The utilization of Generative AI for marketing is helping teams write more emails, ads, blog posts optimized for SEO, website copy, social media posts … the list goes on. In fact, 14% of the respondents to a recent survey about marketing technology adoption were already using Generative AI for marketing use cases.
In these early stages of new tech adoption, it can be difficult to understand the differences between technology providers. For organizations considering them, returns from Gen-AI investments increase if you can align what you want to achieve (your goals) with what the solution you’ve purchased can do (its functionality).
In this blog post, Persado presents a framework for thinking about the kinds of marketing problems Generative AI can solve today. We follow with an overview of the capabilities of some example platforms in the context of that framework. For more on the core capabilities that enable Generative AI to deliver business value, see our executive brief, Generative AI for the Enterprise: The 5 Capabilities Businesses Need for Optimum Impact.
The large language models (LLMs), in which Generative AI operates, fall into the category of natural language processing (NLP). LLMs can craft original content by responding to a prompt with the words that most likely follow, based on millions of similar sentences in the model’s training data. Solutions built on an LLM, like GPT-3 or 4 from OpenAI, can solve a few different types of marketing challenges. Those challenges fall into two broad categories: efficiency and effectiveness.
In marketing, efficiency challenges are about creating enough language. They’re about scale and speed. Do you have all the language you need for your web site, blog posts, social copy, advertisements, etc., and can you update or create more for different channels quickly?
Over the past few months, the discussion around Generative AI for marketing has largely narrowed on the efficiency story. However, that’s not the only way to think about it.
Effectiveness challenges are about driving outputs that are better in terms of results. While efficiency aims for scale, effectiveness aims for quality.
In marketing, for example, more language can divert customers if the language is irrelevant or otherwise doesn’t speak to them in a personal way. More, in this case, can lead to brand fatigue, whereby the customer becomes numb to the message. Generative AI that delivers more effective messages can produce higher rates of customer engagement for marketing.
Generative AI is already proving useful for solving efficiency problems (i.e., generating a lot of content quickly). GPT-3, and now GPT-4, the popular LLMs that third-parties can build on at low-cost, are powerful efficiency tools. Since GPT-3 was trained to be broadly useful, third-parties can leverage the underlying model to build a wide array of solutions. OpenAI recently revealed a list of the applications third-parties have built using the model.
The evolution of Generative AI for marketing will move past efficiency and toward effectiveness as businesses look for competitive advantage and bottom line results. Enterprise users will seek more effectiveness in a number of ways (non-mutually exclusive):
Businesses that tap into more specialized, effectiveness-focused solutions will achieve greater impact.
With that framework in mind, consider the following Generative AI for marketing point solutions and how they could be used:
ChatGPT is a general, all-purpose chatbot launched by OpenAI in November 2022 that uses a large volume of content from the internet as its knowledge base. More than a million people signed up to pilot it within the first few weeks of its launch. Since the original ChatGPT is free, curious consumer end-users can easily access it and start experimenting with it to experience its amazing capabilities first-hand. The results of early experiments have appeared throughout social media and news reports. In one colorful example, a judge in Colombia used ChatGPT to craft his ruling. The chatbot has also raised questions about plagiarism rules on college campuses, while elevating concerns about the tendency of LLMs to make up facts, or to stray into uncomfortable and uncanny conversational terrain.
The current consensus is that ChatGPT is an extremely powerful language generation tool. However, it is not yet clear how enterprises will use it, for what purposes, and to deliver what value. Microsoft, a long-term partner of OpenAI, has confirmed increased investment in the open-source provider (of which it now owns 46% percent), and is integrating ChatGPT, GPT-3, and GPT-4 capabilities across its product suite, including Azure and Microsoft Office 360.
At the time of its launch, ChatGPT was based on GPT-3.5, which uses content from the internet (up to 2021) as its knowledge base. In March 2023, OpenAI launched ChatGPT Plus, based on GPT-4, available on a paid subscription basis.
Since it first launched in November 2022, Microsoft has invested an additional $10 billion in OpenAI and has integrated ChatGPT into its consumer search engine, Bing.
