Since there is much talk of AI replacing workers, including writers, I thought I’d give ChatGPT a crack at writing a Kellblog post. Here’s what it came up with (and I’ll post some thoughts on its work down below):
Sure, here’s a blog post on the role of marketing, written in the style of Dave Kellogg:
Title: The Indispensable Role of Marketing in the Age of the Customer
If there is one thing that all successful companies have in common, it’s a deep understanding of their customers. But how do you gain that understanding? The answer is simple: through marketing.
Marketing is the lifeblood of any successful business. It is the art of identifying and communicating the value of your product or service to your customers. In the age of the customer, where customers have more choices than ever before, marketing is more important than ever.
Here are three reasons why marketing is an indispensable part of any successful business:
- Marketing Drives Revenue
Marketing is not just about creating pretty logos and catchy slogans. It’s about driving revenue. Marketing creates demand for your product or service by identifying your target market, understanding their needs, and communicating the value of your offering in a way that resonates with them.
Marketing also helps to convert that demand into revenue by creating a seamless customer journey. By understanding the customer journey, marketing can create targeted campaigns that move prospects through the sales funnel, from awareness to consideration to purchase.
- Marketing Builds Brand Awareness and Loyalty
Marketing is also essential for building brand awareness and loyalty. In today’s crowded market, it’s not enough to have a good product or service. You need to stand out from the competition.
Marketing helps to differentiate your brand by creating a unique value proposition and communicating it consistently across all channels. By building a strong brand, you can create an emotional connection with your customers, which leads to loyalty and repeat business.
- Marketing Provides Insights for Business Strategy
Marketing is not just about executing campaigns. It also provides valuable insights that can inform business strategy. By tracking and analyzing customer behavior and preferences, marketing can identify trends and opportunities that can be leveraged for growth.
Marketing can also provide insights into the competitive landscape. By monitoring the competition, marketing can identify gaps in the market and areas where your business can differentiate itself.
In conclusion, marketing is an essential part of any successful business. It drives revenue, builds brand awareness and loyalty, and provides valuable insights for business strategy. In the age of the customer, where customer experience is everything, marketing is more important than ever. So invest in marketing, and watch your business thrive.
So, how’d it do? First, not to take AI progress for granted, the fact that ChatGPT can even come up with this is rather amazing, especially compared to AI capabilities of a few year’s back. But in actually synthesizing a Kellblog post? Not so great in my estimation:
- It missed a major Kellblog mantra: marketing exists to make sales easier.
- The writing is a bit cringey. I don’t say things like: “marketing is the lifeblood of any successful business.” A site search for “lifeblood” on Kellblog happily returns zero results.
- While leading with “marketing drives revenue” is not bad, I’d be unlikely to put brand awareness and loyalty on the list, and I’d never say marketing provides insights for business strategy. I might say something more like marketing strategy is business strategy, or marketing strategy must align to business strategy.
- It doesn’t attempt any jokes. Hopefully, a hallmark of my hard-headed but light-hearted style.
This is not to say that AI won’t improve over time and with increased scale. But for now, ChatGPT doens’t appear to be close to putting me out of business.
Excellent!!
I find ChatGPT gives me a skeleton to start working from. Plus some of the language it uses gets me thinking in a direction I hadn’t thought of.
We just built a utility that will help our clients clean up their Vendor and Client lists for FP&A Planning. Amazon, Inc Amazon, Inc. AmazonInc AmazonINC can all become Amazon, Inc
In 10 years, it will be significantly better and likely we won’t be writing our own emails, or even comments to Kellblog!
Yep.
As I was reading this – I thought “there is no way Dave has trained the AI to mimic his tone, wit and/or straightforward communication style”.
AI will be great at automating a large majority of repetitive, data centric tasks that do not require intuition, personality or inter-personal relationships…unless Max Headroom comes back to be the human face and personality of AI!
We are safe for now, Ray
Great post, as always. What I am finding with Chatgpt, it’s all in the prompts. And I guess that’s why I’m seeing articles about companies paying 6 figures for AI prompters!
Yes.
