Fastr Blog

10 Generative AI Prompting Tips for Ecommerce

Written by Jenny Holton | Apr 1, 2024 5:26:54 PM

There is something to be said about the comfort of familiarity or fear of the unknown, but any wise ecommerce leader knows that stagnation is business suicide. Being innovative means being experimental and unafraid to not only do more and do better, but also to learn. However, when a technology tidal wave like AI floods the scene, even the most confidently creative and exploratory ecommerce teams may hesitate – wondering how to dive in, or which toe to use to test these new waters. 

 

So, in the spirit of Fastr's emphasis on and enablement of ecommerce brands' creative ideation, experimentation, and learning, we're sharing some tips to help you level-up your prompt engineering skills so you can better collaborate with and leverage the power of generative AI. Try these 10 tips, explore uncharted waters, and be fearless in your consistent pursuit of ecommerce innovation and growth.

 

 

1. Let's Chat

 

Also referred to as "iterative prompting", conversational prompting is not only an easy way for amateur prompt engineers to get the hang of collaborating with AI, but it also makes it easier for AI tools to produce more intuitive responses. 

 

It's hard to think of AI like a human – and maybe you don't want to – but to leverage generative AI's full potential, try conversing with it as you would a colleague and work in steps. And don't be afraid to circle back – "reframing" is another prompting technique you can use to subtly tweak one of your previous prompts in order to have AI subtly tweak its output. It's a great way to learn what prompts cause what responses and to have a variety of options generated.

 

Try these:

  • Instead of piling your request into a single prompt, provide initial information to start, follow up with more context as you go.
  • Ask questions such as, "Do you have enough information to complete the task?"
  • Reframe your prompts, get reframed responses. For example "What are some ways to increase AOV on my ecommerce site?" will elicit a different response from AI than "Can you list some techniques to encourage online shoppers to purchase 2 or more items?" 

 

 

2. Flip the Script

 

Most often, it's humans asking AI questions, but why not turn the tables and have AI ask you the questions? Some prompt engineers call this "flipped interactions" or "flipped prompt engineering" and it can be done at any point in your series of prompts.

 

Examples:

  • Start your prompt "conversation" by saying "For every prompt instruction I give, please generate 2 questions that I can answer to give you more context or information." – and later, instruct your generative AI tool to include all of your responses for the task.

 

 

3. Set the Tone

 

Knowing and using your brand's tone of voice is important for any content piece your brand publishes, so it's equally as important for AI to do the same. You can include directions in your prompts regarding the tone of voice you'd like the Large Language Model (LLM) to use – the more descriptive, the better. Tone of voice descriptors could include adjectives, audience reading level, and even formatting directions like sentence structure.

 

Example:

  • "Use an authoritative and knowledgeable tone of voice with highly intelligent word choices and keep the sentences short and concise." 
  • Bonus Tip: Have an ultra-specific tone of voice in mind – a particular someone? Maybe it's Barack Obama, Joe Rogan, Amy Schumer, or Elon Musk? Whoever it is, prompt the LLM to use their tone.

 

 

4. Never Say "Never"

 

Providing limitations is extremely helpful. More parameters means clearer guidelines in which the AI can work. However, if you're going to ask your generative AI tool not to do something, try including the ideal alternative – what it should do instead. This builds upon the context that AI has to work with and means you'll get more fine-tuned responses. 

 

Try This:

  • Instead of "Don't use complex sentence structures" try, "Don't use complex sentence structures and do write at a 7th grade reading level." 

 

 

5. Role Play

 

If you've ever participated in theatre or seen either a phenomenal or terrible movie, you probably know how important both character knowledge and character development are. Just as you would in a game of "Who Am I" charades, a LLM will perform much better if given context on the role it is intended to play. In addition to prompting AI to complete a specific task (the "what"), provide information on the identity you want it to assume (the "who") and include details like job title, experience, what they're responsible for, or what they're trying to accomplish with the task at hand.

 

Example:

  • Clarify the role (the "who"): "Imagine you are a fashion editor with 20 years of experience working in women's luxury apparel. You are a leading voice on capsule wardrobes and how to combine clothing items to create outfits."
  • Then outline the task (the "what"): "You are responsible for writing a 750-word guide outlining 10 staple articles of clothing that are required to create countless outfit combinations for timeless, casual style in the spring season." 

