Recently our Principal Facilitator and Coach, Carolyn Moore, presented a session to a group of international HR Directors on how to Elevate Your AI Fluency. During the insightful session, one participant observed that learning to prompt more effectively, and still needing to assure the work produced by their business's large language AI model (LLM), was actually not making their work more productive or efficient. Why was this technology - technology that is in the news each week about how it works more effectively at pace than human capabilities - not producing the expected productivity gains it was promoted to create?
In this post, we are going to examine some of the reasons why this is the case - and what you can do about it. Implementing any new technology is a significant business investment, and your business and your function deserve to make an expected return on that investment.
Let's start by getting one thing straight: a Copilot implementation is not an AI-transformation. Yet, sadly, we hear it all the time: we've done our [insert LLM of choice here - Copilot, Claude, ChatGPT etc] implementation, it's job done. IT have ticked the box on AI transformation. Managers actively encourage employees to use it. Internal financial modelling shows how much productivity gain it will create - most often represented as how large the "headcount reduction" will be - sometimes it's implemented early on. Then time passes - 6 months or so, and reality hits: the technology isn't producing the efficiencies expected. Sometimes this manifests in the assumption that people aren't using it, or not using it effectively, so there's a second flurry of frenzied promotion to get people to use it. More time passes, and still there are little to no gains made. In some high profile examples this has led to rehiring strategies, reversing the "headcount reductions" made following the realisation that the technology won't magically make things more efficient on its own.
So, what is actually happening?
There are number of factors that are influencing this situation, and understanding what is going on is the first step towards knowing how to more effectively approach AI transformation.
Low AI Fluency
In most circumstance we have seen, the AI application has been implemented with a fanfare, maybe even a few webinars on how to prompt, and then turned over to employees to make sense of it and use it in their work. This is new technology, requiring new approaches to work, yet very little is being invested by most organisations to ensure their employees have sufficient AI Fluency to use the applications effectively, understand the opportunities and risks it presents, and engage with it in innovative ways that will actually create value. The result of this is that many people use it as an enhanced-Google search or report-formatter. This is low-level, low-value work that is never going to produce return on investment or significant business value.
Cognitive Surrender
The term cognitive surrender has been coined by Gideon Nave in "Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning
and the Rise of Cognitive Surrender" and refers to the process whereby employees give up their decision-making capabilities to the machine. In businesses where the brainpower of their employees is their unique selling point, the risk of cognitive surrender is significant. The desire to maintain critical thinking has been mooted to be behind Goldman Sachs' restrictions on the provision of Claude for their investment floor employees. The risks of cognitive surrender have also played out publicly with large consulting companies being caught out producing reports for large government clients with hallucinated citations, statistics and conclusions. Cognitive surrender is a manifestation of the outcomes of unfettered use and experimentation, where there is insufficient guidance to employee users on how to best engage the technology. In other words: it is what happens when people do not understand how to use the technology to its best effect. The result being that it produces unexpected outcomes and unintended consequences, that directly impact business - all of which are preventable.
Individual Use and Lack of Scale
Most AI LLMs deployed in businesses are being used by individual employees to do individual tasks. Research from Capgemini has identified that just 7% of all AI use is for teamwork. By definition this lack of scale restricts the return on investment of the deployment of the LLM. The lack of employee AI fluency - as mentioned above - means that most use is not high-value use. That takes real, intentional, strategic design and implementation to plan and deploy effectively. To achieve these results, employees need to develop new skills in innovation, experimentation, customer/client experience, to understand how this technology can add real value. To gain real value from the technology, the use-cases need to be elevated from individual-use to functional- and organisational-level use: embedding the capabilities of the LLM into the flow of work.
Leadership Intentionality and Design
Put simply: LLMs are not magic beans. It is not a matter of sprinkling the organisation with this magical system and discovering a golden egg. AI and digital transformation require different approaches to leadership: a deeper level of listening and intentionality that enables true understanding of customers, employees, shareholders and societal needs. This then needs to translate into intentional systemic and organisational redesign, that embeds LLMs - with human teams - to produce new ways of working and new business models that create value. Without this intentionality and design process, business adopt remains small, individualised, flounders, engagement is patchy and limited (if any) real value is created. Leaders need to be enabled to lead credibly through AI transformation by developing new leadership skills (including their own level of AI fluency). Leadership development and coaching of leaders to be able to effectively navigate ambiguity, engage multi-disciplinary teams to innovate and deploy new business models, and evolve the skills of their people effectively, will bring the most significant return on investment for any LLM deployment.
Productivity Does Not Mean Redundancy
Most measures of LLM efficiency relates to how many human employees it can displace. This is entirely the wrong measure, especially for businesses in the knowledge economy. Productivity is not a simple measure of reduced headcount. LLMs present significant opportunities to create value through innovation, to enhance jobs by automating repeatable tasks, and to even create more engagement and flourishing teams, with human-centered design approaches.
The good news is that all of these issues are (relatively) easy to address.
It starts with the basic principle of increasing the AI fluency for everyone in an organisation - not just a few early adopters or "Ambassadors". Everyone needs to understand:
- What LLMs are and how they work
- What value they bring when they are used effectively
- What risks they present when they are not used effectively
- What effective use looks like (and what is not effective use)
- How to think about LLMs and what they can do within the flow of work
- How to accelerate LLM and AI adoption through customer-focused innovation
This is what we do. We help organisations and leaders build the workplaces of the future, through better AI and digital adoption, engagement and leadership. Comment below how we may help you to build the workforce of the future - creating value for your business, your clients, your shareholders and your employees.
Add comment
Comments