Let's talk about the skill and mindset transformation that's going to be needed...
AI is not just 'more tech' - it's a paradigmatic shift. What new skills will your teams need to develop - including how and when?
Quick day 4 check in! How are you getting on with this series? Please do ping me an email letting me know if this is useful or not - what's missing and what questions remain unanswered. I'll answer as many as I can in tomorrow's email.
Today, we're look at the specific skills your teams will need to develop in order to adapt to the arrival of AI. In fact it's not just about skills. It's a whole mindset shift.
Microsoft's Future of Work researcher, Alexia Cambon shares some numbers on where employees' heads are at.
- 49% people say they’re worried AI will replace their jobs
- 70% would delegate as much work as possible to AI to lessen their workloads
Now, we don't know how much crossover there is between these two groups. Are there two groups of people - one group feeling excited, the other feeling anxious? More likely, based on my own research, most people are feeling ambivalent (i.e. mixed or contradictory feelings - not a lack of opinion either way).
What we can take from Alexia's Microsoft research is that employees fear AI less than they welcome it - or perhaps fewer employees fear it than welcome it!
What we do know - it's going to require a big shift.
Let's start with skills - what exactly do they look like?
I'm just going to look at the two main areas for now. Fusion skills which cover how we each partner with GenAI and Prompt Engineering Skills which cover how we specific generate the deliverables we are looking for.
Fusions skills
Fusion skills describe our ability to work at the human/computer interface. What we integrate, how we do it and how we 'handle' it.
The book Human + Machine: Reimagining Work in the Age of AI categorises eight fusion skills as follows:
1. Rehumanising time: The ability to reclaim time for "human activities" like interpersonal connection and creative exploration. As more is handed off to bots and agents, we need to rethink and relearn how humans connect.
2. Responsible normalisation: Shaping the perception and purpose of human-machine interaction ethically and strategically. This involves actively questioning the impact of these interactions on individuals, businesses, and society at large.
3. Judgment integration: Recognising when human judgment and ethical considerations are crucial, even in highly automated systems. This necessitates strong decision-making and analytical skills, coupled with a firm grasp of ethical frameworks.
4. Intelligent interrogation: Mastering the art of asking insightful questions to extract crucial information from complex data systems. This demands strong research and information analysis skills, alongside the ability to formulate relevant and meaningful queries. We'll talk about Prompt Engineering Skills in more detail in the next section.
5. Bot-based empowerment: Embracing the power of AI assistants to augment your capabilities and enhance productivity. This necessitates an open mind to technological advancements and the ability to adapt workflows to integrate these tools effectively.
6. Holistic melding: Cultivating a deep understanding of how machines work and learn, while allowing machines to capture your behaviour data to refine their interactions. This requires both technical literacy and an openness to co-learning with AI systems. For most of us, this is a massive step.
7. Reciprocal apprenticing: Engaging in a two-way learning process where you acquire new skills through collaboration with AI agents, and they learn from your expertise. This emphasises adaptability, active learning, and strong communication skills. Again, most people have never done anything like this before.
8. Relentless reimagining: Possessing the vision and drive to continuously imagine how AI can transform your work, processes, and even entire industries. This requires creativity, critical thinking and a willingness to embrace radical change.
To this list of how to combine distinctly human skills with machine skills, I would specifically call out one more skill:
9. Ethical discernment - how to use AI-assistance ethically and responsibly. It should be everyone's responsibility in every role to think about the correct, ethical and appropriate use of AI - this can never be delegated, or passed up to the leadership team or passed off to the technology itself.
Prompt Engineering Skills
Now we've explored that human/AI interface, let's deep dive for a moment into Prompt Engineering Skills (PES) - the ability to know the right questions to pose to your AI agent or copilot and how to train it to develop its intelligence. These are crucial for your teams to know HOW to achieve what's possible in these early days of mainstream AI.
Specifically, PES covers:
- what to ask and how
- what context to provide
- how to interpret and check the results for hallucinations (if you remember from yesterday, this is AI speak for a right royal cock up)
Everyone will need to develop Prompt Engineering Skills to some extent. And being a Prompt Engineer will be a job in its own right.
Here's what you'll need to think about developing for your teams - and a great place to start is with yourself. If you're not already doing some 'heavy tinkering', I absolutely think you should be.
1. Understanding LLMs: Deep knowledge of how LLMs learn and generate text is crucial. This includes understanding their training data, strengths, weaknesses, and potential biases.
2. Defining needs: Clearly outlining the desired outcome is key. What specific task do you want the LLM to perform? What kind of output are you aiming for?
3. Crafting clarity: The ability to craft concise, unambiguous prompts and how to chunk down instructions to get the best result from the LLM.
