Where Does AI Save Time at an Agency, and Where Does Judgment Still Matter?

Quick Answer

 

Where does AI save time at an agency, and where does human judgment still matter? A COO's practical guide to drafting, contracts and client calls.

Where Does AI Save Time at an Agency, and Where Does Judgment Still Matter?

Twenty-two years of grading English essays gave me a very specific radar for prose that sounds right and isn't. Every AP teacher knows the type: the paragraph with perfect transitions, confident topic sentences and a quote from a chapter the student clearly met only through SparkNotes. AI writes that paragraph faster than any sophomore ever could, which is both the good news and the problem.

Researchers have a better name for it. In a 2023 field experiment with 758 Boston Consulting Group consultants, a team from Harvard Business School and other schools described a "jagged technological frontier," where AI is brilliant on one side of the line and confidently wrong on the other, and nobody marks where the line is. That's the most useful phrase I know for anyone rolling out AI at work. In this article, I break down where AI saves us real time and where we keep a human in charge.

Quick answer: At an agency, AI saves the most time on first drafts, summaries, document prep and research synthesis. Human judgment still matters most in relationships, negotiation, creative decisions and choosing the right partner for a brand. Ian Drummond, COO of Hollywood Branded, recommends letting AI draft while a named person reviews and owns anything a client or partner sees.


Where Does AI Actually Save Time at an Agency?

The honest answer is the blank page. Most of the hours AI gives back come from work that used to start at zero and now starts with something on the page: a first draft of a scope of work built from our templates, a summary of a long email thread, a meeting recap with action items, a research brief pulled together from a stack of sources.

Our workflows run across Microsoft 365, PandaDoc and Claude. In general terms, that means drafting and summarizing in the tools our team already lives in, preparing contract documents faster from approved language, and using AI to organize research before a person decides what it means. None of it is exotic. It's the administrative weight that used to sit between a good idea and the work itself, and nobody on the team has asked for it back.

The BCG numbers back up how big that inside-the-frontier gain can be: consultants using AI finished 12.2% more tasks, worked 25.1% faster and produced work rated more than 40% higher in quality. That's also why I made AI use mandatory in our hiring assessments. I'm not looking for candidates who treat a chatbot like a magic trick. I want people who can direct it, check it and know when to ignore it.


The First Draft Is Cheap Now. The Final Draft Isn't.

When a draft costs almost nothing, it's tempting to treat it as almost done. The same study shows why that's risky. On a task the AI wasn't good at, consultants using it were 19 percentage points less likely to get the right answer than consultants working without it, because the output looked right enough to trust.

In agency work, "looks right" is where trouble hides. A contract summary that drops a usage-rights limitation, a recap that misstates who agreed to what, or a brief that invents a statistic can all read smoothly, the same way the SparkNotes essay did. So our rule is that AI output is a starting point that a named person reviews before it goes anywhere. Speed on the draft only counts if the review is real.


Where Judgment Still Matters

Negotiation is the clearest case of work that doesn't get faster with AI, because the value is the judgment. Knowing when to hold a position, when to concede something small to protect something big, and when to stop typing and pick up the phone comes from reading the people on the other side of the table. I've spent enough time on SOWs, MSAs and talent fee conversations to know the most important sentence in a negotiation is often the one nobody writes down.

Creative calls belong in the same category. AI can generate fifty integration concepts in a minute. It can't tell you which one a showrunner will actually welcome, which one fits a brand's long-term positioning, or which one sounds clever in a deck and lands badly on screen. Those decisions depend on relationships and context that live with people.

Then there's fit. A partnership works when it creates Mutual Benefit for the brand and the partner, and judging that takes experience with both sides. A model can sort a list, but deciding what belongs on it is the job.


A Simple Test for Who Owns the Task

The rule of thumb I use is about consequences. If a mistake would be caught internally and fixed in five minutes, AI can do most of the work with a quick review. If a mistake would reach a client, a partner, a contract or a check, a person owns it from start to finish and uses AI the way a good writer uses a thesaurus: helpful, occasionally brilliant, and never in charge.

That test keeps the team fast on the low-stakes work, which is most of the volume. It also makes ownership obvious on the high-stakes work, where a smooth paragraph and a correct one can look identical until someone checks.


What Does This Mean for Brands Choosing an Agency?

At Hollywood Branded, we help brands become part of pop culture through a repeatable pathway: Playbook → Radar → Unlock Access → Produce → Amplify → Grow. AI helps at several of those steps, especially where volume is high. Radar tracks 6,000+ opportunities, and tools help us sort and summarize at that scale. Deciding which opportunity actually fits a brand, and getting the right people to say yes, is still human work.

If you're evaluating any agency partner, ask two plain questions. Where do you use AI, and who signs off before it reaches me? A good answer names specific tasks and specific owners. A vague answer tells you the frontier isn't being watched.


Map One Workflow Before You Automate Ten

Pick one recurring task this week, such as meeting recaps or first-draft proposals, and write down who reviews the AI output and what they check for. Run it for a month, then decide what to expand. The teams that get the most from AI tend to be the ones that know exactly where the line is, and they rarely got there by automating everything in week one.

New call-to-action


Frequently Asked Questions

Q. What tasks should a marketing agency use AI for?

A. AI is most useful for first drafts, summaries of long threads or meetings, document preparation from approved templates, and organizing research. These tasks have high volume and are easy for a person to review. Final decisions and anything client-facing should still have a named human owner.


Q. What is the jagged technological frontier in AI?

A. It's a term from a 2023 Harvard Business School and BCG study describing how AI performs very well on some tasks and poorly on others that seem similar. Consultants using AI improved speed and quality on tasks inside the frontier but were 19 percentage points less likely to be correct on a task outside it. The boundary isn't always obvious, which is why human review matters.


Q. Can AI replace negotiation and relationship work in brand partnerships?

A. No. AI can help prepare for a negotiation by summarizing terms or drafting talking points, but reading the other side, deciding what to concede and building trust are human skills. Brand partnerships depend on relationships that last beyond a single deal.


Eager To Learn More?

Keep reading for more on how brands and agencies work smarter.

Want to stay in the know with all things pop culture? Look no further than our Hot in Hollywood newsletter! Each week, we compile a list of the most talked-about moments in the entertainment industry, all for you to enjoy!

New call-to-action

 

Table Of Contents

    WHAT'S HOT IN HOLLYWOOD

    Want to keep your pulse on what Hollywood has to offer? Sign up below to get on our HOT IN HOLLYWOOD partnership updates!