An Industrial Marketer's Guide to Generative AI 

Why Generative AI?

 

How did we end up here, anyway?

While artificial intelligence (AI) for content development, or generative AI, is not new to the scene, it is rapidly gaining traction as a way to augment content creation, streamline processes, and increase efficiency. 

ChatGPT Adoption Explosion

OpenAI's ChatGPT exploded on the scene, gaining more than 100 million users within the first two months of its launch in late 2022. Until this point, large language models (LLMs) were used behind the scenes by developers to create software applications. ChatGPT offered an easy, conversational (and free!) user experience not limited by any one type of task. Users continue to uncover new ways to apply this (and other similar LLM technology), from coding to research, but one popular use is to aide in the content development process.

The widespread hype over ChatGPT has inspired marketers to evaluate other more purpose-built generative AI tools to augment their marketing processes. Many of these tools launched before ChatGPT but are now enjoying more attention and consideration. 

Another driving factor for the adoption of generative AI is....

Google Updates

Google's most recent set of algorithm updates have prioritized user experience and unique content that answers questions.

Think of Google Snippets -- the drop-down Q&A boxes that show up when you search a specific question. 

The same updates have penalized duplicate content -- which was once a strategy used to win search traffic. 

So, where there is a demand for high volumes of unique content that can be published on websites to hopefully win a top ten ranking, AI tools provide a solution.

In theory, the promise of generative AI tools is that a user can enter a few keywords, click a button, and generate paragraphs upon paragraphs of useful, relevant content.

Generative AI Tools for Technical Audiences

 

Do generative AI tools apply to B2B content development for technical audiences?

Of course. Any new tools should be considered carefully, and with technical audiences in mind. We will be most successful if we can learn how to strategically use these tools as they evolve.

Take CRM tools as an example:

• Do we have a digital contact database built carefully over time, by providing helpful, gated content? Yes.
• Does having thousands of email addresses mean we send emails to our database daily? Definitely not. 

Our CRM database is a tool that we use to meaningfully communicate with our audiences.

That said, technical audiences have some nuances that are important to remember when evaluating new tools.

 

Search Habits of Technical Audiences

 

Research shows that while searching online, 43% of engineers filter through at least 5 pages of search results to find the information they're looking for. In the process, about half of engineers actively avoid paid ads. 

When searching for work-related information using a search engine, how many pages of results are you willing to review before you select a page to visit or restart your search? (n = 697)

Page 1 Google rankings aren't the Holy Grail for technical B2B companies like they may be for B2C companies.

 

Trustworthiness of Content Authors

 

Content attributed to engineering experts, technical product managers, and industry analysts are considered the most trustworthy from an engineer’s perspective. 

Technical content can be written and published by a variety of professionals. To what extent do you trust the following authors? 

Engineers are looking for content written by other engineers -- so any content created, whether by an engineer, an internal marketing department, agency, or AI tool, should carry the level of technical detail that a conversation with an engineer would bring.

 

Sales Cycles for Technical Audiences

 

B2B companies have longer sales cycles. Companies that sell $200k+ contracts and nurture leads over months (or years) need content that gradually guides prospects through their purchase decision. This content needs to be both technical and proprietary -- it should be specific to the company's expertise and products as the prospect gets closer to making a decision.

What Generative AI Tools are Currently Available?

 

While it seems new tools are added daily, the standouts for us at the moment are ChatGPT, Claude, Jasper, CoWrite, MagicWrite, and HyperWrite.

It is important to understand the difference between LLMs and apps built on them. LLMs are large language models -- models built and trained on mass amounts of data. ChatGPT and Claude give you direct access to the GPT4 and Anthropic LLMs respectively, where Jasper, Writer, and the others sit on top of LLMs. 

 

LLMs with chat functionality (ChatGPT, Claude, Bard) 

The chat functionality in these LLMs allow you to ask questions and have them answered.

