Applying Generative AI
in Marketing
The ChatGPT Earthquake
The release of ChatGPT in November 2022 was a watershed moment. Because it was free, accessible, easy to use, and often produced surprisingly good results, it went viral and made generative AI a mainstream topic. Most recognized that it would be disruptive. Within a month, it forced educators to grapple with how to adjust their assignments and grading.
The rapid disruption in education is just the tip of the iceberg. Marketing is among many areas that will be greatly affected. The ability to create personalized content at scale will be the most significant impact, but change will be broader than that. Generative AI will be used in most parts of modern marketing, including to generate marketing plans, calendars and content, support audience and competitor research, and engage communities.
Marketing leaders must quickly understand how to use and integrate this technology to deliver on rising expectations for efficiency. The marketers on their teams will need to learn how to stay relevant in a landscape where automation plays a much larger role.
This paper by BrandOps explores the ways generative AI will change how marketing is done. We’ll also discuss how to manage this change and offer strategies for using generative AI in ways that minimize disruption while maximizing its transformative potential.
Imagining the Future with Generative AI
Before we consider how to apply generative AI today, it’s useful (and fun) to imagine some longer-term possibilities. Here are a few questions to “prompt” (pun intended) your thinking:
- If content can be generated in real-time, why would it be created before it’s requested?
- If a web visitor can share their preferences, could the entire site they see be uniquely generated to appeal to them? Will that replace today’s more limited personalization?
- If a buyer can use an AI at Google to get full answers, would they click to your site or just chat with the AI that has read your site, your competitors’ sites, and all analysis and reviews to learn about your offerings?
- Will Google only generate text responses, or will they use generative AI to create personalized audio and video directly on their own site?
Consider a future product discovery scenario that takes advantage of generative AI:
Jamie needs new financial software to support her growing business. She asks Google, “What’s the best financial planning software for a 500 person industrial supplier that manufactures in Vietnam and sells in the US and Germany?”
Instead of a list of links, Google responds with a video narrated by a voice Jamie recognizes as a favorite actor. The video describes Jamie’s business and her likely requirements to confirm it understands what she’s after. It describes the three options that Jamie is most likely to select.
The video goes on to show clips of YouTube demos of those options. It also has clips from video testimonials from review sites. It shows a table of strengths and weaknesses, prioritized by Jamie’s needs. It illustrates the growth rate of each provider and notes that one company had a recent spate of negative employee reviews.
Jamie asks a follow up question about which product best handles potential expansion into Asia. Google continues the dialog with a recommendation based on this new requirement.
The underlying tech to support this is available now. It will take some time to assemble the pieces and tune the experience, but it will probably arrive incrementally, just as search has evolved over time. Regardless of exactly how it plays out, it’s clear that AI’s ability to analyze and produce content will force marketers to change how they operate.
In some ways, succeeding in the environment described in that scenario may seem like a natural progression. Brands already cater content to search engines. It’s more complicated though. Today’s search engines try to highlight content that’s useful for humans. That dynamic may change if the engine itself consumes the content, analyzes it, and presents it back in a new form. It’s not yet clear how marketers will succeed in that environment. Buckle up.
AI in Marketing is Already Here
AI already plays a significant role in marketing. An early example of its use was the introduction of lookalike audiences by Facebook in 2013. Rather than manually segmenting audiences using demographics or interests, machine learning is used to consider more attributes about users and their behavior. Ads are then targeted to people most similar to a given audience.
A more recent example is Google’s Performance Max campaigns which were introduced in 2021. They demonstrate a broader use of AI in marketing. Per Google, this tech “allocate[s] budgets dynamically to the highest performing channels, including YouTube, Display, Search, Discover, Gmail, and Maps …”
Beyond eliminating the need for a marketer to manually allocate budgets across channels, Performance Max creates ads using a brand’s assets. If assets it needs for some channels aren’t provided, it generates them, including videos. Agencies that were used to manually control all of this, and being able to explain the workings of campaigns, have had to quickly adapt.
Before Performance Max though, the core technology used in today’s ChatGPT was made available as GPT-3. Businesses launched in 2020 and 2021 to make this AI text generation available to marketers. By 2022, tens of thousands had used these services to generate blog ideas, long-form text, ads, social posts, and other marketing content.
As you’ve likely seen, generative AI isn’t just for text. In 2022, offerings that use natural language to drive image generation emerged. Many marketers already use this technology, but some have been cautious due to uncertainty around copyright issues.
