Kelly Dell, Fluid Topics
May 15, 2024

Following the explosion of Generative AI (GenAI) across industries, it’s no wonder that many companies are launching new projects centered on this technology. A staggering 85% of executives plan to increase their spending on AI and GenAI in 2024, yet 66% of them are less than satisfied with their progress and subsequent business opportunities.

Typically, IT teams direct and oversee these projects. However, as organizations navigate these new initiatives, they are quickly finding the benefits of ensuring content professionals play a leadership role. This shift is critical as GenAI projects tend to rely heavily on the expertise of tech writers. Therefore, to maximize the impact and ROI of GenAI projects, companies need to start with the right team for the job: content professionals who will transform content into dynamic, user-centric solutions.

With cross-team collaboration in mind, we’ve outlined three tips to get the most out of your GenAI projects from the start.

#1: Don’t Ignore the Complexity of Content and the Value of Content Professionals

Many AI projects are led by IT teams that are proficient in programming, but struggle with content integration. For example, when building an AI-powered chatbot, IT teams often rely on random data sets to test their technology.

While this approach works at first glance, the initial success quickly turns to user disappointment. That’s because, while the chatbot can execute simple queries, it struggles to perform more complex tasks. However, the core problem here is not a technological failure, but a problem with content integration. Resolving this and optimizing your AI is worthwhile for any business. In the case of chatbots, experienced professionals report they can complete tasks 40% faster with GenAI chatbots and increase their output quality by 18%.

The key to reaching this level of success is recognizing that content is the foundation for any AI-driven project. Without the intervention and leadership of content professionals, AI tools are subject to the threats of hallucinations, inaccuracies, privacy issues, and ethical dilemmas.

Of course, this isn’t to say that technical writers are the only ones who know content. At most organizations, several teams are familiar with existing content and can help identify incorrect information generated by AI. By gathering input and feedback from subject matter experts (SMEs), your projects are more likely to succeed, using comprehensive content and generating accurate, contextual answers.

Determining the resources and expertise needed for a project is essential for integrating all team members. By leveraging inputs from SMEs and content professionals, you’ll be well on your way to completing a successful AI project.

#2: Upgrade your Prompt Engineering with Internal Collaborations

In the early 90s, search engines were new and contained basic capabilities. Most people didn’t understand how to best use them. Over time, both users and search technology advanced. The general population discovered how to search effectively, and search engines learned how to interpret queries for more accurate results.

Now, 30 years later, a similar evolution is taking place with GenAI. Prompt engineers are pioneering this discipline, aiming to perfect their queries for optimal results. However, this is only the beginning. Both humans and robots will continue to learn and improve — us at prompting and them at interpreting our requests.

Crafting effective prompts requires defining and refining instructions to produce a beneficial GenAI solution. Ideal results typically refer to specific, accurate, and contextually relevant answers. Understanding how to do this is essential for working with artificial intelligence and natural language processing (NLP).

Good prompts are specific, direct, and include commands. When used to answer questions about or search through technical documentation, prompts should include personas. For example, a prompt might contain text that says, “As a [role], I’m trying to do [x]”. To make this work, companies need content professionals to lend their expertise to prompt engineers. Technical writers know how to ask and answer the right questions:

  • How do I translate my understanding of context into a series of prompts that generate the right answer?
  • How do I back-engineer prompts to get to the right content?
  • How do I train the AI to deliver appropriate responses?

IT teams have a big role to play when developing new GenAI tools. However, they need to collaborate with other departments to ensure their systems provide accurate, context-aware, and ethical responses.

#3: Optimize Metadata to Improve Generative AI’s Relevance

The quality of data directly impacts the results of your AI projects. DITA-structured content offers pre-built-in intelligence. When content is structured, modular, purpose-driven, and includes metadata, AI will understand its meaning. There’s no confusion or guessing. In contrast, NLP can attempt to infer the intent of unstructured content, but it is both resource-intensive and less accurate.

Integrating structure into the content adds context, purpose, and relevance. Teams can further enhance data by including descriptions of audience and relevance in the content architecture. Together, this metadata clarifies the context of the content, reducing the chance of AI hallucinations. By feeding the algorithm with high-quality data, you’re likely to get equal-quality results.

However, the metadata alone isn’t enough. The GenAI project then needs to connect to new silos of information to work as efficiently as possible. To put this in place, the project team needs to identify additional metadata and possibly build a taxonomy. Other technologies can then use this information to identify correct answers. In short, the goal is to reconnect each of the silos by establishing an additional metadata layer over the top of the existing layer of metadata. This ensures consistency across the various silos.

With these elements in place, your GenAI project will be well-positioned to understand and generate relevant responses. The applications for this are wide, and approaches like retrieval augmented generation -aka RAG- are revolutionary for helping customer support, maintenance, and sales teams quickly access key information.

Key Takeaways

Looking for the SparkNotes version of our tips to help you launch successful GenAI projects? We’ve got you covered!

  • IT-led AI projects prioritize tech development and performance, often leaving out other stakeholders. Involving content professionals in this process is essential; without them, AI projects may suffer from hallucinations, inaccuracy, privacy issues, and ethical dilemmas.
  • IT teams tend to collect random sets when they need content for their projects. This is typically sufficient for testing the technology and tools you have selected, but it can rapidly fail to answer even simple questions, not mentioning complex queries.
  • Metadata is essential to improving Generative AI’s relevance. It brings context and helps reduce hallucinations.
  • Prompt engineering extends beyond the IT team. Good prompts connect to user personas — “as a [role], I’m trying to do [x]”. This level of prompting requires the skills of content professionals.

The Bottom Line: What’s the Outlook for the Future?

The IT team is essential for building and integrating the technical elements of AI systems. However, alongside them, the perspective and skills of content professionals are necessary for taking your system to the next level. They know the content value, can design additional metadata schemes for siloed content, and can help downstream operations by evaluating the accuracy and relevance of the model.

As we explore new horizons in the field of GenAI, it’s crucial to acknowledge the impact of teamwork. Partnerships between content experts and the IT department create more meaningful, personalized, and contextualized experiences.

 

Note: This article was co-authored by the teams at Fluid Topics and Dita Strategies.