Unlike Jasper and Cowrite described below, LLMs  don't stick to a specific format. Instead, they rely on prompts to generate specific content. Users input a question -- anywhere from "How does a PXI chassis work?" to "Can you write a poem about PXI?"

Here's an excerpt from that poem:

Compact PCI Extensions for Instrumentation,
A modern marvel, a technical innovation,
A platform that's reliable, sturdy, and strong,
A tool for engineers, a place they belong.

Entertaining and it does, indeed, rhyme.

• Cost: Free, though chatGPT does have paid ($20/month) versions which allow for file inputs, collaboration, and enterprise-level privacy protections.

 

Jasper AI

 

Jasper AI is a content generation tool that can be used to create blog posts, web content, emails, and ads. Jasper also touts content planing, SEO, and image generation tools

• Cost*: $59/month
• Trial: 5 days, requires a credit card
  
  

CoWrite

 

CoWrite is a content generation tool by Writer. Use CoWrite to generate headlines, outlines, and product descriptions. CoWrite also has a custom style guide tool, where organizations can input style rules and create content within the bounds of your defined style. 

• Cost: $18/month
• Trial: 14-days, no credit card required

 

 

Canva Magic Write

Canva Magic Write is a content generation tool within the design platform, Canva. This tool is free for existing Canva Pro users. 

• Cost: $12.99/month

 

HyperWrite

HyperWrite has a browser extension to help with predictive text and an application to rewrite and generate new content from existing content.

 

• Cost: Free for 15 generations (what we experimented with below)/month, 15 rewrites/day, and 500 typeaheads/day. Plans with more generation for $20-45/month.

Other AI-Powered Content Generators

 

In addition to the generative AI tools mentioned, we also evaluated Rytr, Twinword, and WordLift, but felt the output from Writer and Jasper was superior.

That said, these are specifically content generators. Other AI tools also abound, from content planning and optimization tools like MarketMuse and DemandJump to SEO-specific tools like Surfer. These tools overlap in different ways, and we're seeing connections between them. Hubspot is an investor in Jasper and has integrations with Surfer.

As a reminder that there are many options- Brandfolder, Surfer, GrowthBar, Article Forge, CopyAI, Bard, WriteSonic, Ink for All, Peppertype, and Frase...are all sitting on the to-be-evaluated list.

*Note that pricing for these tools changes often (as in...weekly for some). 

TREW Marketing's Evaluation of Generative AI Tools

 

Taking the five tools outlined above (Jasper, CoWrite, ChatGPT, Magic Write, and HyperWrite), we evaluated AI capabilities for technical content development.

Our goal was to answer a few pressing questions. Can these tools make content development faster or easier by:

1. Creating technical content we would otherwise take the time to write
2. Creating better options for headlines or outlines
3. Creating derivative content from our original inputs (build a three-blog series from a created white paper)

 

 

Here are a few examples of what we input into an AI tool, what resulted, and what we learned.

 

6 Examples of TREW's Generative AI Tool Capability Experiments

 

Using CoWrite to Generate Headline Options for a Blog Post

 

TREW input:

Optical sensors are a key technology in autonomous vehicles. Their design and integration determines a vehicle’s ability to perceive the outside world and react accordingly. In this paper, we’ll deconstruct three major optics engineering challenges in the autonomous vehicle space and show how multiphysics simulation technology can help engineers address them.

Headlines Generated:

• The Three Major Optics Challenges in Autonomous Vehicles
• How Multiphysics Simulation Can Help Address Optics Challenges
• Key Technology for Autonomous Vehicles
• The Importance of Optical Sensors
• How Optical Sensors Work
• What Determines a Vehicle's Ability to React?
• The Design and Integration of Optical Sensors
• What Are the Benefits of Optical Sensors?
• Why Are Optical Sensors Important?
• What Are the Drawbacks of Optical Sensors?

