Recently we explored how AI is helping small, time-strapped and budget-conscious marketing teams work more efficiently. Today we're taking a deeper dive into two of the most practical areas: AI tools and prompt engineering.

‍

Both are critical when it comes to really seeing how useful AI can be when used in the right way, for the right things.

‍

Remember: these tools should be used to help enhance your creativity and efficiency, not to replace the skill of a marketer or copywriter. The marketers who are using AI most successfully are prompting it with real human insights from their own customer research.

‍

Let’s get into it.

AI tools to experiment with

ChatGPT: Your research assistant 🕵️‍♀️

‍

Good for:

‍

➡️ Deep research, idea generation, analysing large datasets, creating first drafts and acting as a strategic thinking partner.

➡️ Supporting analysis projects e.g. interview transcripts, survey responses, campaign data

➡️ Creating preliminary competitive landscape overviews

➡️ Generating content ideas based on source materials

‍

Top Tip: There is also a research specific GPT - ConsensusGPT which specialises in complex academic and scientific research - which may be useful if you operate in that field!

‍

Claude: Your content sparring partner ✍️

‍

Claude specialises in content creation across various formats and styles. Claude is particularly strong at working with long documents, strategy work and refining written content.

‍

Good for:

‍

➡️ Summarising several documents, inputs, sources of information to help form a structure (for blogs, emails, etc) you can then mould

➡️ Brainstorming content ideas based on source materials

➡️ Refining content you’ve already written - quickly adapting style, tone or structure based on specific guidance

➡️ Helping overcome writer's block

‍

Watch outs:

‍

➡️ Requires editing… Don’t expect to be able to copy & paste from Claude into your email editor!

➡️ Will need very specific guidance to meaningfully apply tone of voice

➡️ Not ideal for highly technical or specialised content without expert input

‍

Granola: Your meeting companion

‍

Granola is an AI-powered meeting assistant that quietly captures notes, summaries and action points, allowing you to stay focused on the conversation rather than typing everything down.

‍

Best for:

‍

➡️ Capturing meeting notes and action points automatically

➡️ Summarising customer interviews, discovery calls and internal meetings

➡️ Identifying recurring themes and customer insights across conversations

➡️ Creating searchable meeting notes that are easy to revisit and share

‍

Watch outs:

‍

➡️ Always let participants know if AI meeting notes are being used and follow your organisation's privacy policies.

➡️ AI-generated notes are a great starting point, but it's still worth reviewing key decisions and action points before sharing.

➡️ Granola works best as a note-taking companion rather than a replacement for actively listening and engaging in the conversation.

‍

Opus Clip: Your video editing wizard 🎥

‍

Opus clip helps turn long-form videos into engaging, shareable short-form content.

‍

Best For:

‍

➡️ Creating several social media clips from longer-form content

➡️ Repurposing webinars

➡️ Generating quick marketing videos

➡️ Repurposing podcast interviews into social media snippets

‍

Watch outs:

‍

➡️ The selection of clips can sometimes miss nuanced human context so be prepared for every suggestion to not be right

➡️ You’ll need to review all of the suggestions and sometimes make adjustments

➡️ Works best with clear, well-produced source content - so make sure the source content is good!

‍

Prompt engineering

Prompt engineering is still important, but today's AI models also respond well to iterative conversations. Rather than trying to create the perfect prompt first time, treat AI like a collaborator—refine, question and build on previous outputs.

‍

We’ve been experimenting with the RACE framework after listening to an inspiring talk by Joanna Wiebe at a growth conference recently.

‍

‘Role’ is about matching the AI's 'character' to the context of your task E.g. you might give AI the role of a ‘subject matter expert’, a ‘researcher’, a ‘conversion copywriter’. Doing this can help get more in-depth and appropriate responses.

‍

'Action' is clearly defining the task you want the AI to perform e.g. ‘write a comprehensive report’ or ‘create 10 intriguing email subject lines’.

‍

'Context' is about sharing as much detail as you can within your prompt that guides the AI's response. E.g. specific details about your target audience, the tone you want to follow or the specific objectives of the output.

‍

'Execute' is your instruction to the AI to begin working!

‍

Inspired by an example Joanna shared, here’s how we might use this framework to analyse 1000+ free text survey responses (which doing manually would take hours!).

‍

“You are a conversion copywriter, expert in decision-making, persuasion and conversion rate optimisation. You are also a highly skilled investigative expert. You want to better understand what motivates people to sign up for [product name] in order to use those messages on a website landing page to encourage more people to sign up in the future. You’ve collected feedback from customers in the attached CSV. Review the responses to the question ‘What was going on in your life that brought you to [product website] today?”. After reviewing the responses to this question identify the motivators to sign up for [product name] and create categories to group them. For each category, provide the number of responses within that category, list them in order so the category with the most responses is listed first, and give a few examples of exact responses under each.”

‍

Depending on the results of that initial search you may want to ask it a few more questions. E.g. if there is one standout category, ask it to “divide [category] into subcategories to get more granular”.

‍

In this survey data example, you could also ask it to divide the responses by another field e.g. Job title and then ask again for it to provide categories of responses by job title. Hopefully you get the idea… It's about using AI to quickly give you useful ways of looking at (in this case) data that you can then work with. It means you can move more quickly, based on real research, with more confidence.

‍

AI is evolving incredibly quickly, so our toolkit and prompting techniques continue to evolve too. The important thing isn't using every new tool - it's finding the ones that genuinely improve the way your team works.

‍

P.S You can catch up or rewatch any of the Sustainable Growth sessions here.