While AI has been infiltrating the marketing industry for over a decade, the release of ChatGPT brought renewed attention to the potential of AI in marketing, particularly in the realm of content generation. The technology crystallized recent advancements in natural language understanding and generation, and is pushing marketing leaders to re-evaluate their strategies and workflows to properly integrate and use the technology.
The Generative AI Difference
While we referred to artificial intelligence broadly in the prior section, this paper focuses on generative AI. This is a type of AI that is used to create new content, such as text, images, or videos. Generative AI leverages large language models, like GPT-3. These text generation models are trained on massive amounts of data, allowing them to understand and generate human-like language, or use language to guide the creation of images and videos.
Generative AI differs from other forms of AI that have been used in marketing. Most prior work has used what are referred to as “discriminative models.” Those models don’t generate content. They are used to classify data or predict outcomes based on existing data. Both forms of AI are useful, but generative AI is a more recent advance and will likely have more impact on marketing in the near future.
What Can Generative AI Do?
We already touched on some of the common uses of generative AI in marketing, such as writing blogs, social posts, CTAs, and ad copy. Tools like ChatGPT rapidly provide useful responses to prompts like, “Write a blog post about the need for financial consolidation and close software in mid-sized companies that are preparing to go public.”
If you haven’t done something like that already with ChatGPT, stop reading and try it. When you see an output that can be improved, tell ChatGPT to do what you’re thinking. This dialog model makes the tool incredibly helpful as an assistant.
Text generation is powerful, but as we touched on earlier, generative AI can do more. Image generation based on natural language prompts is already reducing the need for human artists in some contexts and accelerating the work of those artists who are open to collaborating with the technology. The generation of video from text is not yet as advanced, but rapid progress is being made.
These content creation capabilities are core to how generative AI can be used in marketing, but there are some less obvious capabilities of the technology that are important to understand, including:
Summarize - The large language models used for generative AI can absorb, analyze, and summarize large amounts of data from multiple sources. This can include customer reviews, surveys, social media posts, sales calls, or competitor marketing. This, at a minimum, increases the performance of human analysts. BrandOps is doing this to increase the value of its broad set of marketing data.
Personalize - Generative AI can take into account a user's individual preferences and behaviors in order to create content that is more likely to resonate with them. Marketers may start with personalizing content to specific roles/personas, but more fine-grained personalization based on knowledge of an individual is possible.
Stylize - Generative AI can adjust the point of view, style, or the voice of content. This includes writing in particular styles or even using cognitive biases to persuade audiences. For example, if you used the prompt above to write a blog, try a prompt like, “Rewrite as a VP of finance who went through an IPO.” This ability to stylize to maintain a brand voice when using AI.
Repurpose - Marketers often need to create new content based on content that exists. Generative AI can create the base content, such as a blog post, and then repurpose it into other formats such as an FAQ or a LinkedIn post to promote the new content. In ChatGPT, you can simply provide the content and prompt it to, “create a LinkedIn post.”
Research - Generative AI has the ability to retrieve facts from the content it was trained on which includes most web content. That makes it valuable in identifying relevant facts, statistics, or trends to use in the content you want it to produce. It’s useful to prompt the AI to also cite the source of its findings and confirm the information as sometimes it gets things wrong.
Act - Generative AI can also simplify interacting with marketing systems to retrieve data or execute marketing tasks. For example, a marketer may pose a question that causes a generative AI to retrieve, analyze, and summarize data from a marketing analytics tool. Or, a marketer may use natural language to both create a blog post and have it published on their site.
This is a powerful technology that can generate a wide range of marketing content, taking into account individual preferences, brand voice, and even cognitive biases. Additionally, it can be used to take action by connecting with other systems and automating tasks.
How Will Marketing Work Change?
The most immediate effect of generative AI will be helping teams create more content. That’s highly valuable since most teams are resource constrained. Generative AI automates or accelerates many tasks that people do today. The tech may act as an assistant, co-author, or primary author. In that latter case, writers shift to being editors.
The shift to co-authoring or editing is significant, but a bigger change will be in the type of work marketers do. Let’s consider those changes, roughly in order of when they’ll occur:
Guiding AI: For generative AI to consistently produce content using a brand’s voice, it’s important to provide a set of “pre-prompts” to set context prior to requesting content. The pre-prompts guide the AI and may include descriptions of your tone of voice, mission and values statements, do and don’t statements, and examples. Marketers will need to create this guidance and share it with their team. They’ll also need to review outputs and update the guidance as the technology evolves and they learn more about it.
