Before you read any further, I want you to know I’m not talking about your marketing or PR agency and how AI is upending what they do (and how they bill).
I’m talking about YOU and YOUR JOB.
What you’re good at, what you like to do … And how your role at work is being reshaped by generative AI.
I’m guessing you’ve experienced many changes in the past couple of years (months?) as you’ve become more proficient with AI tools.
The quality of your LLM outputs — though still not perfect — has vastly improved.
So has the way you interact with AI. You’ve figured out how to communicate with machines by being specific, detailed and preparing your requests in a structured and organized way.
In a sense, you and AI have gotten to know and learn from each other.
My AI Progression
Like you, I’ve been using AI for more and more tasks. For instance, it’s replaced the way I search for simple informational requests.
Here are a few more sophisticated examples.
Expense Tracker
I developed an AI expense tracker and feed it receipts that it reads, organizes by date and type, enters into a spreadsheet that totals the amount and tells me it’s ready to review. Once I give it the OK, it combines the receipts and spreadsheet into one PDF, names it and places it in a folder where I can easily email it to a client. That took a bit of drudgery off my plate, though I did wonder if mundane requests like these bore the crap out of AI.
Presentation Visuals
I’ve never been very good at finding visuals for slides. So I started asking AI to generate them one at a time. But now, I’ll feed in a slide deck without images to Claude, describe the style, tone and colour palette I want, and Claude prepares a detailed visual description — it calls it a ‘style spine’– and a content-specific image prompt for each slide. I then feed the style spine and prompts into an image generator like ChatGPT and it creates the visuals in a series, stopping to ask for my edits or approval. When it’s done, I manually paste the pictures into the slides. I’m sure with a little more thinking and the right connections, I could create an agent to do that part, too.
Client Agency Simulation
I used to write fictional marketing briefs, hand them out to my Schulich School of Business – York University Digital Marketing students and ask them to develop a strategic plan. Last year, I ‘vibe coded’ a full agency client simulation, complete with made-up companies and new products, founder stories, target audiences, launch goals and even clients with varying personalities. Students get a brief, can ask the client questions and then use AI tools to build a marketing plan for the company in real time. It’s quite a sophisticated setup and something I could never have built on my own.
Then Along Came Agents …
The many tool-using AI systems that let you assign varying degrees of autonomy and automation to them, and then they go off and perform a repeatable workflow while you focus on something else.
You can connect Claude Cowork, Gemini or ChatGPT to your accounts on Gmail, Canva, Slack and other work-related apps. You can also give them access to your files and folders and let them read, modify and even reorganize them.
You can also set up projects, systems designed to perform a recurring task.
For instance, I often prepare training proposals based on a template and fill in the blanks. So I developed a project to do that with detailed, step by step instructions about what I want it to do and how I like the end result to look. The LLM generates a new proposal based on the (non-proprietary) info I feed into it. Then, I edit it in my voice, remove the AI-isms and make changes before sending it to the client. That saves me a lot of time.
But Here’s the Rub
I’ve often said I don’t use AI for writing. For this newsletter or my lectures, I prefer staring with a blank page or screen. Procrastinating. Agonizing. Sweating through the process of putting my thoughts together in a coherent way.
As Kurt Vonnegut once said, that lets you to “revise [your] stupidity, to edit [yourself] into something like intelligence.”
But what if the writing you’re trying to do is more functional than creative? More copy than writing?
Like the proposal project I mentioned. Is that a such a bad way to use AI? After all, you’re starting from a template anyway. Why not let a machine do the heavy lifting?
How about getting AI to update slide decks for presentations you created and maybe add in an idea or two of its own? Or asking it for a first draft based on your initial idea and a rambling conversation?
As I share my more of my working day with AI, I wonder if my thinking is getting tangled up with the machine?
Am I dulling my creative senses? Redefining who I am and what my contributions are? Or is this a hurdle I have to overcome as I learn to work with AI in new ways?
This is something I’m struggling with: maintaining my human agency while collaborating with AI agents designed to take work off my plate.
How is it affecting you?
Staying Grounded on a Slippery Slope
It feels like we’re heading toward an inflection point, where it will soon be difficult to claim the work is ours alone. If we’re making the big decisions or if they’re being driven by a machine.
When you begin using agents, you proceed with caution because you’re more hesitant of what they might do. You read and reflect on the AI agent’s suggestions, consider them and make the choice. You’re in the driver’s seat.
But then, as the agent’s memory of what you want improves and you’re facing a tight deadline, you start skimming and pressing, approve, approve, approve … Till your default becomes ‘approve all’.
And you’ve moved to the passenger side of the relationship.
Don’t Use AI on Autopilot
Whenever you’re using an agent, be intentional about the model you choose and the level of permissions you assign. I’d advise you to be on the conservative side and structure it with multiple check-ins and approvals.
Be absolutely clear about your goal and what you want your agent to do. That requires planning, forethought and detailed, specific instructions that could go on for pages.
And always aim high. Differentiate between excellence, the human element you provide and outputs that are simply good enough.
Ultimately, it’s up to you to decide what do you want the AI to do and which parts of the creative process you keep for yourself.
This post was written by Martin Waxman, MCM, APR, Associate Director, Future of Marketing Institute and an Adjunct Professor at the York University Schulich School of Business. A version of this post was originally published on Martin’s AI and Digital Marketing Trends newsletter.