ChatGPT is available for general, self-service use related to summarizing text, conducting research, generating text, and other use cases. With clear prompts, ChatGPT can deliver outputs in a range of formats, including as subject lines, paragraphs, a letter or an email, to name a few. Microsoft has integrated it into search.
Websites, social media, emails, etc.
Currently, adoption appears to steer toward efficiency problems. ChatGPT does not have the ability to self-edit to ensure the text is accurate. Nor does it have built-in intelligence that recognizes how well its language engages readers.
A trio of serial entrepreneurs founded Jasper in January 2021 by in Austin, Texas. Based on the GPT-3 LLM, Jasper focuses on enabling copywriters, founders, and everyone who deals with content to generate more of it, fast—10X “faster” is the promise Jasper makes on its website.
Jasper’s subscription model has attracted a reported 100K+ users and a large influx of venture capital from Y Combinator. Anyone can sign up for a subscription and the tool is easy to use, with helpful videos and tips on writing prompts in a way that enables the AI to produce the outputs you want. Jasper’s reported data on impact focuses on team productivity and the amount of copy they can produce. There is no information on how well the AI-generated messages performed with any target audience.
Jasper is based on GPT-3. Like ChatCPT, the knowledge base is the entire internet (prior to 2021).
The company gained initial traction with individual users. It recently launched Jasper for Business, however. This includes a set of features such as a brand voice element, as well as capabilities to integrate the solution into common business applications.
Jasper focuses on marketing and sales copy, including for SEO, social posts, and videos.
Websites, social media, emails, etc.
Jasper’s brand promise appears to focus on efficiency.
The Persado Motivation AI Platform is a Generative AI for enterprises that delivers language and messages designed to motivate customer engagement and action. Built on a proprietary language model, the Persado knowledge base includes ten years of performance data about billions of marketing messages. Persado is the only Generative AI we’re aware of with the ability to predict how well a marketing message will perform based on years of data on customer engagement. And since it is the only proprietary model for marketing, Persado-generated content is not derivative of other content sources. As a result, it avoids some of the challenges that GPT-based solutions are currently experiencing, as brands ban their use in the enterprise.
Across retail, financial services, travel and other sectors, enterprise brands like Ally Bank, Gap, JPMorgan Chase, and Marks & Spencer leverage Persado to personalize their messaging for specific customer segments and across channels. On average, Persado-generated messages produce 40% higher engagement than a human-generated variant with the same intent.
Learn More about the core enterprise capabilities of the Persado Motivation AI from our white paper: The 5 Capabilities Your Generative AI Needs to Boost Marketing Performance
Persado has a proprietary language model trained on enterprise data.
Yes, the Persado Motivation AI Platform was designed for the scale and performance needs of enterprise marketing, customer experience, and e-commerce teams. Persado AI-generated content motivates customers to engage and act based on decades of digital marketing performance data.
Marketing message performance, multi-channel personalization, as well as customer experience.
All digital channels, including email, IVR, social platforms, text/SMS, and web/e-commerce
Persado provides enterprise brands with somewhat more efficiency and significantly more effectiveness at scale. On average, brands achieve 40% higher performance from Persado-generated messages than they would have achieved with their human-crafted original. The platform realizes these results by personalizing the digital experience with messages that are optimized to perform. It does that by tapping into years of data on the language that engages customers and motivates them to act.
This market space for Generative AI in marketing is growing rapidly. We will continue to refine the consideration framework and update this post with more companies over time—keep checking back!
The post Generative AI for Marketing: How to Compare Different Generative AI Marketing Applications appeared first on Persado.
Podcasts are a fun and easy way to learn about machine learning.
TL;DR We asked o1 to share its thoughts on our recent LNM/LMM post. https://www.artificial-intelligence.show/the-ai-podcast/o1s-thoughts-on-lnms-and-lmms What…
Palantir and Grafana Labs’ Strategic PartnershipIntroductionIn today’s rapidly evolving technological landscape, government agencies face the…
Amazon SageMaker Pipelines includes features that allow you to streamline and automate machine learning (ML)…
When it comes to AI, large language models (LLMs) and machine learning (ML) are taking…
Cohere's Command R7B uses RAG, features a context length of 128K, supports 23 languages and…