5 Proven Strategies to Maximize your Marketing ROI with Automation and AI
- AI as your new favorite intern
- AI as a powerful extension
- Start small, document and scale
- Automation as the special sauce
- The all-important human touch
What’s the big deal with AI and automation?
It’s not as if it’s new, right? From Google’s RankBrain, YouTube recommendations and spell checkers to email filters, Siri and Alexa, these applications have been part of our daily lives for more than half a decade. Designed to rank information to provide better experiences, the first AI wave was all about handcrafted knowledge with limited to no self-learning capabilities.
Enter the second (tidal)wave: Generative AI and Large Language Models (LLMs) which are giving rise to a series of “what’s next” engines, most famously ChatGPT and Bard. Generating unique human-like responses based on the data it’s trained on, these applications are, quite literally, transforming every aspect of the world as we know it.
The numbers alone justify the hype and anyone thinking AI will be a short-lived fad is sorely mistaken. In a mere five days after its November 2022 release to the public, ChatGPT reached the magic million user mark, earning it the title of fastest growing technology in history. To put this in perspective, ChatGPT reached the 100 million user mark in the same time it took Instagram (the second fastest growing app in history) to reach one million. It’s unprecedented times and the pace of change is only accelerating.
#1 Use it as your new favorite intern
Think of AI as an enthusiastic intern who has unlimited time, access to an infinite library of information and can give a near instant response in almost any language. Now imagine the productivity benefits if everyone in your organization had an intern like that who sat next to them, eagerly waiting to help lighten the load.
On the other hand, much like a fresh-out-of-college student, it lacks real-word experience and is often only as good as its mentor, requiring hands-on coaching, very defined briefs and bite sized tasks. Also never forget that it believes everything it reads and was primarily trained on old information.
Our advice? Treat it like you would any other intern:
- Never provide it with sensitive information
- Make sure you give it clear, step-by-step briefs
- Check everything (and we do mean everything) and give clear feedback on how to improve
- Discover what tasks it does best and use it for that
- Set up robust Quality Assurance (QA) processes (known as Hallucination Mitigation Frameworks that scrutinize every AI output to confirm it hasn’t gone “rogue” by providing inaccurate, irrelevant or potentially damaging information) and ensure nothing leaves your desk without a healthy dose of human intervention
#2 Use it as a powerful extension
Stay up to date, and take advantage of, the latest features with OpenAI Plus that allows you to, much like Chrome and other browsers, add a wide variety of extensions to help you do certain tasks.. By changing your settings, you can use the Beta Web Browsing to, for example:
- Access more recent, relevant and latest information: Enable AI to search the web, visit websites and retrieve latest information, or act as a real-time assistant (effectively bypassing the limitations associated with the model being trained off old information ).
- Install plugins: Explore the wide and ever-growing range of available plugins such as Zapier for seamless integration with external data sources, There’s An AI For That to find the AI tools for your specific needs and Bramework for SEO analysis.
#3 Start small, document and scale
When it comes to implementing AI and automation in your marketing workflow, the only way to learn is by doing. Start small, document your findings and scale up to larger projects over time. Here’s an example workflow for SEO optimization.
- Starting small: Keyword structure. Review a website to find the overall keyword site structure and expand on the categories. Your prompts could look something like this:
- Go to this logistics website and find the top SEO keyword categories
- Expand on this list with SEO keyword subcategories
- Put this list into a data table organized by category
- Expand on sea freight keywords
- Going bigger: SEO Writing. After defining your strategy, use AI to write a first draft for a new SEO optimized page. Prompts could include:
- Write 500 words of unique SEO copy from this top-ranking website
- Write the post in this tone of this brand
- Include the brand’s USPs
- What are the key Q&As for this topic
- And even bigger: A Nordic translation. After adapting and expanding on this first draft, translate the approved version into any Nordic language. Prompts to experiment with:
- Translate this text into Swedish
- Optimize the article for SEO using these keywords
- Replace the information about Oslo with info about the port of Stockholm
As part of a larger SEO workflow, identifying these clearly defined, bite-sized tasks, then experimenting with the prompts to obtain the feedback you need, is an excellent way to achieve consistent AI results. Once you’ve cracked the code, it’s important that all employees document their tasks, collectively working towards an armory of institutional knowledge that can be replicated and perfected (as opposed to starting from scratch every time). Consider using ChatGPT’s new feature that allows you to share your chats with others, making it easier to spread knowledge.
Over time, efficiencies across the whole workflow will be greater than the sum of these AI improved tasks, empowering significant productivity and performance gains across the entire organization.
#4 Automation is the special sauce
If AI is the brains of the operation – enabling us to be significantly more efficient and greatly broadening our skill base – automation is the muscle that allows us to replicate ourselves with well defined, rules-based processes.
Case study: A dash of special sauce for a leading dairy client
We have been using the power-combination of AI + Automation for many of our clients, including a leading dairy company. Here’s what we did.
The challenge: Find a more efficient way to scale search presence and attract more customers without increasing media spend.
The strategy: Evaluate the competitive and performance landscape of each prioritized keyword, only bid on keywords that the client doesn’t rank well for in Google and automate the bidding strategy for best results.
The solution: We developed an extensive keyword map, then automated the process of monitoring the client’s top-ranking keywords and removing these bids in Google Ads. This saved both time and valuable ad spend which could be allocated to other, more competitive search keywords to gain new traffic.
The results (over a three-month period)
#5 The all-important human touch
While many headlines prove how impressively smart AI can be, an equal number show that it can be just as...not so smart. Original thought, critical thinking and reasoning, knowing fact from fiction, experience, cultural or racial sensitivity, these are all very human traits that AI cannot replicate, at least not with 100% accuracy or efficiency. This makes the human touch more important than ever before, and your every AI action must be accompanied by a rigorous QA process (i.e. Hallucination Mitigation Frameworks) that scrutinizes every aspect of its output. Failing to do so can have serious, and often embarrassing, consequences as a New York lawyer recently found out.
ChatGPT, in particular:
- Is only as smart as the data in its LLM which includes vast amounts of inaccurate, sometimes harmful, material from the web
- Cannot produce new ideas and is essentially a “master-plagiarizer”
- Has limited knowledge about anything that happened before September 2021
- Is multilingual but monocultural, in other words, it speaks Swedish but thinks like an American
- Is biased, perhaps even slightly racist, and doesn’t know true from false
- Can be pretty bad at maths
Get ready for the third AI wave
Since the launch of Chat GPT-4 in March 2023, the capabilities of AI have vastly improved in terms of accuracy and performance. Becoming multi-modal (as opposed to a text-engine only) is arguably one of the most exciting developments where text prompts can now also generate images and video. This is opening up huge possibilities for designers and creative teams to start seeing the productivity benefits with AI enabled tools like Midjourney, Adobe and Photoshop.
This AI revolution is only eight months old and the next six look equally exciting (check out our recent Nordic Morning Talks where our team explored the topic in more detail). We expect Chat GPT-5 to land early-2024, alongside Google’s Generative Search Experience which is currently being tested and tentatively predicted to roll out in December 2023. Watch this space... we’re not in Kansas anymore.
At Nordic Morning, we actively use these five strategies to speed up time-consuming processes, test new ideas, accelerate prototypes and, perhaps most importantly, free up our time for critical, strategic and creative thinking. Want to do the same? Get in touch with our team of experts to find out how.