AI Doesn't Write Our Content. It Helps Us Make It Better.
AI has opened the floodgates to content. There is more being published than ever before, and with so much of it now produced or assisted by AI, it's natural to wonder how much of what we're reading is actually being written by the people whose names are on it.
For almost two years, we've been publishing steadily. Most of our articles and interviews run between 2,000 and 3,000 words, and producing that kind of long-form content on an ongoing basis is an enormous amount of work. AI is very much part of how we do it.
But AI doesn't write our articles.
Fortunately, we're never particularly short of things to talk about. We already know the subjects that matter because they're the same issues we work on with clients every day: strategy, business development, growth, brand positioning and communications. Those are the conversations we're already having, and they naturally become the conversations we write about.
What AI does, and how we work with it, is a different story.
THE MORE THINGS CHANGE, THE MORE THEY STAY THE SAME
Funny enough, the way I work with AI today isn't really new at all. I've been working this way for most of my career.
When I ran my PR agency, Buzz, we produced prodigious amounts of copy. This was back when a media kit could contain ten different documents, from news releases to backgrounders, fact sheets, executive biographies, articles and case studies, each two or three pages long. There was an enormous amount of writing.
I learned fairly early what I was good at and what I wasn't. I'm not a graceful writer. Crafting beautiful sentences has never been my particular talent, so from the beginning, I worked with writers.
My role was different. I knew what the story was. I knew what we needed to say, what was interesting about it, what needed to be emphasized and how the argument needed to develop. I knew how the story should flow. Someone else would then put the words together.
What initially perplexed me was that sometimes I would get something back that was beautifully written, far more beautifully written than I could have written myself. But somehow, you couldn't quite understand the story. There wasn't a logical progression to the argument. You got to the end and all the information might have been there, but it didn't quite hang together.
I would take the copy and start moving paragraphs around. I wouldn't necessarily change a word. I'd move this paragraph up, put that one after it, shift something else to the end. And suddenly the story made sense.
Eventually I understood that this is my particular skill. I know how to find the story inside a lot of information. I know how to decide what matters, what doesn't, how the argument should develop and how to structure it so that someone else can follow it. I've worked that way for decades.
And in many respects, that's exactly how I work with AI today.
I DON’T REALLY PROMPT AI. I TALK TO IT.
When I first started using ChatGPT, I used it very simply. I would paste in a paragraph and say, essentially, clean this up. Fix the grammar. Make this sentence clearer. For probably the first year, that was most of what I did.
As I began to understand what the technology could do, I started experimenting. I tried asking it to write an article about a subject I knew well. I tested different ways of giving it instructions. I wanted to see whether I could simply tell it what I wanted and have a finished article come back.
What came back was generally correct. It was also generic, basic and completely inadequate for the kind of work we do.
Oomph operates in a highly specialized world, and our approach to the work is unusual. We combine decades of experience in consumer marketing, brand development and communications with knowledge gained from years of working inside architecture, engineering and other design firms.
Marketing in this industry is still a relatively young discipline, and much of what you need to know to do it well isn't something you learn in a marketing textbook. You learn it by working inside firms: understanding how they operate, how they pursue work, how they communicate highly technical expertise, how decisions get made and how marketing and business development actually work in practice.
We bring those two worlds together. We take the principles and disciplines of mainstream marketing and apply them to the very particular realities of design firms.
As a result, much of what we write about is fairly arcane. There simply isn't the same enormous body of existing material about marketing and business development in architecture and engineering that exists around mainstream marketing.
And that matters when you're working with a large language model. Ask AI for an article about branding or a general marketing strategy and it has mountains of material to draw upon. Ask it how a building sciences firm should develop a go-to-market strategy for entering a new geographic market, or how highly technical expertise inside an engineering firm can become the foundation of a sustained thought leadership program, and the pool gets considerably shallower.
When I experimented with asking AI to write that kind of content for us, what came back wasn't necessarily wrong. Generic. Predictable. Superficial. It was Mickey Mouse.It didn't reflect the complexity of the firms we work with or the knowledge and experience we bring to the problem.
AI can tell you what business development is. It can produce 750 perfectly competent words about strategic planning. It can summarize conventional wisdom about positioning. But that's not what we're selling. We're selling our thinking, our experience, our judgement and our particular way of looking at these problems.
AI HASN’T MADE THINKING FAST
I am probably one of the world's worst typists, so I dictate everything. And now that I've become comfortable working with AI, it has gone beyond dictation. It has become a conversation.
I don't really prompt ChatGPT when we're developing an article. Usually, I start by talking. A lot. I explain the idea I'm thinking about, why I think it matters, what I've been seeing with clients, what I've been reading about changes in the industry, and why I think this is something we should be talking about. Sometimes I know exactly what the story is and where I want it to go. Other times, we talk our way into it and shape the story as we go.
AI starts putting structure around all of that and producing copy. And then we start fighting. The conversation sounds something like this: “No, that's not what I mean. That's too consultant-y. We've already said that. This needs to come first. That's completely wrong. No. The real point is this. I'm sorry, this reads like a cheesy consultant. Make it sound like me. Put the whole section back together.”
