Can Learning Help Overcome the AI Capability-Absorption Gap?
Organisational readiness will take time, but that should not stop us from preparing our people to thrive in an enterprise AI workplace
We are reaching a curious time for enterprise AI adoption. Organisational AI readiness is advancing far slower than the technology, and yet more and more large firms are warning employees that their jobs are likely to disappear or be transformed by automation and agentic AI. In some cases, this is a convenient cover for post-pandemic downsizing or an opportunity for cost reduction in a time of economic uncertainty; but the long-term trend seems clear.
The New York Times recently profiled a Bay Area firm dedicated to automating as many so-called white collar jobs as possible, as fast as they can. The article’s introduction gets right to the point:
Years ago, when I started writing about Silicon Valley’s efforts to replace workers with artificial intelligence, most tech executives at least had the decency to lie about it.
“We’re not automating workers, we’re augmenting them,” the executives would tell me. “Our A.I. tools won’t destroy jobs. They’ll be helpful assistants that will free workers from mundane drudgery.”
It goes on to describe how this is changing now that the technology has caught up.
There is clearly a risk that layoffs and crude replacements of human labour will happen faster than the organisational changes that are necessary to make enterprise AI adoption successful. Replacing people with AI in advance of making the organisation AI-ready risks creating the worst possible outcome: a Frankenstein’s monster of old management processes and practices, with unreliable automations grafted onto the old org chart to cover the widening gaps in function.
We cannot pause the game and give organisations a chance to catch up, so we will need to deal with these challenges as we go along. Perhaps the best mitigation strategy right now is to focus on enhancing human capabilities and giving people free rein to learn, experiment and build the confidence that they can - and should - be part of the new world of work that is coming.
That world will be imperfect, jagged and uneven, bringing new uncertainties and risks, but I would challenge any business leader to watch Andrej Karpathy’s recent presentation about the future of of LLMs and software without feeling hopeful, positive and ready to jump in.
The capability-absorption gap
The history of the last two decades of digital transformation is littered with failed examples of tool and method adoption without the necessary organisational change, from social collaboration platforms deployed within an unreformed management system to mandating agile practices within a wider culture of top-down task assignment.
The German automotive sector’s capitulation - first to Tesla and then to Chinese competition - is a good example of this kind of management myopia, and Volkswagon’s expensive failure to understand software is particularly instructive. In a world where cars are defined by software, it is no longer an ‘add on’ that can be sourced externally. It is a key competence - but one that you can’t just buy by the kilogram. And the same goes for the wider organisation: in a world of software, you need to operate like software to succeed.
This is one reason why we are so interested in the future of organisational operating systems and the ‘world building’ that will be possible on top of them. Agility, adaptability and autonomous agentic services all require a more connected orgOS, rather than the divided silos of the last century.
But the reality of AI adoption and adaptation today is more prosaic.
Azeem Azhar wrote a few days ago about the widening gap between the speed of technology innovation in AI and its adoption in the enterprise:
This sluggish absorption, rather than frontier innovation, is the main reason we can see the AI boom everywhere except in the economic statistics. That remains true even as AI startups rack up millions to tens of billions in revenue at record speed. But the global economy is huge—about $100 trillion a year—so that is a lot of OpenAIs. It will take several years, not a few quarters, before even the fastest-growing AI startups contribute one percent to global income. (Incidentally, I have little doubt they will, and that AI-native newcomers will replace many incumbents across industries over the next two decades, just not in the next two years.)
The rest of the economy is dominated by incumbents. Those incumbents are laced with friction. They need to tackle it. Three distinct institutional frictions underpin this capability-absorption gap, and each one is structural rather than technological.
He went on to outline the three frictions holding us back:
Learning time, in common with all major new general purpose technologies, such as electrical power and even the typewriter**.**
The complexity of the technology, which has cross-functional impact and amplifies organisational complexity
Non-determinism and reliability - hallucinations and so on
These are not trivial challenges to overcome. Instead of thinking about change management or tool adoption, companies need to be thinking about proactive, ongoing adaptation and the new capabilities and architectures that are required to orchestrate a fast-growing platform of automations and agentic services.
The big blocker, of course, is management culture and inertia as socio-political control systems give way to socio-technical architectures of collaboration. Some organisations won’t make it, and that is probably a good thing.