 

 

6. Show and Tell

 

Ask any elementary school-aged child to recall stories from "Show and Tell" and you'll quickly realize they retain an impressive amount of information. Human brains of any age process more and better with cues that reach more of our senses. And although it's not a human, AI works similarly. Alongside any context you share through your prompts, include reference examples to demonstrate and further cement what exactly you want AI to generate. 

 

Try This:

  • If you're prompting AI to use a certain format, persona, tone of voice, or aesthetic design style, copy and paste an example of what you would like to imitate into a prompt (i.e. a banner image you used last year in your holiday campaign that received a higher-than-average number of clicks, a URL showing your PDP layout, a social media post that's the length and tone you're aiming to recreate, etc.).

 

 

7. Speak the Language

 

Although we have grown accustomed to having translation tools at our fingertips, AI is now another. But if you've ever heard those horror stories of people using a translation tool to generate a tattoo design, you're well aware that languages don't precisely mirror one another. So, if you're using generative AI to help you speak another language, play it safe by providing context around idioms, metaphors, and meaning. 

 

Example:

  • If you're asking a LLM to translate "Let's put this on the back burner until next week." be sure to include context such as "Keep in mind this is a work metaphor for pausing further work on a project until next week".

 

 

8. Spitball

 

Like we previously mentioned, it's helpful to think of generative AI as a colleague – a collaborator. And like a good teammate, it's not only able to help generate a task, but can also help with idea generation. So, if you're drawing a blank or your creativity tank is running on empty, prompt AI to assist you with some inspiration. 

 

Examples:

  • "Give me 3 catchy headline ideas for this webpage." – and include your URL in the prompt.
  • "Suggest 5 hashtags for this social media post." – and paste your copy into the prompt.
  • "Give me 10 blog post ideas for interior design based on current trends."

 

 

9. Ask the Audience

 

A lot of ecommerce strategy is about knowing your target customer base and how they operate while shopping in the marketplace where your brand resides. But understanding the audience and market requires a lot of data collection and analysis – and therefore, a lot of time. Employ AI to share insights to leverage in your SEO efforts, your product descriptions, your understanding of your industry or competition, or even have it predict how your content will be received by your audience.

 

Try these:

  • "Show me a list of keywords someone might use when searching online for a [product type]." 
  • "What are the top 10 questions people ask about [company name]'s [product name]?" 
  • "Give me 5 ideas for a meta title and meta description for a [product type] product page." 
  • "What are 10 questions someone might have after reading this" – and paste your content into the prompt.

 

 

10. Co-Editor

 

Already finished drafting a new content piece? Tasked with revising an already-published page? Maybe it's a product description, an "About" page on your company website, or a social media post. Whatever it is, you could spend the hour, day, week, or longer combing through it, you could hand it over to a colleague to look through and provide feedback, or you could save yourself – and your team – the time by sharing it with an AI tool and requesting a review

 

Examples:

  • "Highlight spelling and grammar errors and suggest corrections for each."
  • "Shorten this to 1000 words."
  • "Provide feedback on the tone of voice used in this."
  • "Create a 280-character summary of this for a post on X/Twitter."

 

 

How are we leveraging the power of AI?

At Fastr, we believe in creative freedom, bypassing dev bottlenecks, experimentation, consistent growth, and – with the Fastr Frontend platform and FastrAI – empowering ecommerce teams to make these beliefs their every day reality.

 

Fastr Frontend is purpose-built for the teams behind innovative ecommerce brands – enabling them to design, deploy, and test shopping experiences with unparalleled flexibility and ease.

 

Supercharging your productivity, our in-platform AI co-pilot, FastrAI, can be customized to your brand voice, tech stack, and asset library and is ready to assist with both high-priority and previously time-consuming tasks including experience generation, accessibility checks and corrections, and experimentation.

 

Craving the agility to do more, learn more, engage more, and sell more? Let's chat about equipping your team with enterprise-grade AI capabilities and the power to create the optimized ecommerce site you've always wanted – faster and more easily than you thought possible.