4. Tailoring content: Providing relevant context and information specific to the task at hand. The more the LLM understands the situation, the better it can respond.
5. Leveraging examples: Providing illustrative examples to demonstrate the desired style, tone, or format of the output. This helps the LLM align its response with expectations.
6. Fine-tuning iteratively: Understanding how to develop a basic prompt, observe the output, and refine the prompt based on the result to get the best outcome.
7. Domain expertise: Understanding the specific terminology and nuances of the field to craft better prompts.
8. Adapting to changes: Adapting prompts as the models and needs change.
No doubt, prompt engineering will change massively over the coming decade. But the above is a good place to start.
What about mindset? How will that need to change?
The reality is that the technology is evolving faster than humans can keep pace right now. A shift in mindset will up the pace of learning. Here's what I see:
Machine/human fusion will need to be socalised... and normalised.
It's new. It will weird. It might feel wrong. And it will almost certainly feel LESS efficient to start off with! There is a big period of socialisation and normalisation.
This is the era of continuous learning.
I know, I know - learning has always been continuous! But with previous technology shifts, there was a sense of some clear-ish milestones and destinations. Now it's different. We are in an 'opening window' where the technology will shift and develop faster than most people can keep pace...
Getting comfortable with not knowing
... and for that reason, people will need to adjust to no longer being 'on top of' their domain. There is no new hurdle to jump over - rather a series of unknown hurdles to keep finding and scaling. Previous technical shifts have evolved at a pace most people could keep up with. The internet developed in the mainstream over a decade, with services and routines changing gradually over that time. Not so with AI. This is massive for our sense of identity at work. We are all used to feeling we have expertise and experience that counts. Many people will find that eroded - with no end point in sight to reclaim that identity as someone who is now on top of this new thing. There is no finish line and people will need to get used to this without feeling they have been deskilled.
From competition to collaboration
For some, it will be a big shift to view AI as a collaborator, not a competitor. It can automate repetitive tasks, freeing up employee time for higher-level thinking, creativity, and innovation. Employees will need to really understand (and believe) that though AI excels at data processing and repetitive tasks its lack of empathy and other uniquely human talents will make people even more valuable in an AI-powered workplace.
From instructions to exploration
Traditionally, people are used to receiving relatively clear instructions or at least a strong brief. AI requires a more proactive and explorative approach. Employees will need to learn to ask the right questions, experiment with the technology and interpret its outputs creatively. This will come more naturally to some people than others.
From individual contribution to collective impact
While individual contribution will remain important, AI necessitates a shift towards collaboration and collective impact. Employees will need to effectively communicate with AI systems and colleagues, leveraging the strengths of each.
A shift to data-driven decision making
Intuition and experience remain valuable, but AI provides access to vast amounts of data. Employees will need to take a big step towards using data-driven insights to make informed decisions and justify their recommendations, no matter their starting point.
Embracing transparency and "explainability"
Employees should understand how AI works and the data used in its decision-making processes. This promotes trust and acceptance of the technology.
Prioritising ethical considerations
As AI interacts with human lives, ethical implications cannot be ignored. Employees will need to consider potential biases and advocate for ethical deployment of AI systems.
A gear shift
You may already know the concept of deep work (coined by Prof Cal Newport in his book of the same name) - the periods of uninterrupted time we need to do complex or detailed high value work. Conversely ‘shallow work’ is the opposite - the low value work that requires us to work in a lower gear is either atomised (small in size, high variety - therefore inefficient) or high volume (monotonous, manual - and also inefficient!). Spending more time doing deep, complex work will be a mindset - and gear - shift in its own right and will draw down harder on people's cognitive load.
Some final things to think about...
Roles are likely to become less specific and merge together into broader job categories as AI will help people work more widely than their own expertise and experience currently allows.
There are worries about skills atrophy which can leave organisations exposed. John Sicard of Kinaxis compares it to when something weird happens on a plane, and the autopilot switches off, leaving the pilots in charge. That's when they really need to know what they're doing! So, keeping people's skills sharp will be important.
You may also find that GenAI is most valuable for more experienced employees with higher levels of expertise who can discern problematic or hallucinogenic responses more easily. There will be different training needs for people early in their career who might end up with little exposure to 'what's inside the box' of the job they do - the simply won't see people do it any more.
Things will get worse/slower before they get better. Employees will almost certainly need some space and downtime to acquire and apply these new skills. This will mean a temporary loss of productivity to ‘put down the axe and pick up the chainsaw’.
Annnd, in case you hadn't already deduced this, GenAI is not going to save you tonnes of money and time in the short term, sorry! You’ll need to invest in a huge amount in people development.
Got all that? 🙂
Next:
Ok, it's going to be big. What should I do now..?
References
1) Taking the Reins on Managing Generative AI from Betterworks