Even tech companies and AI pioneers like Microsoft are facing difficult decisions at every level, which demand brave and visionary leadership to avoid being left behind. Bloomberg recently published a long piece about Microsoft’s CEO Satya Nadella and the complexities of their OpenAI partnership, which shared some insights into how he sees the various shifts that AI brings:
For now, Nadella … [has] sought to rally the troops by telling them to forget about the successes of the past and be ready to abandon what’s worked up to now, says Charles Lamanna, a vice president who oversees Copilot Studio and some other products. “The last five years we spent building, it doesn’t matter. It’s not worth anything anymore,” he recalls Nadella telling them. “Burn the ships.” …
With AI promising (or threatening) such existential change, there remain weighty questions about who will win and who will lose from this next technological transformation. In public, Nadella often sounds like a macroeconomist. He draws parallels between this moment in AI and the industrial revolution, and the convergence of the Global North and South, and cites economic and labor theorists from David Autor to Friedrich Hayek to Herbert Simon.
Learning and superagency
Whilst we wrestle with these thorny, long-term issues, there is one thing that both organisations and workers can do to help navigate what could be a bumpy transition: learning. Not just learning new tools or skills, but (re)learning how to learn, learning how to adapt, and learning how to navigate uncertainty.
Cultivating curiosity and escaping the workplace culture of learned helplessness that C20th management created is vital, as Geoff Livingstone recently emphasised:
*The curious-minded soul will go further. Curiosity in the AI age isn't just about adopting new technologies—it's about developing a questioning relationship with them. It's the difference between being an AI user and being an AI collaborator.*
Users follow instructions and hope for good results. Collaborators understand the underlying logic, recognize failure patterns, and continuously refine their approach.
Despite our age, those of us who grew up having to ‘hack’ or program home computers to get them to do what we want already have some of this drive to tinker in our DNA. And also gamers, especially those who grind away at real-time strategy games, according to David Hoang. But for too many people, the experience of work has been to do what they are told and stick to their swim lane.
There will be ‘jobs’ that increase in value as AI gradually encroaches on process work, most notably those that involve human contact, traditional or artisanal skills, etc. But in the workplace, the most valuable and remunerative roles will probably not be general management, but high agency, multi-domain contributors who can wrangle AI and other technologies to get things done.
Neil Perkin recently wrote about this emerging concept of Superagency, which can result from cultivating curiosity and continuous learning:
There’s something very attractive about the idea of ‘superagency’, the emerging concept that sits at the intersection of AI, organisational design, and talent. It refers to the amplification of individual or team capability through AI, allowing people to operate with outsized influence, speed, and creativity. The idea is that we’re creating a compounding effect where AI augments human capacity in scope, scale, and speed, potentially turning a five person team into the equivalent output of fifty.
Organisations will change and careers will change. Whether you find yourself leading an organisation or working for one, founding a startup or running a portfolio of roles, the best preparation and the best defence against obsolescence comes from:
Cultivating superagency and embrace the possible
Creative thinking and curiosity to explore new topics
Learning how to learn, and multi-domain knowledge acquisition
Managing and orchestrating systems and tools
Exploring the unknown and mapping what you find
Time for L&D to step up?
Right now, HR and Learning & Development functions ‘own’ learning within large organisations. But often their KPIs are about how many people are trained for their role, how many leaders are developed, and how to ensure everybody gets baseline training in areas subject to standardisation or compliance. This kind of learning can sometimes open minds to the art of the possible, but because it happens periodically, outside the flow of work, the pain of re-entry to the world of day-to-day management means it is quickly forgotten.
People can achieve almost anything with the right conditions and incentives, but they tend to go further when they are in control and need to succeed. Necessity really is the mother of invention.
So how can we get out of the way and provide the tools, opportunities and conditions that will lead to greater agency and ownership of their own experimentation, learning and development This requires breaking a lot of habits and assumptions embedded in how learning was done before. But it is also a massive opportunity for us as individuals and organisations.
Just as the internet gave us access to content and opened up new forms of online learning, so AI and LLMs will give us the ability to have our own teacher or coach, and explore further and faster than ever before.
With such an abundance of information available, learning to frame your goals and ask the right questions to learn in an exploratory way can be a real eye-opener for people who are used to being spoon-fed pre-packaged courses, methods and models.
For L&D functions, rather than just select the courses they think people need, perhaps they can focus on helping people learn how to learn with AI tools, encourage learners to formulate questions and shape their own learning quests and missions, and then give them access to AI tools trained on the corpus of learning content available to them so that they can shape their own learning journeys.
I suspect the way we deliver learning in enterprises will change a lot in the near future, and I don’t think anybody has all the answers at this point, but this is a topic we will continue to think and write about.
Next week, we will deep dive into some of the techniques that will allow us to learn with and alongside agents and LLMs, and which have the potential to create a symbiotic learning loop as we both teach, guide and learn from our AI tools.
Thanks Keith - good resource
Anthropic’s AI Fluency course is a fantastic resource on learning how to engage and think with AI - going beyond basic prompt tips.
https://www.anthropic.com/ai-fluency/overview