Stop Scrolling Your Feed:How I Built a 3-Step AI Scout That Reads YouTube for My Business

My YouTube "Liked videos" list had turned into a 508-video graveyard. Today I built a three-step AI scout that studies what I have already saved, filters new uploads against my current business goals, writes a one-page brief in Google's free NotebookLM tool, and queues only the top three picks to a private playlist.

Why bookmarking videos for Saturday stopped working

Like many business owners, I used the YouTube thumbs-up button as a bookmark. I would spot a 40-minute interview on pricing, client acquisition, or AI tools while waiting in line for coffee, tap Like, and tell myself I would watch it on Saturday.

Saturday never had time. Between building custom websites and AI assistants for my service-business clients by day and taking the night shift with our baby boy, the list only grew. By the time I audited my account, 508 videos were sitting in that list untouched.

Yesterday I built the first half of the fix. Three times a day, at 6 am, 2 pm, and 10 pm, a background helper checks my new likes, pulls the full transcript through Google's free NotebookLM tool, and writes a one-page business brief mapped to my current goals.

Instead of sitting through a 40-minute video to find two minutes of useful advice, I could open my reading page and get the exact steps in two minutes.

That solved the watching problem. Today I realized I still had a scrolling problem.

Auditing 508 saved videos to find my real taste profile

Even when I used the Like button as a two-second bookmark, I still had to open YouTube and scroll through clickbait and algorithm noise to find videos worth saving. So tonight I asked my AI assistant to take over the scouting too.

First, I had it analyze all 508 videos I had already liked, plus the 102 non-Short, non-music long-form videos I saved over the last 90 days.

When you look at all 508 saves at once instead of one by one, the pattern is narrow. Almost every video I actually cared about came from 16 creators and five business topics:

• AI coding and multi-agent workflows that run real business tasks reliably without constant babysitting.
• Agency client acquisition and pricing, including how small consulting shops land and deliver for service-business clients.
• AI media, product photography, and visual tools that reduce production costs for online brands.
• Content packaging and hooks that earn attention on YouTube and LinkedIn without hype.
• Local service-business operations, such as voice receptionists, lead follow-up, and workflow automation.

Everything outside those 16 channels and five pillars - generic news roundups, 60-second Shorts, music mixes, and pure hype demos - was noise I never went back to read.

Writing that taste profile down into a single file gave my assistant a clear scorecard. It now knew which creators I trust, which topics connect to our client work and active projects, and which uploads to ignore on sight.

Turning that taste profile into an automatic scout and playlist

Second, we turned that written taste profile into an automatic daily scout. Instead of waiting for me to scroll my feed and tap Like, my assistant now checks new long-form uploads from those 16 favorite creator channels and my personalized recommendations on its own.

Every candidate video gets scored against my taste profile and my active 60-day business goals. Anything under three minutes gets dropped immediately so Shorts never clutter the queue.

In tonight's first live run, the scout scanned 31 new long-form uploads across those 16 channels and found 16 that cleared my quality bar. It took the top-scoring pick - a 13-minute head-to-head test of two coding models - pulled the 14,916-character transcript through NotebookLM, and added a one-page action brief to my reading dashboard.

Third, my assistant created a private YouTube playlist on my account called "AI Scout - Worth Watching" and automatically added the top three videos. Now I read the written brief with my morning coffee, and I only open the private playlist when a video has a screen walk-through I want to see with my own eyes.

How one brief immediately improved our client workflow

The value of a system like this is not how many transcripts it saves. The value is whether those briefs change how you run your business the same week.

One brief from earlier today broke down how a creator studied 27 of his best-performing videos to turn his editing and visual pacing style into a repeatable checklist. Instead of relying on guesswork every time he sat down to edit, he turned his own past winners into a written rule sheet that an AI helper can check.

We are already using that exact pattern to package workflows for our clients. That is the difference between passive watching and active scouting: a passive watch list makes you feel behind, while a filtered brief gives you one concrete improvement you can ship today.

How to build a simple version of this in 3 steps

If your saved links, Watch Later list, or browser bookmarks are piling up, you do not need custom code to use this same three-step pattern:

• Step 1: Audit what you actually saved. Copy the titles and channel names of your last 50 saved videos or articles, and ask an AI chat tool to group them into your top five business topics and name the noise you always skip.
• Step 2: Filter against your next 60 days of goals. Give the AI a one-page sheet of your current business targets, paste a video link or transcript into Google's free NotebookLM tool, and ask for the single change that helps your numbers this month.
• Step 3: Send only the winners to one private playlist. Cap your daily filter at three to five picks a day so your watch list stays a shortlist instead of a second job.

If you want my taste-profile template and NotebookLM setup, let me know and I will share them. How many unread videos are sitting in your Watch Later list right now?