Takeaways

•  A few of these options are meaningful, though don't generate anything new that wasn't created from the overview
• This was a good list to use as a catalyst for a brainstorm

 

Using Jasper to Generate a Blog Post Outline

 

TREW input: 

Title: Making Electric Vehicle Wireless Charging Systems a Reality

Overview: We explore how wireless charging works, identify the challenges presented by wireless charging of EVs both at a system and a component level, and discuss how [company] components can help address these challenges so that you can make widespread wireless charging a reality.

Outline Generated

• How does wireless charging work?
• What are the challenges of making EVs wireless charging systems a reality?
• How do [company's] help address these challenges?
• What making wireless charging a reality?

Takeaways

• This outline was a simple set of questions that didn't add value to our content generation process
• It didn't make content development faster or easier

 

Using ChatGPT to Generate a Blog Post

 

TREW input: 

Title: Creating a Test System for Longevity

Answers Generated

Modular hardware allows for greater flexibility and scalability in production test systems. It allows for easy replacement and upgrading of individual components, reducing downtime and maintenance costs. Additionally, modular hardware can be reused across different test setups and products, leading to cost savings in the long term. Furthermore, with Modular hardware it is easier to test different configurations and features of a product, improving the overall test coverage.

When creating a test system for new products, several things should be considered:

1. 1. The product's specifications and requirements: The test system should be designed to verify that the product meets these specifications and requirements.
   2. Test coverage: The test system should be able to test all of the product's features and functions to ensure that it is fully functional.
   3. Test efficiency: The test system should be able to test products quickly and accurately, in order to minimize downtime and maximize throughput.
   4. Scalability: The test system should be able to adapt to the production volume and be easily expandable for future products.
   5. Test cost: The test system should be cost-effective and easy to maintain, in order to minimize long-term costs.
   6. Safety: The test system should be designed with safety in mind, to protect both the product and the operators of the test system.
   7. Flexibility: The test system should be adaptable to test different products and configurations, in order to minimize the need for new test systems.
   8. Automation: Consider the possibility of automating the test process to reduce errors and improve efficiency, test planning and execution.
   9. Data Management: The test system should be able to collect, store and analyze data from the tests and have the capability to share it with other systems if required.
   10. Standards: Compliance with industry standards for design, safety, and performance should be considered.

Takeaways

• This was decent basic background information and provided a catalyst to ask some questions in an upcoming SME interview -- for example, "how do your solutions handle data management, or do you integrate with other organizations' data management tools?"
• The content itself was repetitive and unusable as-is

 

Using CoWrite to Generate a Blog Post

 

TREW input: 

Title: Creating a Test System for Longevity

Section Questions:

• • What factors that most affect longevity in a test system?
  • What components of a test system are most at risk for obsolescence or part failure?
  • How do longevity requirements vary across industry applications?
  • How do you best select a test partner to work with?

Blog Post Generated

Instead of pasting the entire post, here's an image of the final technical blog post we created for this client, starting with the AI-generated post.

• • Blue text was original inputs into the AI tool
  • Red text is what we deleted and/or changed
  • Black text is what we kept from the original AI-generated post

Takeaways

• We kept a few overview sentences!
• Using this AI tool actually made the writing process longer -- what the tool generated was distracting. It wasn't technical enough in detail or professional enough in tone

 

Using Canva Magic Write to Product an Overview

 

TREW input:

Why is radar different than an RF sensor?

Content Generated:

Radar is different than an RF (radio frequency) sensor because it uses radio waves to detect objects and measure their distance, velocity, and direction. Radar does this by sending a signal out and then measuring the time it takes for the signal to bounce back. An RF sensor, on the other hand, is used to detect changes in electric or magnetic fields. It is often used to detect the presence of objects, as well as changes in temperature, light, and humidity.

Takeaways:

• This content is usable -- it's a definition/overview, not company specific, and sounds similar to a wikipedia entry.

