So much digital ink has been spilled regarding how generative AI is a first-class productivity booster.
Estimates on this score remain speculative, even if the expectations are robust.
Ninety-one percent of business and technology leaders surveyed by Deloitte1 expect genAI to improve their organizations’ productivity, with 27% expecting a significant productivity boost.
Such stats sound great and are available in dozens of permutations from many researchers, analysts and pundits. But what does a productivity boost look like? And how might it work?
GenAI: An Experiment
Practical evidence from researchers and analysts remains scant, but LinkedIn is full of posts about LLM (Large Language Model) prompting experiments and best practices.
For instance, Ethan Mollick tried this 30-minute experiment more than a year ago and revisited it recently with a sub-1-minute exercise. Of note was the performance and accuracy leap between his experiments.
Try something similar here, with marketing collateral as the context. As the end user constructing the prompts, you’re going to create a content roadmap for a new mobile application. Picture a Starbucks mobile app or something similar. You pick the industry, but retail probably makes the most sense.
This marketing plan must cover five customary bases: a corporate messaging guide, a snazzy presentation, a unique product logo, a bubbly blog post, and a video script designed to educate folks. Each piece of content explains how the product works and why the company is building it.
Normally, a marketing chief might delegate these tasks to staff, giving them days or a week (mileage will vary) to produce such a broad swath of content, with time for feedback cycles and collaboration. But genAI is a force multiplier for content, so let’s roll with a deadline of one hour to complete all five tasks.
Now give it a shot. Log into genAI applications in 5 browser windows (one for each piece of content). You can use your favorite service for text, but you’ll want to use a genAI presentation app such as Gamma for the presentation and an image generator such as Midjourney for the logo.
Next, tailor prompts to the content you want created. For example, try something like: You’re a content marketing manager. Create a messaging one-pager for a new mobile app that lets people pay for goods and collect and redeem rewards points. Do it in a formal tone, using no more than 800 words.
For the logo, try: You’re a graphic designer embedded with marketing. Design a memorable and unique logo for a new mobile order and pay application that will appeal to a broad range of consumer demographics.
Again, be as specific as possible. What product are you aiming to promote? The image description in your prompt should augment the desired output.
Then draft similar prompts for the remaining assets in the remaining browser windows. Hit enter and wait for the magic to happen. You should get complete responses within seconds.
Where the machine hands off to the human
Now read each one. What did you get in return? How did the LLMs do? Most likely, they produced something credible approaching pretty darn good. These will not be finished assets, but they should be passable first drafts. Prompting is both art and science, and opinions vary on the approaches.
Now consider this: a software program did all these tasks in seconds—not the full work week it might take a human, or a day for each task, if split among five employees.
Remember this pilot flight of fancy runs for one hour. Say you’ve spent 15 to 20 minutes reading the content. That still leaves 40 to 45 minutes for refinement.
The machine performed the magic: the real prestige, or trick reveal, lies ahead. You have your work cut out for you by becoming the human collaborator in the human-machine partnership. You and the genAI services are essentially copilots.
You’ll tweak the prompts for context, richness and tone. For example, strike an informal tone for the blog post. You’ll spend most of your time tailoring the output to your liking, editing out dull or inaccurate text and asking the service to flesh out more details that suit your intended messaging.
Multimodal and Beyond
As impressed as you may have been by the outputs, the future looks even brighter for genAI.
More GPTs are evolving beyond processing and answering in text mode to incorporate multimodal capabilities. This makes them capable of responding to multiple inputs, such as images and sounds in addition to text, and generating similar output.
Increasingly, vendors leading these genAI creations are partnering to drive business value.
Some are integrating their new genAI products with their existing software to boost worker productivity. And interactive genAI, in which systems actively engage with users, will enrich human-machine collaboration.
Regardless of what winding path genAI ventures, as employees build up their prompting muscle memory and learn how to harness these tools, they will boost their productivity and skills. Enterprises will tap personalized technology skills development to drive $1 trillion in productivity gains by 2026, enabled by genAI and automation everywhere, according to IDC research.2
Don’t have the resources in-house? Most organizations don’t. Fortunately, a growing and open ecosystem of partners with experience working with these systems can help you get the right combination of people, process and technology to achieve your desired outcomes.
Dell provides high-performing servers, client devices, and professional services to help organizations deploy genAI on premises in trusted, secure environments.
Learn more about how Dell Generative AI Solutions.
1 Deloitte’s State of Generative AI in the Enterprise, Deloitte, March 2024
2 Workforce Upskilling for the AI Era, IDC, Jan. 2024