Tuning Base Assets: We previously referenced Google’s automated generation of search, display, and video ads. Meta and Microsoft have similar technology. These ads are generated using base assets provided by marketers. The engines create the actual ads and find which are most effective. As generative AI advances, this will be the general approach. Marketers will spend less time creating finished ads or other content and instead will tune the assets the AI uses to generate ads and other messages that are highly personalized.
Monitoring Generated Mentions: Marketers currently monitor media and social channels for brand and product mentions. They also track placement in search engine results. As generative AI produces content for prospects on third-party sites, marketers will need to know when their offerings are referenced and find ways to grow these new earned AI mentions. This is a very new area, but one that provides both a great opportunity if a team can understand how to increase their visibility and a great risk if their visibility diminishes during this transition.
Envisioning Opportunities: Marketers will need to change from a focus on execution and set aside time to explore what generative AI technology makes possible that was previously infeasible. For example, it may have been cost-prohibitive to create a series of microsites to boost a brand’s search visibility. Generative AI has already made this much faster, easier, and lower cost. It’s important to recognize these opportunities sooner than later to stay competitive.
Creating Experiences: If prospects are able to learn about solutions through AIs hosted on other sites, they will have less need to visit a brand’s site. Similarly as third-party info and review sites become less used, B2B marketers will lose intent signals they rely on for much of their work. All of this likely means marketers will need to shift their focus from conveying information (which AIs will provide) to creating engaging or entertaining experiences that attract their audiences and allow for the collection of first-party data.
Getting Started With Examples
There’s no better way to quickly learn about generative AI than to walk through some practical examples. The examples below assume you’re using a tool like ChatGPT and that provide some context before the example prompt.
To provide context, you may state information about your target audience, your brand or product, and the desired tone or style of the output. As you get more advanced, providing sample text, such as a prior version of the content or similar content from competitors, can also be helpful in setting the context for a session.
Whitepaper Creation
Start by creating an outline of the paper. Try a prompt like this:
Create an outline for a whitepaper on digital transformation in finance targeted at VPs and CFOs. Emphasize the value of real-time data visualization, process automation, machine learning, and artificial intelligence.
You may request changes to the outline or simply follow up with: Write section 2.
Again, you might request changes or just take the content to use the output as a draft and start editing.
Content Planning
Creating a content plan requires more specifics on what should be included, such as the channels, the resources available, and what to highlight. A prompt like this provides a useful first draft.
Create the Q2 content plan for [your company]. Include topics for five social posts per week, one blog per week, one video every four weeks, and one whitepaper every eight weeks. Detail the plan by week. Highlight general brand info, our annual conference, a new integration with Microsoft 365, and updated scenario planning functionality.
Video Scripting
Creating a simple brand video script is easy, especially if information about your brand was on your website when the AI was trained. You can likely use this prompt to start:
Write a script for a corporate video that highlights [your company]'s values and mission.
Of course, you can also provide context before the prompt if the information is more recent.
Other examples
There are dozens of uses, with almost every marketing task able to benefit from an assistant like ChatGPT. Providing context and crafting prompts can be a challenge though. Sharing what works well among your team helps everyone quickly ramp up.
Here are a few more examples to experiment with:
- Create an FAQ from this blog [followed by blog text]
- Create an email invite and three social posts to promote a webinar on [date] about [topic] with [presenters]
- Create a LinkedIn post to promote this blog post [followed by blog text]
- Create a series of Instagram captions and hashtags for this product [followed by product description]
- Create meta tags for this web page [followed by page content]
- Write a competitive intelligence report comparing us to [list of brands]
Smartly Adopting Generative AI
Generative AI has the potential to revolutionize the way marketing leaders approach their work. With the ability to create content, images, and even entire campaigns, it is a technology that can't be ignored. Here are some things to consider as you begin to integrate the technology into your workflows.
Prepare your team: To get the benefits of generative AI, it's crucial to prepare your team for the change. Start by stating your goals and what you hope to achieve with the technology. Acknowledge those who focus on the flaws and limitations of the technology and ask them to keep an open mind as everyone learns. Assign homework like the examples above so everyone gets hands-on experience. Share learnings about what does and doesn’t work. Create a change management program to ensure a smooth transition.