 

 

Using HyperWrite to Generate a White Paper Abstract

 

TREW input: 

Full, completed, edited white paper >2,500 words

Content Generated

• Three variations of summaries

Takeaways

• This was helpful and the summary could be edited and used for a landing page or technical paper submission for a conference. 
• While the abstract didn't create anything new, that's actually what I was looking for. I didn't want my technical content to be paraphrased (which often leads to inaccuracies), instead, the tool took subheadlines and intro sentences and put them together to form an abstract.
• This requires a strongly written piece of content to start with. 

Pros and Cons of today's AI tools

Our original evaluation set out to answer three questions. Can these tools make content development faster or easier by:

1. Creating technical content we would otherwise take the time to write -- no. The content generated isn't technical or detailed enough to stand alone, or even to be a start for a rough draft.
2. Creating better options for headlines or outlines -- possibly. The headline options generated weren't necessarily better, but they did give some quick options. 
3. Create derivative content from our original inputs (build a three-blog series from a created white paper) -- not yet, though it seems possible. These tools seem to focus on culling existing material from the internet and creating new content, rather than deriving new content from proprietary information.

Though AI tools have become increasingly popular in discussions around generating content, they offer some advantages and disadvantages. 

 

Pros of AI for Content Generation

 

1. Accelerated Definitions or General Content: AI tools can generate generic content quickly and accurately, which saves time and effort for content creators when working on definitions or glossary pages
2. Brainstorm Tool: Generated headlines or outlines can help a writer go from blank page to rough draft. 
3. Research Tool: Similarly, AI tools could be used to research a topic, though because of stated possibilities of inaccuracies, it would be preferable to use vetted sources like industry publications.
   
   

Cons of AI for Content Generation

 

1. Limited Creativity: AI tools are not as creative as humans, and may lack the ability to create content with the same level of originality and imagination. Specifically, AI content generators use a lot of repetition. In example 3 above, eight of the 10 considerations start with "The test system should..." Much of the content created feels like it's written at about a 6th grade reading level.
2. Limited Context: AI tools can be limited in their ability to understand the context of a topic, which can result in inaccuracies or mistakes. We found specific inaccuracies when trying to use a specific company's name. AI content generators seemed to paraphrase company descriptions in ways that made the descriptions of a specific company or product completely inaccurate. Some of the tools seem to “forget” what they already said in earlier sections -- in fact, this must be a known issue as Jasper touts that it reads your past 3,000 characters every time before writing for better context.
3. Lack of Technical Detail: Content generated is general, not specific to a company's expertise or products. Since we know engineers trust content written by other engineers, we must ensure that our technical content is detailed and accurate.
4. Lack of Content Strategy: AI-developed content isn't designed to guide a prospect through a buying journey. It lacks the emotional intelligence to personally connect with customer pain points or use prospect's nuanced language.

While AI tools can be effective in producing text quickly and accurately, they may not be suitable for all types of content, and may require human input and oversight.

Frequently Asked Questions

 

We will continue to answer questions that arise from our team and clients here. Have a question? Complete the form at the bottom of the page.

 

Should we be concerned about plagiarism?

Based on what we've seen, plagiarism doesn't seem to be an issue. Because these tools pull from existing material, they also can check generated text against existing text. 

 

How do generative AI tools handle brand messaging?

Since AI content generators are creating new content, they're often rephrasing messaging. This is something brands need to watch out for -- your audience typically doesn't interact with your brand every day. They need to be reminded of your message, and the best way to do this is to use the same language across content. If AI tools are consistently rephrasing, in using them you risk confusing readers and losing brand identity.

 

How will Google adapt to the mass of content being generated?

Though we can't predict future Google changes, one thing we're waiting to see Google handle is bounce rate -- if companies are producing content to answer simple questions, will websites be penalized for decreased time on site?

We suspect that Google will quickly be able to determine whether content has been written by a human or created through generative AI tools. Too much content generated by AI may actually have a negative impact on a website's SEO scores since recent Google algorithm changes are prioritizing user experience.

 

How does HubSpot's "ChatSpot" AI tool work?

