In this episode of the Cherryleaf Podcast, we look at what Techcomm problems can generative AI solve.
Technical writers today face numerous challenges that can hinder productivity and impact content quality. From managing high volumes of documentation to keeping pace with rapid product changes, it takes great effort to produce and maintain helpful, accurate, and engaging technical content. This is where AI comes in.
We provide an overview of the common pain points technical writers experience and demonstrate how generative AI can effectively address many of these issues. From automating repetitive tasks to enhancing writing quality, AI-powered tools offer new capabilities that will transform the field of technical communication.
We discuss key applications like boilerplate text generation, text summarisation, grammar corrections, writing style adaptability, and more. With the right human-AI collaboration, technical writers can be more efficient, consistent, and creative.
This is the Cherryleaf podcast. Hello again. Welcome to the Cherryleaf podcast. We’re going to go back and look at generative AI again in the context of technical writing, and this time look at what problems could generative AI solve for technical authors, or what’s called technical writers in America.
I’d like to take you back to the late 1980s or early 1990s when desktop computers, the PC and laptops were introduced and it gave this opportunity to take a personal computer out of the office and take it home to show somebody’s relatives to your parents and the like. And one of the questions would be, “Well, that’s very nice, but what can it do for me?” You’ll see from programmes at that time, it really boiled down to recipes, perhaps it might help somebody with their accounting, personal finances, but often there wasn’t a great deal that a laptop or personal computer in those days in terms of practical use rather than just games that it could do for somebody at home.
And to an extent, we’re at an equivalent stage with generative AI today. When people look at the technology, they say “Ohh, that’s very nice.” But for me, as a technical author, what can it do? It might be that somebody struggles and is in the same situation. Maybe saying recipes or the equivalent of that for today. However, there are a number of problems that generative AI can solve today, and it’s inevitable that generative AI is going to get better and better and solve more and more problems that we face.
So let’s go through and look at some of the problems that technical authors face today and where generative AI for some of those could help.
So what are some of the common problems faced by technical authors? Well, one is the amount of work, the volume of content that technical authors need to create or manage. And linked in with that is often there are repetitive tasks that still have to be done manually, for example extracting information from word documents and PDFs. That work can still today involve a lot of copying and pasting from one document to another. Many technical communicators have to wade through a lot of meeting notes and information to find the gems that they need for the user content to extract it from them.
Yeah, where you’ve got more than one person contributing content, there can be challenges of maintaining consistency, particularly over structure, style and tone. There can be a need to update the information you’ve started to write and to update legacy documentation, and that can often be due to changes in product names that mean that that has to be changed across all of the documentation set or changes in the technical requirements to run a particular product, so there are time pressures and consistency challenges.
There can also be the challenge of meeting the diverse user needs. You might have a product with different types of users – the pro and the like – version beginner users and advanced users; users in different international countries. The need to meet the needs of regulatory bodies as well.
Another common challenge is integrating with other teams – integrating with development who are producing the product or other departments like training and marketing and support. And you might have related to that challenges in getting your draughts reviewed by subject matter experts, getting code samples from developers, and you might feel that what you’re doing is isolated from the rest of the organisation.
And another challenge is increasing or improving the value of what’s delivered – what’s delivered today in a knowledge base or a user guide or a help file in many ways is similar to what was produced and delivered 10-20 years ago. So there’s this challenge of how can we create content that’s better for end users that helps them not get stuck? Can it be more personalised? Can it be customised to their needs? Can it be more reactive and responsive to help them?
So let’s look at how generative AI can help with some or all of these challenges that technical communicators face.
So #1 is the role of generative AI in automating repetitive tasks. You can now upload a collection of documents to a chat bot and ask it to extract information from that. So it might be that there’s product specifications and you can get those extracted for multiple products. Or you can get a transcript from a subject matter expert, somebody who’s done a video of how to use a product or walk through, and you can ask the chat bot to take that transcript and convert that into a set of instructions to a first draught quality.
If you’ve got lots of meetings, one of the common features within chatbots is summarising information so you can ask it to condense lengthy material into concise bullet points, and that might be suitable for background or overview sections.
Another feature of large language models and chat bots is that it can do what’s called autocompletion. So you can start writing a sentence and ask it to complete it again. That’s useful if you’re looking at things like definitions or overviews where it can predict what would come next. It’s also quite useful if you’re looking at writing code snippets for things like API documentation or SDKs, and you can start to write the code snippets and ask it to complete the code, the commands, the parameters, and so on, and it should be able to do that without typos that you might if you’re doing it yourself, might slip in.
