can build you a sales agent that can read all the emails exchanged with a potential lead, do deep research on their company, and craft response messages, but I don’t have the sales experience to tell if the agent is working well.

Is “hey idiot, buy our stuff or we’re blocking you” an effective sales pitch? How should I know? I’m not a salesperson.

AI agents were not the first type of software to have this problem, however. Robotic process automation (RPA) was a hot field—some might say a fad—in the late 2010s with a very similar promise to agentic AI: automating repetitive white-collar tasks using pre-coded workflows. Unlike agents, however, RPA lacked the intelligence and flexibility given by large language models (LLMs), and easily broke down in the face of even slight variations in its inputs or procedures.

But RPA did one thing that would one day make agents accessible to the masses: popularized low-code and no-code tools. Unlike programming languages, low-code/no-code platforms allowed non-technical users to piece together their own RPA workflows using icons on a screen, visually building their automations as in a video game.

The true breakthrough came when these no-code building tools began to be applied to LLM-powered workflows like agents. Now, non-technical users could build automations that were much more powerful and flexible than RPA and use them to speed up their work. Enterprise agent platforms like Microsoft Copilot, Salesforce’s Agentforce, AWS Bedrock Agents, and Google’s Vertex AI Agents (to name a few of many) have brought no-code agent building to large enterprises, while others like n8n, Flowise, and Langflow have done the same for individuals and smaller businesses.

Now, AI agents are in the hands of anyone who can master one of these platforms. Sure, the most advanced enterprise-scale agents may still need to be built by software engineers, but many simpler ones—like the sales agent I mentioned above—can be built by the people who are in the best position to evaluate them: their users. Increasingly, agent-building isn’t just a profession—it’s part of everyone’s profession.

🎙️ EXCLUSIVE INTERVIEW

Tomorrow night, Tuesday, July 1, I’ll be interviewing Vishal Sachdev, a professor at the University of Illinois at Urbana-Champaign Gies College of Business whose work focuses on real-world applications of AI agents.

We’ll discuss…

• How aspiring agent builders can get started in the field

• Agent use-cases that are delivering ROI today

• How freelancers can market their agent-building skills to businesses

…and more

This exclusive event is only for members of Building AI Agents’ course and community, so be sure to check it out and sign up!