I've launched a few products on platforms like Brain and Udemy, but regularly checking sales figures is a pain. You have to log into multiple sites, navigate to each dashboard, and stare at the numbers. And when those numbers are zeroes, it's a quiet, soul-crushing experience.
I started wondering if I could offload this somewhat depressing routine task to AI. And if the results weren't great, could AI objectively brainstorm the next steps? This is my experiment.
The process was simple. I tasked a browser-operating AI with visiting the admin screens of multiple sales platforms and reporting the sales data.
But just reporting "zero" would leave me paralyzed. So, I took it a step further and added these instructions:
"If sales are zero, propose three possible causes and three actionable improvements I can implement immediately."
This way, I'd get not just the stark fact of "zero sales" but also an action plan for what to do next.
You are my business assistant.
Please perform the following steps:
1. Access the Brain admin screen and check the sales count for product "XXXX" over the last month.
2. Access the Udemy admin screen and check the sales count for course "YYYY" over the last month.
3. Compile and report the results.
4. If both sales are zero, identify one main probable cause and propose three immediately actionable improvements. For each improvement, mention the benefits and the estimated effort/cost.
After a short while, the AI's report arrived. As expected, reality was harsh.
"Reporting: Sales for both Brain and Udemy were 0 units within the specified period."
Yeah, figured. Where a human might sigh and get discouraged, the AI calmly moved on to the next task. The requested cause analysis and improvement proposals immediately followed.
[Cause Analysis]
[Proposed Improvements]
Enable Brain's Introduction Feature (Affiliate Program):
Issue Limited-Time Free Coupons on Udemy:
Share Product Use Cases on note.com and X:
These proposals were more specific and immediately actionable than I expected.
AI is just a proposer. I couldn't just blindly accept the suggestions. So, I threw a few more questions its way.
"Tell me more about the fees for Brain's introduction feature."
"Show me the steps to issue a coupon on Udemy."
The AI answered these questions as well, referencing each service's help pages and terms. I got all the necessary information, including specific fee calculation methods and exactly where to click in the admin screen to issue coupons.
With all this information, all that's left is for me to make a decision.
By having AI handle data collection, analysis, and solution brainstorming, I can focus on the most critical part: decision-making. I feel this is a highly effective way to move stalled projects forward.
If you've built a service or product that feels a bit stagnant lately, I recommend trying to consult AI.
The key is not to just ask, "How can I sell more?" but to phrase it like this: "Here's my current data (accesses, sales, etc.). Based on this data, tell me three probable causes and three specific improvements."
Basing it on objective data helps generate actionable plans, not just wishful thinking. When you want to consider your next steps without getting emotional, AI can be your best sounding board.
If you're interested in automating tedious tasks with AI to free up your time, I've also prepared a detailed guide on implementation.
It explains how to semi-automate information gathering and article generation using Python and Google services, complete with code. It covers setting up the environment from scratch, so it's for those who want to build the same system themselves.
I build and run small Python systems — trading bots, RAG APIs, scheduled automation — and write up whatever breaks along the way.
If a provider-agnostic RAG Q&A API is useful to you, mine is MIT-licensed on GitHub: rag-faq-api. It runs and passes its full test suite with no API key (offline stub LLM + hashing embedder), swaps to Claude / Gemini / OpenAI via one env var, and ships a retrieval-quality harness (Hit@k / MRR / Recall@k) with a chunking sweep.