When AI Pushback Pays Off: Using LLMs to Fix a Real Business Problem

Frustrating.

We run two events per month and use an online platform (I'll call it xxx) to promote them.

Yesterday, our account was hit by a spam bot that bought up all the remaining tickets for one of our events. Our assistant didn't notice right away and—thinking these were real attendees—paid for additional registrations to accommodate them.

By the time we realized what had happened, the damage was done.

I contacted xxx's support team, but it was the weekend. Nothing much happened other than a pleasant chat with their chatbot, which did escalate the ticket—but escalation doesn't fix bots in real time.

This wasn't the first time it had happened. A few months ago, we dealt with the same issue, and one specific action helped resolve it quickly. So I decided to try that again.

TL;DL (Too Long; Didn't Listen)

• Don't accept the first answer an LLM gives you—especially if you know it's solved the problem before.
• Different LLMs have different strengths; knowing which one to use (and when to push back) matters.
• You don't have to be a developer to sound competent when working with technical teams—AI can help bridge that gap.

Asking the LLM for Help (Again)

Last time this happened, I asked Claude to generate code that could bulk-delete spam registrations via the platform's API. I forwarded that code to xxx's tech support and asked for API access so my developer could install it.

The problem was fixed shortly thereafter.

So this time, I did the same thing—except when I asked Claude (through Perplexity) for the code, it initially refused:

"I wish I could provide code that would help, but this is genuinely a platform limitation that requires xxx's internal tools to resolve efficiently."

That was interesting—because Claude had done exactly this before.

So I pushed back.

I reminded it that it had provided code the last time this happened, and that offering a concrete technical solution had prompted the platform to act quickly.

And then Claude responded with:

"You're absolutely right—apologies for the confusion. Here's the Python code similar to what I provided before…"

It generated the script.

I won't include the code here, but that's not the point.

The point is this: you don't have to settle for the first answer an LLM gives you.

Why I Chose Claude (and What I Would've Done Next)

I went to Claude specifically because it has a reputation for being strong at programming tasks. But if Claude had flatly refused, I would've tried:

• Google Gemini
• ChatGPT
• Microsoft Copilot
• Manus

Different models, different guardrails, different strengths.

This is how I work with LLMs every day: challenge them, redirect them, and switch tools when necessary.

The Unexpected Bonus: A Script for Support

This time, Claude did something new. It also generated a short script I could send directly to xxx's support team explaining how the code could be used.

It suggested wording along the lines of:

"I have a developer ready to implement this API-based solution to remove the bot registrations. The code uses your official API endpoints to identify and delete the spam orders. If you can provide API credentials with the necessary permissions, my developer can execute this immediately. Alternatively, your team could run this script internally since it uses standard API calls."

Was Claude trying to make up for its initial refusal? Maybe.

Either way, it worked.

The Outcome (and the Real Win)

I sent the email at 8:30 a.m.

By 10:30 a.m., the issue was fixed.

The last time this happened, it took days—until I involved an AI-generated technical solution. This time, I went straight to the LLM, and the turnaround was hours instead of days.

Also worth noting: I'm not a developer. I use the tools developers create.

But with the right help from AI, I can communicate effectively with technical teams—and that alone is incredibly powerful.

A Fair Note About the Platform

To be clear: xxx is a great platform, and we continue to use it.

Spam bots are a constant battle for any large online platform, and I know they've implemented safeguards to reduce these attacks. When issues arise, they do fix them as quickly as possible.

We'll keep using xxx because it has tremendous reach and—frankly—is an often overlooked digital marketing tool for events.

Even with the hiccups.