I Built an AI Video Review System Using Gemini 2.5 Pro

A financial advisory firm managing over $500M came to me with a straightforward problem: their senior partner was spending 18 hours every week watching advisor videos and giving feedback. Client-facing recordings, market updates, educational content — the kind of videos that either build trust with wealthy clients or quietly destroy it.

The feedback was inconsistent. Some advisors got detailed notes. Others got a quick "looks good" because the partner was drowning. Most videos never got reviewed at all.

In wealth management, a single client relationship can mean $50K to $500K+ in annual fees. When an advisor looks nervous on camera, reads from a script, or rushes through a tough market explanation, clients feel it. Maybe not consciously. But trust erodes, and eventually they pick up the phone when a competitor calls.

The firm knew this was costing them money. They just couldn't fix it with human effort alone.

So they asked me something I hadn't been asked before: can AI actually watch a video and give real, useful coaching feedback?

I said yes — but let me prove it first.

Most People Still Think AI Is Just a Text Tool

When most business owners hear "AI," they think chatbots and email drafts. That's fair — those are the most visible applications. I use AI for text-based work all the time in my DTC fashion brand in San Diego. Writing product descriptions, managing 313 blog articles, handling customer questions.

But video? Almost nobody is using AI to analyze video for business purposes. And video is everywhere. Sales demos. Training recordings. Compliance footage. Marketing content. The average mid-size company produces dozens of hours of video every month, and almost none of it gets reviewed because human review simply doesn't scale.

I built this system using Google's Gemini 2.5 Pro, which is one of the few AI models that actually understands video the way a human watches it. It doesn't just read a transcript — it tracks body language, speech rhythm, eye contact, pacing, and how all of those change over time.

It's not perfect. It struggles with subtle facial expressions and sometimes misreads a deliberate pause as nervousness. But for structured coaching against clear criteria, it's remarkably capable.

This is why I use different AI models for different jobs. One might be great at writing. Another is great at watching. I'm not loyal to any single tool. I'm loyal to results.

How the System Actually Works

Think of it like an assembly line for video coaching.

Step one: recording. Advisors open a link in their browser and hit record. No special equipment, no apps to install. The system automatically crops and frames their face so every video looks consistent — no more tiny figures in huge rooms or extreme close-ups. This matters because the AI needs to evaluate the person's performance, not their webcam setup.

Step two: multiple takes. Advisors hated the pressure of getting it perfect in one shot. So I built a system where they can record as many attempts as they want. The AI reviews all of them and recommends the best one, explaining why.

This single feature changed everything. Advisors went from dreading recording sessions to treating them like practice with a built-in coach.

Step three: AI analysis. The AI watches each video and scores it across 8 specific areas — eye contact, speech pacing, filler words (um, uh, like), body language, vocal confidence, content clarity, how well the video is organized, and how strongly it ends. Each area gets a 1-10 score with specific observations tied to exact moments in the video.

Getting these scoring criteria right took three full rounds of refinement. My first attempt produced the kind of useless feedback you'd find on a motivational poster: "Good energy! Consider working on your pacing." Not helpful.

The final version gives feedback like: "Between 1:42 and 2:15, your speech rate dropped significantly while explaining the fee structure — this may signal uncertainty to clients who are already sensitive about costs." That's coaching. Not cheerleading.

Step four: human review. This is the part most people miss. The AI's feedback goes to the senior partner first — never directly to the advisor. The partner reviews the AI's analysis, makes edits, adds context the AI can't know, and approves it before it gets delivered.

The AI is a smart assistant that does the heavy lifting. The human makes sure it's right.

The Results After 60 Days

The senior partner's review time dropped from 18 hours per week to 3.2 hours. That's an 82% reduction. At $400+ per hour, that's over $300,000 per year in recovered capacity.

Average advisor scores across all 8 areas improved 23%. Feedback is now consistent — every advisor gets the same depth of analysis every time, not just whoever the partner had energy left for.

One result stands out. An advisor who was about to be put on a performance improvement plan improved enough through the AI coaching that the plan was cancelled entirely. The firm estimates this kept a $2M book of business. That single save covered the entire cost of building the system.

And here's what surprised me most: after calling it "creepy" during the first week, 10 of 12 advisors now voluntarily record practice takes on their own time. They're using the AI as a personal coach without being asked.

I want to be honest — this is one system, early results, still being tuned. But the signal is clear. AI video analysis isn't theoretical. It's producing real business results today.

This Pattern Works Way Beyond Coaching

The same approach — AI watching video against clear standards — works anywhere video exists in business. Compliance teams could review recorded client meetings for regulatory issues. Sales leaders could analyze which reps actually close deals and what they do differently in the first 90 seconds. Training departments could evaluate whether their instructors are effective, not just whether they showed up.

Most businesses I talk to haven't even considered this as an option. They're focused on chatbots while sitting on hundreds of hours of unreviewed video. Finding these kinds of overlooked opportunities is exactly what I do as a Chief AI Officer. It's not about chasing the flashiest technology. It's about finding the $400-per-hour bottleneck that nobody else has noticed and building a system that eliminates it.

Thinking About AI for Your Business?

If anything here made you think about video — or any other process — sitting unreviewed in your company, let's talk. I do free 30-minute discovery calls where we look at your actual operations and figure out where AI could move the needle. Not in theory. In the next 90 days.

Book a Discovery Call