The many ways creative folks are using generative A.I. in their work

While intelligent machines haven't taken over, they're surely influencing the way we work.

Around a decade ago, most agency folks were scrambling to open Facebook pages for their clients, figure out the right image resolutions, and fit a post’s copy within the permitted character limit.

Today, the same folks or their younger colleagues are scrambling to master the many generative A.I. tools popping up every second day, and the prompts to feed them.

ChatGPT, Dall-E, Midjourney and their ilk are the comings of the second age in India’s digital creative and marketing timeline. They are called generative A.I. because they possess the ability to create new content, including text, images, video, audio, and many other formats.

These tools, in the past three to six months, have captured the creative and marketing world’s zeitgeist. Hollywood actor Ryan Reynold’s Mint Mobile, a virtual mobile network, in January 2023, used ChatGPT to script a TV ad while AB InBev’s Beck recently used A.I. to produce a limited-edition beer and its full marketing campaign.

Their rising indiscriminate use has fuelled an atmosphere of excitement as well as a fear of the machines taking over people’s jobs, and ultimately humanity like the movies.

To look at it as a threat or as something that will improve your efficiency depends on the outlook of the receiver says John Paite, chief creative officer (Art & Tech), Media.Monks, an international digital agency owned by Martin Sorrell's S4Capital. “I think right now no one is running out of jobs… it will create more jobs,” he says pointing to the new jobs for people skilled at doling out prompts to these generative A.I. tools.

Generative A.I. has become a part of the proverbial water cooler discussions, and while there are people on both sides of the conversation, one point has taken centre stage: Everybody wants to use it once. 

“Curiosity” is what led Deepak Gopalakrishnan to try ChatGPT in the first place. He does not see it as a fad and thinks it is the start of a new era. “I wanted to get into it in the beginning and get my hands dirty just to see what it is,” he reveals. Gopalakrishnan is a freelancer who works in the digital marketing and content landscape.

He uses ChatGPT4 (a paid version), and has used Midjourney, a tool which generates images from language description. He looks at them as “the starting point” which he then uses to refine and rework before presenting the outcome to clients.  

Many, like him, are using these generative A.I. tools as a base on which they build their work. Media.Monk’s Paite tells us they use tools like ChatGPT and Midjourney as a kickstart for say a writer facing a mental block or when a client requests last-minute changes.

Independent agency Schbang uses tools like ChatGPT; it takes the output and combines it with human expertise “to adapt the content so that it suits the brand personality and the tone of voice of the client we are catering to,” says Eva R Sachdeva, the agency’s content lead and manager.

She cites the agency’s work for a skincare brand where the prompts ensured the tool generates content which is custom to the client. Here is what she fed to ChatGPT:

Can you give me an outline for writing a blog on the topic '5 summer skincare tips'? It has to be from the perspective of a brand that sells skincare items in India and the point of view and tone of voice needs to be in the second person and you need to talk like a skincare expert giving these tips and selling your product to the consumer.

The content which comes out is not used right away. Once the blog post — a mix of human and generative A.I. work — is ready, it is run through an A.I. detection tool like Copyleaks. “Once detected, we use Quillbot or Copy.AI, these tools will give us a paraphrased version of that particular A.I. generated content,” says Sachdeva. After all the rework, the final piece is run through a plagiarism checker before being published on the client’s website or blog.

Also Read: AI, ChatGPT and its ilk: boon or bane for agencies?

On the other hand, Talented, a Bangalore-based creative agency is using many generative A.I. tools to, as per co-founder P.G. Aditya, “streamline a lot of processes internally while also creating new processes altogether that collaborate with A.I. as opposed to just utilising it.”

Visualisers use visual A.I. models and the agency is seeing writers and strategists use it to their advantage to speed up processes. Aditya says A.I. that uses LLMs (Language learning models) have proven to be supremely useful - be it for research, bouncing ideas off of, or just using them to break down and describe visual work. “LLMs like ChatGPT and Compose.ai have proven themselves useful in cutting down work time and increasing creation time.”