And there are ways of creating boilerplates or templates where you set up standard sections and then ask the AI to complete different sections or content for certain sections, or when it’s doing the conversion of transcripts to enter the information into the appropriate sections.
And you can also ask it to review content and identify where content might have been duplicated so you can consolidate that information. Those are just a couple of ways in which you can use it today to just get rid of the drudgery, some of the repetition that you might have to do, or you might pass on to an intern to do.
Another situation is when you’re actually writing content and getting generative AI tools to help you at that stage. Although tools do have grammar and spell check, with generative AI, you can have a conversation, a flow going back, asking it to review for specific errors or for tone issues, and asking it to change the tone or style of the content so that it’s more consistent. And that could help people where English isn’t their first language.
If you’ve got content that’s been written by different people and there are variations in tone and wording, again you can instruct it in your style, in your standards, and then ask it to adapt those documents so that they conform to your preferred style and tone.
There’s also features like proactive assistance, where you can ask or use the AI to anticipate and answer questions that might come up from users, and you can also ask it to expand a collection of bullet points or notes into more comprehensive paragraphs, again to a decent first draft.
And one of the other features that we’re starting to see emerge is that it can look at code, look at documentation sets, and generate documentation from the code comments, from the documentation strings or from the code, interpret the code itself. Again, it won’t provide a complete manual, but it will help save you time and having written the content, it can help you polish it, to review it.
So within the tools and within generative AI, you can ask it to review your content, to check for particular errors, you can ask it to optimise your content for search engine ranking, for search engine optimization. And you can ask it to apply categories and tags, which again can help with content management and also with your Google rankings.
But one of the biggest benefits is the opportunity it gives us to create better outputs rather than what we provide today. So a chatbot can’t generate a user manual for you. It needs source content in the large language model to answer the questions that users still have, so there’s still going to be a need for technical writers to create that source content that goes into the large language model and that is used when the chat bot provides the answers to users’ questions.
But having a chat bot means that users can have a conversation. They can ask questions they’re not certain about, particular answers they get back. They can ask for clarification. They can ask for something to be explained in simpler terms or using a different analogy. The lines between training and support and knowledge bases become much more blurred. The end result is much more personalised assistance for users to help them when they get stuck. It’s much more user-centric.
There’s also the concept of dynamic FAQs, where the AI’s role is to respond to user queries and update FAQs in real time if the user doesn’t understand the text. Somebody can now ask “Can you draw me a picture that illustrates what you mean?” And with the latest versions of ChatGPT, it can create a diagram, an image for the end user to help them understand in that way.
So we now have this tool that gives us with the ability to write sentences, instructions, to ask it to do something that can give us the abilities to be a computer programmer, to be a graphics illustrator, to be a video producer that we can all do this just by being able to ask a question, being clear in our intent.
So where we’re looking to find ways to resolve those common problems that technical authors face – dealing with the volume of content that’s out there, the speed of change, managing consistency, keeping the documentation updated, meeting the needs of different users, integrating with other departments – then really the first call, the first place to go to look for the solution, the ways, the tools that can help us resolve those problems should be AI.
So if we go back to our analogy about personal computers and initially just being tools to provide a way of recording recipes, as we know that changed rapidly. We got the internet, we were able to provide information on our credit card securely so we could start to buy things. And now you have people of all ages using the computer, people doing their family trees, going back to the 1500s, designing cross stitch patterns, conversing with relatives in Australia, all manner of different things these days.
It seems inevitable that generative AI is going to have a transformative impact in technical writing. And it’s one that we should embrace. It’s going to be able to give us far more powers than we have today in capabilities of things that we can do.
We were reflecting earlier in the week about the delegates that have been taking our e-learning course on using generative AI in technical writing. They’re the pioneers. They’re the ones that are leading the community, starting to learn and apply the techniques, capabilities, the promise of generative AI in technical writing.
So as a course of action, what we would encourage you to do is embrace this change. And if you want to acquire and improve your skills, to consider – and of course it will give you hands on experience, it will help you understand the concepts, the promise of generative AI, and how it can be used within technical writing – we have further details on that either at cherryleaf.teachable.com, that’s our e-learning platform, or if you go to our website cherryleaf.com and go to the training section, you’ll find information on the course and the link to the e-learning platform that way.
So relatively short episode this time for the podcast, but I hope you found it useful. We always welcome feedback and comments – emails probably the best way – firstname.lastname@example.org. It just leaves me to say thank you for listening.