The LavaCon Conference was held in San Diego this year – fortuitously only a few blocks away from and a few days after the MadWorld Conference which I also attended. LavaCon is focused on Content Strategy and covers developing content strategies, as well as the tools and technologies used to deliver those strategies. This year the key topic for the conference was the use of AI tools in and around Content Strategy.
The power of the current generative AI tools has already made it cheap to create BAD content that you can’t trust. This will mean that from now on it will become increasingly expensive to find GOOD content that you can trust.
This is going to be a challenge for companies to ensure they are making it as easy as possible for their customers to find quality, trustworthy content. This will mean using AI tools to help create that content, as well as to help distribute that content. So, content strategies are going to become increasingly necessary as well as well-produced structured content.
AI in content strategy – a sense of exploration with a touch of chaos
Surprisingly, there was no sense of doom and gloom that AI will be taking over from technical writers any time soon. Rather there was a sense of excitement that AI will be a powerful tool to help technical writers … if they can learn to harness it.
There was a sense of exploration with a touch of chaos as tool vendors have quickly pulled together solutions using this first generation of AI tools, but with no one really knowing where this is all going yet. However, there is a growing consensus around the strengths and weaknesses of current AI tools, and therefore how to deploy them.
Sarah O’Keefe reflected on how current AI tools are great for pattern recognition in both structured and unstructured content, creating summaries of content for both text and video, and they do all of that amazingly fast. However, she pointed out that AI tools fall down on accuracy – AI tools “hallucinate” – on privacy – you don’t know where your content is going or where it came from – and on compliance.
So with this in mind, Noz Urbina encouraged us to look beyond the most well-known generative capabilities of tools such as ChatGPT and Dall-e, and look to other AI tools with different capabilities that can be applied to different tasks. We should look at the whole content and customer lifecycles to identify tasks where AI tools can be safely deployed.
Want to find out more?
Content creators vs content consumers
It is important to consider AI tools from the point of view of both the content creator and the content consumer. If AI tools can summarise a video to create a short description thereby saving the content creator time, then it can also summarise a video for the content consumer. We can see these different perspectives in the approaches taken by the application vendors. The vendors for content creation tools have added features to support the technical writers, for example, Heretto, MadCap, Adobe, Writer, etc., whereas the content publishing platforms have added features to simplify content consumption, for example, Zoomin, Fluid Topics, etc.
Structured content vs unstructured content
All those at the conference agreed that although AI is good at understanding unstructured content, it does a much better job when given structured content. AI tools seek to understand the content so they can recognise patterns to then categorise or summarise the content. But they do that based on the data sets they have been trained on. Providing structured content with rich metadata to an AI tool will yield better results for content consumers. And if you also provide the AI with the associated structure and taxonomy, you get even better results.
The converse is also true – if content creators provide an AI tool with a defined structure, for example, a taxonomy or xml schema, the AI is able to create structured content. In addition, if you have enriched your taxonomy with concept definitions and related terms such as synonyms, preferential terms, and deprecated terms, then the AI tool can better classify the content and also apply the correct terms.
Esther Yoon demonstrated how her team used an enriched taxonomy to help an AI tool to apply tags to their legacy content before importing it to a new content portal. They initially used the AI to help them identify terms from the legacy content as a starting point for creating their taxonomy. And of course, her team still needed to review the initial list, refine it, and then enrich, but the AI tool dramatically reduced the time needed.
So AI tools can help you create a taxonomy, apply that taxonomy to unstructured content, and then use that taxonomy to create a better experience for the content consumers. A winning combination.
Demonstrating value – toot your horn
In his key note speech, Scott Able encouraged us to “toot your horn” – a theme echoed by others such as Maura Moran and Kevin Nichols who emphasised linking content directly to business objectives. These could be financial objectives, for example, improved efficiency, increased revenue, faster time to revenue, or non-financial, for example, meeting compliance requirements, mitigating risks (technical or operational), or building brand reputation. So to demonstrate the value of the content to the business, you need to first understand how the content delivers value to the business, and then let everybody in the business know that it is adding value. This is where Content Strategy comes in.
During my talk, I described how a key part of building an effective content strategy is to understand what the content needs to achieve, including business objectives. You can then design the content to meet those needs. Then, by tracking the performance of specific content, you can evaluate if the content is achieving the intended goals, and therefore demonstrate the value of the content.
David Hoare provided some good examples of how his team at ServiceNow has measured how their content performed against business goals such as product engagement, new customers, and customer retention. All of these metrics were higher in customers whose users had higher engagement with the product documentation.
Want to find out more?
Customer experience vs content experience
Kevin Nichols reminded us that most people experience a company’s brand through its content. You might think this is because of marketing or advertising content, but actually more than half of the content a person consumes from a product company is technical content, such as technical documentation, learning content, or support content. So when we consider the customer experience, we can’t separate it from the content experience.
To identify how content can support business objectives, we need to understand the consumer journey, and identify what content customers need at each stage to help them achieve their goals. Jo Ward gave the example of how she and her team at Salesforce look at each stage of a customer’s journey in terms of Modes: Start, Do, Learn, Solve – what is the user trying to achieve.
Quality of outcomes vs quality of content
To demonstrate value, you need to have a way to quantify value. Meghan Gillhooly provided a great anecdote about how at a previous company, management were convinced they had the single, perfect metric to measure content value – spoiler, they didn’t. Similarly, Torsten Machert demonstrated how a simple quality metric like the Flesch-Kincaid score can provide a high score for a sentence even if that sentence doesn’t actually make sense.
When assessing content value, there are two sides you need to consider: the quality of the content, and the quality of the outcomes.
For example, David Hoare and his team at ServiceNow measure the quality of their content using the following dimensions: Documentation debt, Writing quality, Content architecture, Multimedia, Localization, Accessibility, and also CSAT. In addition, as mentioned previously, they measure the quality of outcomes by looking at product engagement, new customers, and customer retention.
Whatever you measure, David reminded us that you need to provide context for the number you are looking at to give it meaning. For example, is it going up or down, which direction should it be going, and how does it compare against others.
So how was it …
I had long wanted to attend a LavaCon, but couldn’t justify the cost of the travel when there were other options closer to home. Having relocated to Mexcio to set up a new 3di office, I was excited to have the chance to finally attend. And I wasn’t disappointed.
There was a great energy at the conference powered by enthusiasm and opportunity. Enthusiasm for the subject and the opportunities opened up by the new knowledge we were learning.