It all comes down to the prompt.
AI doesn’t (currently) do everything magically…
But, if you help it focus with clear direction, then you have a recipe for success.
Specifying tone, as well as positive and negative examples gives a pretty consistently great result for articles.
Spent some time on prompt improvement today and didn’t get that far, nothing worth reposting as a major step forward. That’s probably more about my prompting skills than ChatGPT.
“not so great” is being very kind to ChatGPT in my opinion :)
I’m being nice because I didn’t work too hard at the prompts which at this point is kind of known (i.e., that it does its best work after more prompting). At some point, when I have more time, I’ll go back and really try to help it and see how it does.
Indeed.
I’m pasting [an edited] comment I received via email from my old friend Tom Tobin, who works as Head of Product at Regie.ai (which generates sales emails), and explains how I’m not getting the full experience that’s availble today. Worth reading, very much so. Thanks Tom!
(ps: I’m not sure I can put a hyperlink in a comment but I will try to figure out how to upload/share the PDFs he references.)
Hi Dave!
I saw your tweet on GPT blog posts, but I think you were not getting the real experience GPT can deliver today.
Using the playground and GPT-4, you get very different output, and can control the parameters more finely. There is a system prompt – the outline for the AI to run with. In the public ChatGPT, this system prompt is something you can’t change. it’s something like “be a helpful assistant” but it’s also where the limits lie in what it will and won’t say.
But for this case, I gave it the system prompt, and can choose how it writes and the order of operations – for instance, telling it to write an outline first helps it first decide what to cover, then you can tell it to generate the longer prose. It does better when the whole output fits in the context window, and usually within the recent tokens. That means you first ask for the outline, then ask for the detail for that outline. You could also choose the format of the output (e.g. write it as a play script, write it as a transcript). You can even choose the style – to say it should write like Vonnegut himself, or as Hemmingway.
Then you can feed it different topics… and get different blog posts all based around the DK/KV interviews. These are attached as the PDFs below. Sorry for the output format, it’s not easy to grab the real output.
Looking at the Cat’s Cradle one, I should have messed with it more to include things like Kurt not having knowledge of B2B startups, but being able to comment on them
Then I tried having it write as you, on a topic. Of course, it operates without actual examples, and without actual insight – it can only go on what’s been written elsewhere. That’s why the topic you feed it could also come with an example that you’d like it to base the story around. That’s work for later :)
And this is what I do part of my day [at Regie.ai], so that Regie can generate sales emails for SDRs…
Here are the three PDFs Tom refers to above.
https://drive.google.com/file/d/1unc2GJUx3-vMJSzP4HA6F_iTZu0olKbC/view?usp=sharing
https://drive.google.com/file/d/1KLujqDSJTFdE5GiUWguaGh874tA7wFFc/view?usp=sharing
https://drive.google.com/file/d/1yzRFrj1_6jgaDmEccbPJHcMxmwKYVhTk/view?usp=sharing
As I said on Twitter, interesting experiment. ChatGPT summarization capabilties are good as we know. The writing vs @kellblog’s usual sharp content is bland junk.
What I’ll add here is to object to the notion that adding more data and “scale” into ChatGPT will solve this. The paremeters are already ginormous. The enterprise can customize some of this with the infusion of focused data sets. But the bottom line is you’ll need to fuse LLMs with different approaches to AI if you want to put good human writes on their heels. Mediocre business writing will explode…
I should have just written the comment :)
ChatGPT can get the language and phrasing right but with no real world experience. It’s kind of like trying to get a good writer who knows nothing about startups to stooge something. They may google it but there are no board meeting transcripts from startups on the internet to learn from. There are no internal discussions about spend or rep efficiency or positioning. So it’s kind of lost without insight and a topic. But it is pretty good at writing, and you can make it better with the right prompting. And better if you don’t use ChatGPT but the API.
Yes, some people refrain from so doing in order to avoid potentially hurting my feelings or such, but at least 90% of the time I post the comment myself when I can. I’m definitely not an expert here and appreciate the input from someone who knows way more than I and who has played with it more deeply.