Ben Holliday

Content design as a multiplier for large language models and generative AI

This is another in a series of x24 posts. Expanding on topics from my 2022 book.

A reminder, the idea of multipliers in Multiplied is recognising the tools, approaches, and parts of our work that make more impact possible through digital transformation.

In my bets on digital transformation for 2024 I talked about large language models (LLMs) and generative AI. I described how these might transform how we work and the way we deliver services. I also explored the importance of what generative AI builds from. With LLMs, the underlying content and information the technology uses determines how useful, accurate, and trusted any solution, such as a Chatbot, will be.

In summary:

Generative AI will only be as good as what it works from. This needs investment in areas such as content design, and organisations need better content strategy and governance to support how they deploy LLMs.

To realise the potential of LLMs and generative AI, organisations need to get their content strategy and design in order. As well as the consistency, accuracy and quality of content being produced, this is about how organisations manage information, including governance and assurance models.

Every organisation I’ve ever worked with struggles to maintain and manage content at some level. All organisations depend upon policy and guidance internally, and they most probably publish and communicate information externally, including as part of service delivery. Where we’ve seen a real shift over the past decade is the introduction of content design. As a recognised set of skills and approaches, this has become an integral part of how many services have become clearer, simpler and more accessible. In itself, this has been a big change for many organisations, with content increasingly sitting with delivery, as well as with marketing and comms teams.

Government in the UK has led the way here with a content design profession central to how GOV.UK and other sites like NHS.UK meet user needs. If you’re still not clear at this stage about what content design is, or its value, then I recommend you read Sarah’s book.

I’ve already seen plenty of speculation that AI will be able to replace the need for content design. Also that generative AI will be able to apply style guides to existing or draft content (I’m certain it will be able to)*. To answer this directly, I don’t think anyone’s content design role is going to be replaced anytime soon. My bet is that there’s a much bigger opportunity for how content design will unlock the potential of LLMs and generative AI.

Changes to how we access information and professional expertise

Technologies like GPT-4 give us unprecedented speed and scale in how we can access and synthesis the world’s information. How this type of AI uses content is also connected to an important conversation around the future of how we work.

I’ve found the book The future of the professions, by Richard and Daniel Susskind useful when exploring this topic.

“We anticipate an ‘incremental transformation’ in the way that we produce and distribute expertise in society. This will lead eventually to a dismantling of the traditional professions […] in a ‘technology-based Internet society’, we predict that increasingly capable machines, operating on their own, or with non-specialist users, will take on many of the tasks that have been the historic preserve of the professions.”

The future of the professions – Richard and Daniel Susskind

The central idea here is that the way expertise can now be distributed will disrupt how organisations work. This includes the type of roles we need and how people can be supported to complete a wide range of tasks that might have once belonged to smaller groups of experts or professions.

This links to the idea of co-pilot automation, where personal AIs will work alongside us in the future. They will be able to provide contextual information and guidance in real-time. They will also be able to tailor our individual tasks and workflows, directing our actions accordingly and while automating routine processes.

Future uses of this technology could fundamentally reshape our workplaces. In terms of how we rely on expertise, we could be closer to finding better ways of managing tacit knowledge. Knowledge gained through people’s lived professional experiences but that is harder to capture or distribute more widely. LLMs could be part of the answer here, even helping how we maintain tacit knowledge when people leave our organisations.

The ability to access relevant information in real-time is also a way that services might meet increasing demand. Many organisations already maintain knowledge bases and AI can build on this. In scaling support, this is about enabling how staff can provide advice and manage casework across a wide range of topics. In comparison, organisations might previously have had a reliance on experts to deal with specific types of queries but at the risk of not being able to scale that support over time. LLMs and generative AI gives us the means to scale support and the reach of our services while still working within the human capacity of our organisations.

These technologies could also provide real-time training. Through co-piloting we can continuously develop the experience within our teams, their knowledge and confidence and the types of support they provide.

These ideas take us back to how we manage content and some very real challenges as well as opportunities. Most organisations still suffer from duplicate, out of date, or content not written and maintained to best meet their needs. Content process and governance may be broken, in that, internal ownership of content will be distributed, or may even be unclear. With the challenge of tacit knowledge there’s the question of how we might better document content design decisions. There’s also the challenge of internal policy or guidance not yet being written from a user focused perspective, leaving people filling in the gaps. We should be trying to exclude those gaps through content design work and processes.

Starting with places like our intranets and shared drives we can do the work needed to improve and maintain content. But if we don’t make this investment there’s the very real risk that LLMs will use incorrect, out of date, but available information. This could then lead to errors, with solutions guiding and advising people incorrectly. With the right approach to content discovery and audits we are likely to find contradictory reference points and unclear policy documentation. We will need to consider gaps in information and guidance that already exists. We will also need to make existing information clearer, improving how accessible it is.

Challenging remits and acting as a content platform

With the challenges described here, I’ve been thinking specifically about how content design teams need to increase their remit. This includes the influence they have across all policy, guidance and materials in their organisations. If we see any levels of automation in content design as part of service delivery, then this is where I think we can refocus the content design skills and expertise many organisations already have at their disposal.

As we enter an era of LLMs and use of generative AI all organisations can start to think of themselves as content platforms. This therefore needs to be considered as part of any digital transformation strategy. How you organise as a content platform is about optimising for how people access your information and services, but also how internal work is supported through the distribution of information and expertise. How every organisations approaches this will be key to making the most of digital transformation opportunities.

Your organisation as a content authority

Recognising how and where your organisation is as a content authority is another important part of digital transformation strategy.

Outside of major UK government platforms like GOV.UK and NHS.UK many organisations curate and publish high-quality content. I’m thinking especially here about charities and the third sector.

These organisations act an authority and trusted sources of information in specialist subjects and types of support. For example, the support and information people need following a significant health diagnosis. In the future these organisations will need to ensure they continue to be a content authority in how their information is used by external LLMs and Chatbots. There is also the need to ensure that these types of solutions will inform and signpost people back to them as part of how they scale future services and support.

Now is definitely the time to think more about content strategy and design as a multiplier for digital transformation. At this stage investing strategically in content is going to be important to shaping the future impact of LLMs and generative AI. Do get in touch if you want to chat about this subject and share ideas.

* For further reading, Dom Billington has written this post: Creating a custom GPT to help with content design. It explores some of the limitations and considerations of using ChatGPT to automate and support content design work.

This is my blog where I’ve been writing for 18 years. You can follow all of my posts by subscribing to this RSS feed. You can also find me on Bluesky, less frequently now on X (formally Twitter), and on LinkedIn.