I've Tested Almost Every AI Agent Builder on the Market

Over the last year I’ve built agents on nearly every no code and low code platform out there. Not demos. Actual agents meant to do real work.

Some of them are still running in production today. Most of them aren’t.

Instead of keeping these lessons to myself, I thought I’d write out what I’ve learned. The good, the bad, and the stuff the marketing pages will never show you.

Note: This isn’t me bashing the tools. Some are genuinely great. But there’s a gap between what these platforms promise and what they actually deliver, and I want to share what I’ve seen so you don’t have to learn it the hard way.

The “Anyone Can Build” Problem

Every platform says the same thing. Type what you want. We’ll build the agent for you. No code required. Anyone can do it.

And look, that’s technically true. You can type a prompt and get something back in minutes.

Here’s what they don’t tell you.

That agent is generic. It sounds like every other chatbot. It handles the happy path and falls apart the moment something unexpected comes up.

I’ve done this dozens of times. Build an agent in a day. Feel great about it. Show it to the team.

Then the real work starts.

You need it to match your brand’s tone. You need it to handle the weird questions your customers actually ask. You need it to not sound like a robot reading a script.

And suddenly “anyone can build” becomes “anyone can build something mediocre.”

The platforms lower the floor. That’s real. But they don’t raise the ceiling. Building agents that genuinely work still takes thought, iteration, and honestly a lot of trial and error.

I’ve watched teams spin up agents and abandon them within weeks because the gap between “working demo” and “actually useful” was bigger than they expected.

The “Enterprise Grade” Problem

This one frustrates me the most.

You’re evaluating tools for your business. You see “enterprise” on the pricing page. You see the logos. You think okay, this is serious.

Then you start asking questions.

Does this support custom SSO? Not yet, but it’s on the roadmap.

SCIM provisioning? We’re working on it.

RBAC? Sort of, but probably not the way your IT team needs it.

Integration with your existing identity provider? Let me check with engineering.

I’ve had these conversations more times than I can count. And I’ve started telling everyone I work with: don’t trust the label. Ask the specific questions.

Here’s my checklist before I take any “enterprise” AI tool seriously:

• Custom SSO support (not just Google and Microsoft)
• SCIM for user provisioning
• Proper RBAC with granular permissions
• Data residency options
• Audit logs
• Integration with existing identity providers

If they can’t answer yes to most of these, they’re not enterprise grade. They’re enterprise priced.

The gap between the marketing page and the technical reality can be enormous. Better to find out early than after you’ve already started rolling it out.

The “Shiny New Tool” Problem

A new platform launches. The demo looks incredible. Twitter is buzzing. Your team wants to try it. Leadership asks why you’re not using it yet.

So you deploy. Agents everywhere. Summarising emails. Drafting responses. Automating the small stuff.

Six months later you look up and realise nothing has actually changed.

The agents are running. But they’re not moving the needle. They’re doing busywork, not the work that actually matters.

This is the trap I see constantly.

People see a tool and work backwards to a use case. They ask “what can we automate” instead of “what problem are we actually solving.”

The result is a lot of AI activity with very little AI impact.

Here’s how I think about it now:

Before deploying any agent, I ask one question. If this works perfectly, what changes in six months?

If I can’t answer that clearly, I don’t build it. It’s just a demo dressed up as a solution.

What I Actually Look For Now

After testing all these platforms, here’s what I’ve learned matters most:

1. Refinement, not just creation. How easy is it to iterate? Can I fine tune the tone, adjust the logic, handle edge cases without rebuilding from scratch? The first version is never the final version.

2. Foundations, not just features. Does it integrate with the systems I already have? Can my IT and security teams actually approve this? Shiny features mean nothing if it can’t fit into the existing stack.

3. Clarity on the problem, not excitement about the tool. The best agent I’ve ever built wasn’t on the fanciest platform. It was built for a specific, well defined problem where we knew exactly what success looked like.

The Bottom Line

The tools are better than ever. That part is true.

But better tools don’t automatically mean better outcomes. They just mean faster paths to the same mistakes if you’re not careful.

If you’re building with AI right now, stay skeptical. Not cynical. Skeptical.

Ask what happens after the demo. Ask what “enterprise” actually includes. Ask whether you’re solving a real problem or just chasing hype.

The promise is easy to believe. The delivery is what counts.

To building things that actually work.