From Growth Hacker to Prompt Engineer

In 2014 I co-founded Ladder after seeing a presentation by Bryan Balfour on the scientific method applied to marketing. I was immediately drawn to the systematic process for generating and testing ideas as an alternative to the vibes-based way most marketing was done at the time. The people above me were spending millions of dollars on TV ads with no real measurement of the ROI of each campaign, whereas I was A/B testing every decision and rapidly learning what works (or doesn’t). 

‍

Fast forward to 2024 and I’m about to release an O’Reilly book on Prompt Engineering for Generative AI, and I’m realizing that a lot of the same things that appealed to me about growth hacking also apply to prompt engineering. When I’m optimizing a prompt it activates the same parts of my brain as when I was optimizing landing pages – you have a goal in mind and you A/B test different creative approaches until you find that one magic combination that works.

‍

Most of the people I used to work with in growth became early adopters of AI, not least because it requires a lot of the same skills. It helps to be technical, data-driven, and able to come up with creative solutions to ambitious problems. However, I see a lot of the same problems we had in growth happening in prompt engineering: a focus on short-term tactics instead of strategy and process. I thought it’d be a fun exercise to go back through Balfour’s presentation and map out the similarities between working in growth and AI, for anyone making that transition.

Experimentation and iteration

Growth Hacking: This involves continuously experimenting with different strategies and tactics to find the most effective ways to grow a business. Growth hackers test various ideas rapidly and iterate based on the data they gather. They are constantly optimizing marketing campaigns, user experiences, and product features, making changes daily, weekly, and monthly in order to keep improving performance. It’s not about one magic tactic that grows your company, it’s about building a machine that is always testing new tactics to sustainably drive growth.

Prompt Engineering: Similarly, prompt engineering requires experimenting with different prompts to optimize AI-generated outputs. Engineers test multiple prompts, analyze the results, and refine them to improve the AI's performance. You can think of every prompt or chain of prompts as a campaign, and it’s your job to keep improving performance over time. The difference is that with prompt engineering, you get the results back in seconds without having to wait two weeks to conclude an A/B test! A lot of people think of prompt engineering as finding the right combination of magic words to trick the AI to do what you want, and that’s not it! It’s building a system for testing and learning what combinations work, in order to keep improving the accuracy and reliability of your AI applications.

Data-Driven Decision Making

Growth Hacking: Data is the backbone of growth hacking. Decisions are made based on metrics and analytics to ensure that the strategies employed are effective. Much of the job is in finding the right metrics to optimize to, and moving the needle on them without harming other metrics you care about. Learning how to measure performance and what serves as a good proxy for overall business goals is a key skill.

Prompt Engineering: Data plays a crucial role in evaluating the effectiveness of prompts. By analyzing the output quality, relevance, and accuracy, engineers make data-driven adjustments to prompts. You can define evaluation benchmarks based on a list of answers you know the prompt should return in certain scenarios, and use that to determine an accuracy score as a measure of performance. For harder to define tasks, you can even use an LLM to grade the output of another LLM, which speeds things up versus waiting for a human to grade responses.

Creativity and innovation

Growth Hacking: Growth hackers often need to think outside the box and come up with creative solutions to drive growth. Innovation is key to finding unique ways to attract and retain customers. Understanding user behavior and needs to tailor strategies that improve user acquisition, retention, and engagement. This mix of creativity and technical ability is powerful, and having cross-functional skills can help you spot opportunities for growth you otherwise might not.

Prompt Engineering: Crafting effective prompts requires creativity to guide the AI in generating useful and engaging responses. Engineers need to innovate to solve complex challenges to automate tasks by bringing in the right context for the AI to generate its response. Crafting prompts that are aligned with the end user's needs and expectations to generate content that resonates with them, so it requires a deep understanding of all aspects of the business and domain that you’re attempting to automate.

Everything is always changing

Growth Hacking: The ultimate goal is to find scalable strategies that can drive significant growth without a proportional increase in resources. These are best found in channels that are new and haven’t been saturated yet. Therefore as a growth hacker it pays to be aware of all of the new mediums and tactics that have yet to be explored, where there’s no real best practice yet. You have to develop the best practice yourself through testing and learning what works, so you can exploit it before the competition. 

Prompt Engineering: Every week there’s something huge released in AI, whether it’s a new model, platform feature, or prompting technique. Things are progressing so quickly that you could spend months struggling to get something to work, completely unaware that there already exists a solution out there that can do it out of the box. I have even seen multiple scientific papers which claimed that AI couldn’t do a task, only to find out that they were using an old model, and the latest model is perfectly capable. 

Leveraging technology to scale

‍

Growth Hacking: The ultimate goal is to find scalable strategies that can drive significant growth without a proportional increase in resources. That’s what the ‘hacking’ in growth hacking means – “building something quickly or testing the boundaries of what can be done”, as defined by Mark Zuckerberg in Facebook’s famous letter to investors. Growth hackers are always using the latest tools, platforms, and technologies to implement and track growth strategies, in many cases building things themselves because there’s no established way to do the thing they’re trying to automate at scale.

Prompt Engineering: Effective prompt engineering enables the scalable generation of content, responses, or actions by AI, allowing businesses to handle larger volumes of tasks with minimal human intervention. You don’t need prompt engineering if you’re just prompting ChatGPT. But if you’re making a prompt template that will be used hundreds of times, or building an AI application that will be run thousands or millions of times, then prompt engineering really begins to matter at scale. Prompt engineers don’t always have to be software engineers, but they have to be willing to run a Jupyter Notebook or call a poorly documented API to use a system before someone has had the time to build a nice UI on top of it and make it accessible to the mainstream.

Concluding Thoughts

Both growth hacking and prompt engineering require a blend of analytical thinking and creative problem-solving. They are iterative processes that rely heavily on data and experimentation to achieve optimal results. They require technical skills and a willingness to dive into the details of poorly documented systems. These core skills are useful in any field you go into, but I think that a background in growth is a particular advantage to working in AI today. 

I see a lot of people who come into the field from a machine learning perspective, and they spend all day fine-tuning their own custom models, when with a few hours spent prompting they could have achieved the same result. The growth hacker mindset is always to use technology to drive outsized returns on effort, rather than building technology for technology's sake, and I think that’s what gives you an advantage over people from more technical backgrounds. 

If you are interested in transitioning to prompt engineering from growth you can learn more in my new book Prompt Engineering for Generative AI, published June 25th by O’Reilly. 

‍