And we keep going. We work through the article paragraph by paragraph. Some paragraphs are rewritten once or twice. Others may go through ten iterations. Sometimes we spend an hour developing an argument and then decide the whole thing is wrong. We leave it overnight, come back the next morning, pull it apart and start again. That's our process.
I keep timesheets, so I have a pretty good idea of how long these pieces actually take. I can't remember us producing a serious article in under three hours. Four or five hours is probably more typical. The interviews can easily represent eight hours of work once you include developing the questions, conducting the interview, working through the transcript, finding the story, writing and editing the piece. Some of our more complicated articles have consumed a day or a day and a half of almost continuous back and forth.
We put that amount of work into the content because it reflects what we think, and we take that seriously. We know people use these articles. They learn from them. We've had people tell us they've followed the advice in them. So the argument has to hold together. The advice has to be sound. We have to be precise about what we mean. If we're going to put something out into the world with our name on it and tell people this is how we think something should be done, we'd better mean it.
And then there is the language. I am allergic to consultant-speak. I don't want our content buried in jargon, clichés and highfalutin language designed to make relatively straightforward ideas sound more impressive than they are. I want it to be professional, intelligent and direct. People who know me will sometimes read one of our articles and say, “Oh yeah. That's Johanna talking.” That's deliberate.
Our content today is probably closer to the way I actually speak than much of the material I produced earlier in my career. Working conversationally with AI gives me an extraordinary degree of control over the language. I can keep pushing back until the words reflect not only what I think, but how I would actually say it.
Some of that, frankly, probably comes with having done this for a very long time. I'm no longer trying to prove that I belong in the room. I don't need our writing to sound like McKinsey to demonstrate that we know what we're talking about. Thirty years of experience gives you the confidence to say something plainly. The expertise is in the thinking, not in how complicated you can make the language sound.
There's one slightly ridiculous wrinkle in my claim that I don't prompt AI. That's true when I'm developing an article with ChatGPT. That's a running conversation, not a prompt. But once we've beaten a piece into shape, I'll sometimes take it to another AI tool for a final editorial pass, or use one for market research.
For that, I don't write the prompt myself. I ask ChatGPT to write it. Those prompts run long, sometimes two pages of instructions. So it's true: I don't prompt AI. I just get AI to write the prompts for me.
HAVING THE TOOLS ISN’T THE SAME AS HAVING THE EXPERTISE
We are still in the early days of figuring out what AI can and can't do. And AI isn't the only new technology we're grappling with. There is a parallel situation happening in architecture. Increasingly sophisticated modular construction systems are beginning to provide architects and builders with complete kits of parts from which substantial buildings can be assembled. There are systems with standardized components for exterior walls, windows, interior partitions, floors, kitchens and bathrooms that can be combined to construct an entire multi-storey residential tower.
But having the parts doesn't make you an architect.
A building is the result of thousands of interconnected decisions. What does it need to do? Who is it for? How should people live or work in it? How does it respond to its site? What are the zoning, regulatory and permitting constraints? What can the client afford? How should the building perform? And how do all of those requirements, constraints and opportunities come together in something that actually works?
The components may radically change how the building is constructed. They don't replace the knowledge, judgement, creativity and problem-solving required to create the building in the first place.
We are now going to see versions of this same situation across every area of knowledge-based work, and probably well beyond it. The tools will continue to change, and they will be enormously helpful. But our knowledge will still be the foundation of the work.
So it's on us, the experts, to be clear about how we work and what we bring to the table.
It isn't enough to say that we can't be replaced by AI or by a modular system. We need to be able to explain to our clients the thinking, accumulated knowledge and judgement that go into every situation and into the solutions we propose, and connect that expertise directly to their situation:
What does this mean for you? How does it apply to your particular situation? What problem are we solving? And why does it matter?
FIVE QUESTIONS THIS ARTICLE ANSWERS
1. Can AI write good thought leadership content for a professional services firm?
AI can produce competent content quickly, particularly on subjects where there is already a large body of information available. But strong thought leadership needs to reflect your firm's particular experience, judgement and point of view. AI can help shape and articulate that thinking, but the expertise itself has to come from you.
2. How can you use AI to create content without losing your firm's voice and point of view?
Treat AI as a collaborator rather than handing it the writing assignment. Start with your own ideas, experience and perspective, then use AI to help organize the argument, develop the copy and refine the language. Keep challenging and revising what comes back until both the thinking and the language genuinely sound like your firm.
3. Does using AI actually make writing thought leadership faster?
It can make parts of the process much faster, but it doesn't make the thinking fast. A substantive article can still take hours of discussion, restructuring, rewriting and editing because the difficult work is deciding what you actually want to say and making sure the argument holds together. AI accelerates production; it doesn't eliminate the intellectual work.
4. What role should subject-matter expertise play when using AI to create content?
It should be the foundation. The more specialized the subject, the less useful generic AI-generated material becomes. Your firm's accumulated knowledge, experience and judgement are what make the content distinctive and valuable; AI is a tool for helping you turn that expertise into something other people can understand and use.
5. How can professional services firms demonstrate the value of their expertise when clients have access to AI?
Make the thinking behind the work visible. Don't simply tell clients that your expertise can't be replaced by AI. Explain the knowledge, judgement and problem-solving that go into the decisions you make, then connect that expertise directly to the client's situation: what does it mean for them, what problem does it solve and why does it matter?