Talented uses GPT-powered applications like ChatGPT, Compose.ai, Tome, Gama, and Beautiful.ai. It uses visual generation AI tools, especially ones that take advantage of Stable Diffusion like Midjourney, Gen1 by Runway, Dall-E, Scribble Diffusion, and DeepAgency.

An account win

For all the increased efficiency and time-saving generative A.I. offers, what is the end game for an agency – win new work? “We won two clients in the Middle East with retail store design we'd created using these A.I. tools," says Mahesh Murthy, founder, Pinstorm.

He tells us the agency has subscriptions to several generative A.I. tools and uses them “to look into the product category and see what kind of conventional and unconventional strategies it suggests.”

What is interesting, as per Murthy, is the agency’s internal rule that the employees’ strategy should be as good as the one generative A.I. suggests. “This is the starting point, and if your work doesn't come up to it…”

He says ChatGPT is not good when it comes to writing advertising copy, it's good at writing body copy and content, but "it doesn't know to write ads so I think a copywriter's job is still safe. We think it's pretty good at visualisation, you've to learn how to push it to do that. We're already using it across every single one of our clients."

When asked about Pinstorm’s clients’ response to knowing generative A.I. was being used and if it had affected revenue, says Murthy, “Clients don't pay us to use tools, they pay us for people and the hours we put in. We're still putting the same people and the number of hours.”

The art of the prompt

Like a young pup which needs instruction to keep it and others from harm, a user needs to prompt generative A.I. to get it to do the right thing. “A lot of what A.I. promises to do is only realised when it’s prompted correctly. Much like any new piece of technology, trial and error comes is core to the process,” says Talented’s P.G. Aditya.

Gopalakrishnan gave an example of how he and his wife, used ChatGPT for a presentation. “We'd given it a few situations saying this is the brand, this is the industry, this is the persona we're trying to define with the brand. it gave us a good enough starting point for us to improve on.”

One may assume prompting is the next big thing to learn but it is not. It all boils down to knowing what you want and feeding it to the tool. You can keep giving it prompts and “it has the wisdom to adapt and learn from the prompts you give,” remarks Schbang’s Sachdeva.

The new teacher

An interesting aspect of these tools is their ability to teach. “ChatGPT allows our junior writers to learn,” states Bonsy Mehta, Schbang’s head of content marketing, consumer centricity, and branding.

She references ChatGPT and how young writers, who make errors while writing, do not have to depend only on an editor like in the old times but can use the tool to learn where they err.

Where she feels ChatGPT helps the most is with the tone of voice. “It will give you the same article in the tone of a voice of a 15-year-old and a 45-year-old.” Allying fears like Pait, Mehta believes ChatGPT won't replace writers but will allow them to hone their skills more.

A peek into the near future

Machines, unlike humans, can master data quicker, and therein lies their superpower. As time passes, generative A.I. tools will grow in sophistication and the first signs of this uplift come from plugins.

Think of them as additions which let you customise apps, programs, and browsers. “Plugins will increase the power of ChatGPT by at least 5X,” says Jacob Joesph, vice president, Data Science at CleverTap, a customer engagement and retention platform.

Companies such as Expedia, Instacart, Shopify, Slack, Wolfram, and Zapier have developed the first plugins for ChatGPT.

He, for instance, tells a story about a user who wishes to cook palak paneer. Using Instacart, a grocery delivery and pick-up company, he prompts the dish he wishes to cook and the number of servings. Now, the company app/website will calculate which ingredients in what quantity are needed and will deliver them for you. 

Jacob believes every company will benefit from generative A.I. tools. Traditional mortar companies are developing online offerings and “with tools like GPT the cost of intelligence has come down.”

He explains this using ChatGPT’s pricing models, “Six months back, for 1000 tokens, which is like 750 words, I had to pay four cents which three months back they brought it back to two cents, and now it is 1/10th of it with GPT 3.5.”

Unlike the many flashy trends in the past eight to sixteen months, generative A.I. has fared better and is here to stay. “I feel missing this wave is like someone who's missing the digital wave that started about 10-15 years ago and this is arguably going to be even more powerful,” says Gopalakrishnan.

The machines are here to stay and generate. 

Photo by Alex Knight