HubSpot launched their alpha version of a new AI tool, chatspot.ai, on March 6. This tool combines generative AI and research capabilities with access to your HubSpot portal, allowing it to generate answers and HubSpot reports based on your queries. Because it also links to your Google Docs, it can generate content topic ideas and content drafts that you can save to a Google Doc, or post directly into HubSpot. 

The tool is very simple to use. Once you connect it to your portal, you open a chat window where you are able to give it a prompt or data request and it will serve up ideas or results in minutes. 

We found the HubSpot connection and reporting to be valuable even in this alpha phase, and expect future releases to help us expand our content research and ideation.

 

Does my company need a corporate AI policy? 

Customers are starting to ask questions. Primarily, they want to know how AI tools affect the product or service you’re delivering to them.

Skepticism around AI tools often comes from deeper concerns that you’re automating what you used to do manually and exposing them to errors or causing them to overpay for something they could now do themselves with an AI tool. 

Making it clear how you think about AI and what tools you use for specific purposes with an AI Policy creates transparency. 

Click here to get started on your AI policy. 

Read TREW Marketing's AI Policy.

 

Conclusion

 

While we are constantly experimenting with AI technologies to drive efficiency and performance across all business functions, we don't believe AI writing tools are currently capable of producing the in-depth content highly technical audiences require.

These tools can aid in the brainstorming and ideation process but should not be relied on as a writing source.

Here’s how we think about AI as a B2B technical content marketing agency:

1. We believe in a human-centered approach to AI that empowers and augments professionals. AI technologies should be assistive, not autonomous. 
2. We believe that humans remain accountable for all decisions and actions, even when assisted by AI. The human must remain in the loop in any AI-assisted applications. 
3. We believe in the critical role of human understanding and experience, and place specific value on the expert knowledge of subject matter experts. It’s the individual knowledge and expertise held internally that differentiates a company.
4. We believe in understanding the limitations and dangers of AI, and considering those factors in all of our decisions and actions. 
5. We believe in an open approach to sharing our AI research, knowledge, ideas, experiences, and processes in order to advance the industry and society.
6. We believe in partnering with organizations and people who share our principles. 

To see how we at TREW Marketing are using AI today, read our complete AI Policy.

Watch the Webinar

 

Improve Generative AI Prompts for B2B Content Development

Looking to get more out of AI tools for content development?

The answer lies in prompting. Quality in. Quality out.

Prompting an AI tool is similar to guiding a dinner party conversation or conducting an interview – the way you ask questions has a great impact on the response you get. 

In 55 minutes, you'll learn to improve your prompting skills and get drastically better, more customized results from AI tools every time.

In this webinar, we cover: 

• Summer 2023 updates to AI tools for content development

• Best practices around which tools to use and what to use them for

• Collaborative practices for companies and agencies using AI tools

• Common prompting mistakes

• How to prompt using ChatGPT

• A live prompting demo

• Reusable templates for prompting

• New AI tool features specifically useful for technical B2B content development

• Live Q&A

 

Listen to the Podcast

 

Generative AI for Industrial Content Development: A ChatGPT, Writer, and Jasper Showdown

The landscape is exploding with generative AI tools for content marketing.

But when it comes to creating B2B marketing content, especially pieces targeted to engineers and technical buyers, are these tools actually helpful?

This is what TREW Marketing's writing team set out to find through research and experimentation.

On this episode of the Content Marketing, Engineered podcast, Wendy Covey, Morgan Norris, Jamie Tokarz, and Hannah Schulze weigh in on what they found to be the most and least beneficial aspects of these tools in their current state. 

 

Generative AI for Marketing- Maintaining Trust and Building an AI Policy

Is your team using AI tools? Here’s everything you need to know to maintain trust and authenticity.

In this episode, we highlight the ever-present need to stay authentic and trustworthy when developing and publishing content – something we still need human oversight for. Morgan Norris also shares the liabilities of using AI tools incorrectly and the direction these tools could take in the future.