Establish policies: As you gain experience, create a framework for when to use AI. Be clear about what work is best done mostly by AI, where you expect AI assistance to be used, and what will be done primarily by people. Also set policies about when a human review is required before generated content is used. Ensure your team understands the processes and builds in ways to be sure the new approach is followed.
Create an onboarding package: Develop an onboarding package for your new virtual assistants. This should include information you’d provide to any new marketer such as the company's values, mission, and goals, brand guidelines, your personas, key value propositions, and other background information.
Know your data: Understand the quality of your data and how it can be accessed to consider if and when to integrate new AI tools with your data sources. Identify where the risk lies and take steps to mitigate it. For example, data about what content works for particular customer segments is generally lower risk compared to data about individuals that might be in a CDP. Most teams will need to expand their data to succeed with generative AI, including adding the comprehensive marketing analytics available in BrandOps.
Meet new vendors: Marketers are so flooded with vendor requests that most have put on blinders. It’s important to open those up around this area though. Meet with new vendors and explore the different options available to you. Learn about the latest advancements in generative AI and how they can benefit your marketing efforts.
Give your team access to tools: Provide your team with access to the necessary tools and resources to get the most out of generative AI. This can include training programs, software, and conferences.
Run A/B tests: Conduct A/B tests of human vs. AI work to understand where the technology excels and where it falls short. Use this information to make informed decisions about how to best incorporate generative AI into your marketing strategy. This can help you get the most out of the technology and ensure you're getting the desired results.
Managing Risks
Plenty will go wrong as the use of generative AI grows, but that doesn’t mean it won’t be valuable or that marketers should be afraid to use the technology. It’s just important to be thoughtful about it.
Generative AI can create content and images that could damage a brand's reputation and consumer trust. One recent example was a generated social post referencing a historical event that was highly inappropriate. To avoid the problems, have humans review content before it’s published. That’s still a much more efficient process than human generation.
Another risk is that generative AI outputs may include biases that were present in the data used to train the model. For instance, when generating text or images about people in particular roles or professions, the AI is likely to use stereotypes. Humans will need to overcome those by intentionally directing the AI to be inclusive.
Current AI models are built using a wide range of publicly available content. In many cases, the generation relies heavily on this content. At this time, the degree to which copyrighted material may be used in training and the degree to which generated content must be unique in order to qualify for copyright is in dispute. Marketers should use commercial services that assume legal liability for use of generated content.
Finally, there is a potential for job displacement since generative AI can automate some of the tasks previously done by humans. Leaders need to consider how to manage the risk of talented marketers preferring to work in environments that rely less on AI. Providing clarity on your plans for when and how it will be used reduces anxiety that would otherwise impair performance.
The BrandOps View
BrandOps is a Marketing Performance platform that is well-positioned to support the use of generative AI. The platform offers valuable data that can be used by generative AI to summarize your marketing activity, results, and opportunities for improvement. This includes data on your competitive landscape and important changes, which can be crucial for staying ahead in today's fast-paced marketing landscape.
The crucial significance of BrandOps for teams utilizing generative AI cannot be overemphasized. Without access to critical data, generative AI models are left to simply guess and fail to enhance results. To generate top-performing marketing assets, it is imperative that generative AI models are equipped with information about tried and tested methods, successful marketing assets, the overall marketing strategy, market data, and competitor insights. BrandOps serves as the solution by compiling this information and making it readily accessible for generative AI models, ensuring they can perform at their maximum potential.
Moreover, a frustration faced by users of generative AI is the “amnesia effect" where each time they work with a virtual assistant, it starts over. BrandOps is creating solutions for teams using generative AI that retain data about context, including outputs that were previously produced. This helps avoid the amnesia effect and the production of repetitive content. The BrandOps team has years of experience with natural language technology and continues to build solutions to support and use generative AI.
Let's Go
Generative AI has the potential to revolutionize the way organizations approach marketing. Generative AI will increase the efficiency and effectiveness of marketing teams by automating the creation of personalized and engaging content.
There will be a significant amount of change to absorb including how marketers execute current tasks and how they handle a range of new tasks that arise as a result of the transformation. To implement generative AI in your marketing efforts, start by identifying specific use cases and goals for the technology and take into account the recommendations in this paper. Use the examples we reviewed to get started now and quickly gain a strong understanding of the capabilities and limitations of generative AI models.
Consider partnering with solution providers and other experts in the field to ensure the successful implementation and integration of generative AI into your marketing strategy. Additionally, create a plan to measure the success of the implementation and continuously improve the use of generative AI in your marketing efforts.
