How to Use Opus 5 to PRINT $$$ With AI UGC

generation is cheap now.

video runs at cents per second, images run in parallel batches, and producing content is no longer the hard part of this business.

what's still hard is knowing what to make, who to make it for, and what to do with the attention once you have it.

those are judgment problems, and judgment is what the thinking layer does.

so the model writing your scripts is the same model that should be picking your angles, reading your data, and running your funnel, because those jobs are worth more than the scripts.

---

Load It or Waste It

a strong model does dramatically more with good inputs, and the gap between a loaded session and a cold one widens as models improve.

a loaded session carries your persona docs, your niche research, your banned phrasing, and your performance history.

the same prompts against a cold session produce smarter-sounding generic output, which is worse than useless because it looks like work.

so the standing context files are part of the money system, written once and appreciating every time the underlying model gets better.

---

Job 1: Find the Angles That Buyers Respond To

engagement research finds content people watch.

buying research finds content people act on, and those overlap less than you'd hope.

How to Run It

feed Opus 5 the raw material, hundreds of comments, reviews, threads, and complaints from the niche.

then ask for the clusters that matter commercially.

> "cluster this material into pain points, objections, and buying triggers. rank the pain points by how often people describe actually spending money to solve them, not just complaining about them. then pull the exact phrasing people use at the moment they decide to buy."

that last instruction is the whole difference.

the phrases people use when they're ready to spend are different from the phrases they use when they're venting, and hooks built from the buying language pull a different audience than hooks built from the complaining language.

Why It Prints

content built on a buying trigger attracts people already near a decision.

the same production effort reaches an audience worth more, and nothing about the pipeline changed except the research question you asked.

---

Job 2: Write the Capture, Not Just the Content

a piece with a great hook and a weak close produces views and nothing else, which is the most common failure in the entire space.

How to Run It

every script gets its close written as deliberately as its hook.

> "write the closing beat for this piece. it needs a 1-word comment keyword that's obvious and relevant, a question that invites a real answer about the viewer's situation, and a specific promised deliverable tied to what this piece just taught. keep it in the creator's voice and under 2 sentences."

the keyword captures the lead into the automation.

the question produces the substantive comment the algorithm rewards, so both signals fire off 1 action from the viewer.

Why It Prints

reach without capture pays nothing, and the capture line is the cheapest thing in the pipeline to get right.

the same viewer either becomes a lead or becomes a view, and the difference is 2 sentences the model writes for free.

---

Job 3: Run the Qualification Sequence

thousands of captured leads are worth nothing if nobody sorts them.

How to Run It

Opus 5 writes and runs the 3-stage sequence.

stage 1 delivers exactly what the CTA promised, no strings, since delivering clean builds the trust the rest of the sequence spends.

stage 2 qualifies with a short question about their situation, and the answer sorts buyers from browsers without a human touching anything.

stage 3 routes, moving qualified leads toward the offer and keeping browsers warm in the audience.

> "here are the DM replies from the last batch of captured leads. sort them into ready to buy, needs nurturing, and not a fit, based on what they said about their situation. draft the next message for each group in the creator's voice."

Why It Prints

the sequence turns raw volume into a short list worth human attention.

that filtering is the step that converts a content operation into a revenue operation, and it runs on judgment the model can apply at scale.

---

Job 4: Read Which Content Actually Produces Money

views tell you what got distributed.

captured leads tell you what worked, and those rankings are frequently different.

How to Run It

> "here is yesterday's deployment data plus the captured lead volume per piece. rank the pieces by leads captured per 1,000 impressions, not by views. identify what the top pieces share, the hook mechanic, the angle, the format, the emotional driver. then produce today's brief weighted toward those patterns."

that ranking is the one that matters.

a piece with big reach and no captures is entertainment, and a piece with modest reach and heavy captures is the one to multiply.

Why It Prints

the operation reallocates toward what converts instead of what performs.

over a few weeks the whole content mix drifts toward the patterns that produce revenue, and that drift is worth more than any single winner.

---

Job 5: Multiply the Winners

a winner is a proven structure the audience already voted for, and proven structures are the highest-probability material you'll ever produce.

How to Run It

> "this piece produced the most captured leads this week [paste script and data]. write 5 variations that keep the format structure, the emotional register, and the narrative arc, changing the specific details, examples, and visual notes. each targets a slightly different segment. preserve the closing CTA structure."

deploy the variations across different accounts over the following days, and 1 winner becomes 6 to 12 strong pieces.

Why It Prints

variations built on a converting structure convert at a far higher rate than fresh guesses.

as winners accumulate, a growing share of your output is high-probability revenue content instead of testing, which is the compounding the whole model runs on.

---

Job 6: Package It for Clients

everything above runs per client, and the system doesn't change between them.

How to Run It

Opus 5 handles the client side too, the onboarding research on a new niche, the reporting that shows what the content produced, and the briefs that keep each client's mix pointed at their buyers.

> "summarize this month's performance for the client. lead with captured leads and qualified conversations, then the content patterns that produced them, then what we're testing next month. keep it plain and specific, no filler."

Why It Prints

a client stays when they can see what they're paying for.

reporting that leads with leads instead of impressions is what makes the value obvious, and obvious value is what makes a retainer renew.

Chain the Jobs Into 1 Brief

the 6 jobs are separate for explaining, and in practice they run chained.

Opus 5 holds long multi-step work, so the chain gets handed over whole instead of stepped through.

> "here's the mined niche material, the persona docs, and yesterday's lead data. rank the buying triggers, build today's angles weighted toward what converted, write the full scripts in each creator's voice with the closing capture beats, then flag anything that breaks the client's banned phrasing."

that's the research, the analysis, the scripting, and the capture writing as 1 pass.

on a weaker model that chain drops constraints somewhere in the middle, and the reason to run a strong thinking model is precisely that it holds together to the end.

the time saved is real, and the bigger win is that every stage sees the context of every other stage, which is what makes the output coherent instead of assembled.

---

Route the Model Like You Route the Renders

1 practical note, since the same routing logic that governs generation applies here.

the premium thinking tier earns its cost on the jobs where judgment compounds, the research, the analytics, the qualification, the winner extraction.

bulk mechanical work, the routine batch writing once the templates are stable, can run cheaper without hurting the outcome.

spend the strong model where a better decision is worth real money, and spend less where the work is repetitive execution.

that's the same principle as testing on the cheap render tier and finishing winners on the premium one, applied to the thinking layer.