The respected American venture capitalist Mary Meeker recently shared her first internet trends report since 2019, focused on AI and its future. It runs to 340 slides (with center-aligned text … grrr!), so it takes time to absorb. But in addition to its meta-level analysis of investment and adoption trends, cost and revenue projections, and predictions about where AI is headed more broadly, it also covers enterprise and organisational uses of AI.
As with the original dotcom boom, first movers in AI are spending vast amounts of money on pump priming the sector in the hope that they will dominate and later find ways to make that money back later.
As the report reminds us:
It’s different this time, we’ll make it up on volume, and we’ll figure out how to monetize our users in the future are typically three of the biggest danger statements in business.
Both enterprise computing incumbents and AI-native new entrants are investing to establish a dominant position inside large organisations, but this is not simply a new wave of enterprise software.
The shift from vertical SaaS to horizontal platforms could also change the way software is used and paid for by organisations, as the report notes.
Horizontal enterprise platforms could usher in a new form of monetization: not by selling siloed software licenses, but by charging for intelligence, embedded throughout the stack … The value shifts from tools to outcomes – from CRMs to automated deal summaries, from service desks to AI-powered resolution flows.
At the same time, enterprise AI is likely to disrupt the nature of human work itself by automating some jobs, changing others, and creating new high-agency roles where people orchestrate, manage and guide a variety of AI agents to achieve more than they could before.
The idea of the human workforce re-configured to teach and refine machines as a primary function might sound dystopic. But it’s worth remembering historical parallels. Fifty years ago, this prospect of rows of cubicles and uniformed office workers sitting quietly in front of LED computers ten hours a day likely sounded equally dystopic. Yet here we are. Technology has constantly redefined and evolved the nature of work and productivity…AI is no different.
And although many are concerned about AI potentially making many jobs obsolete, the report quotes Jensen Huang’s remarks at the Milken Institute Global Conference last month, where he argued:
We don’t have an abundance of workers. We have a shortage of and for the very first time in history, we actually have – we can imagine the opportunity to close that gap to put 30-40 million workers back into the workforce that otherwise the world doesn’t have. And so you could argue that artificial intelligence is probably our best way to increase the GDP, the global GDP, and so those are two other ways to look at it. In the meantime, I would recommend 100% of everybody you know take advantage of AI and don’t be that person who ignores this technology.
Is agentic AI already priced in?
There is a growing consensus that agentic AI is a done deal, and we need to prepare workers to operate alongside a growing population of agents and automations.
A recent episode of the podcast McKinsey Talks Talent is a good example of how we are taking this for granted and not really asking hard questions about organisational readiness either in terms of technology, data or the architecture of work.
When Mary Meeker talks about the shift from vertical SaaS to horizontal platforms, this is not just a shift in the software organisations use, but also a shift in their fundamental architecture. We are moving from a place where an organisation is defined by its divisions, departments, processes and management – and uses enterprise software to improve the way this system operates – to a place where software defines and perhaps embodies the organisation, its rules, processes and its coordination of work. From organisations using software to organisations as software.
When the key components and capabilities of an organisation’s work are available (and findable!) within a common digital platform and shared data layer, it becomes possible for anyone (in theory) to create workflows, experiences, tools and disposable apps on top of the platform to innovate or just to get work done more effectively.
Leaving aside the question of where this leaves the enterprise software firms who have provided much of this middleware at great cost in the past, the more important question is how many organisations are really, truly doing the architectural work needed to make this happen.
In smaller, specialist sectors, it will probably be possible to buy or rent entire sector-specific platforms that function as a basic organisational operating system. Meeker’s report alludes to this in areas such as legal, accounting and healthcare, where rather limited, one-size-fits-all platforms such as practice management systems have been used in the past.
But for large complex organisations, the bigger question is whether they can handle the change management and the re-architecting work themselves, or whether new AI-native market entrants will start with the platform and then build out the right type and size of organisation on top. Or, as hinted at by the strategy of AI investment pioneer Elad Gil, perhaps firms with strong processes and IP but lacking visionary technology and management might be acquired or rolled up to accelerate their transformation.
Jurgen Appelo is excited by the potential for what he calls Agentic Network Organisations (ANOs), which seems like an evolution of previous models like Holacracy, DAOs and RenDanHeYi that were difficult to implement due to the complex coordination required. Appelo believes agentic AI will – eventually – solve that problem.
Assuming an organisation has a comprehensive platform of services and components to work with, creating this kind of organisational architecture on top would enable much more direct coordination of work, more agility and adaptability than conventional management allows. And all of that creates the conditions for programmable organisations, as we have written about previously.
Agentic AI needs strong foundations
In the first dotcom boom, it was widely understood that without some kind of micro-payment or transactional system, the web would not be able to fulfil its potential as a powerful new connective medium for people, business and society.
Largely because of this foundational weakness, advertising and the monetisation of attention became the default, with disastrous results. The web, and especially the social web, was a much better experience twenty years ago than it is today. And we had more or less functioning democracies that had yet to be fragmented by bad actors seeking influence and money through exploiting our psychological weaknesses – the resulting social media world that emerged from this arguably contributed to the socio-political problem that Bellingcat’s Eliot Higgins calls disordered discourse.
There is a non-zero chance that we might inadvertently build the agentic AI era on similarly shaky ground unless we consider all the foundational elements that need to be in place, from renewal energy and scalable power grids at the bottom of the stack through to copyright law, deep fake protections, algorithmic transparency, cybersecurity and so on.
And at the firm level, we need to think about the horizontal platforms Mary Meeker mentions, and not just embed agents into the vertically divided structures of the Twentieth Century corporation or use AI to super-charge bureaucratic management.
Some of these challenges are non-trivial and will inevitably be addressed at a linear pace, whilst AI continues advancing exponentially. But that doesn’t mean we shouldn’t try.
Perhaps the best place to start is with human capabilities and skills, helping people adapt to AI augmentation and discover a sense of agency and empowerment in the face of fears about jobs and change.
Using AI to make things or automate work is now an accessible area of human-AI collaboration for many people, and whilst it is not without risks and flaws, it is a pretty magical and inspiring entry point to the potential of agentic AI. This funny piece by a programmer rebutting the anti-AI stance of some of his peers touches on the superpowers we now have access to:
Kids today don’t just use agents; they use asynchronous agents. They wake up, free-associate 13 different things for their LLMs to work on, make coffee, fill out a TPS report, drive to the Mars Cheese Castle, and then check their notifications. They’ve got 13 PRs to review. Three get tossed and re-prompted. Five of them get the same feedback a junior dev gets. And five get merged.
If we take a glass-half-full approach to what this means for jobs and work, there is clearly a lot of empowering potential to celebrate.
As Arafat Kabir puts it in How AI Polymaths Are Quietly Rewriting The Social Contract Of Work:
The social contract of work is being renegotiated in real time. AI polymaths are at the forefront, blending human ingenuity with digital power. The future is being written—one project, one agent, one polymath at a time. The question isn’t if this change will come, but if you’re ready to move with it.
There are of course potential downsides for workers, not least those who invest years in education only to emerge into a job market where much of what they learned has been automated or commoditised.
But as Addy Osmani advises in an article on avoiding skill atrophy in the age of AI:
Use AI it to amplify your abilities, not replace them. Let it free you from drudge work so you can focus on creative and complex aspects – but don’t let those foundational skills atrophy from disuse. Stay curious about how and why things work. Keep honing your debugging instincts and system thinking even if an AI gives you a shortcut. In short, make AI your collaborator, not your crutch.
Learning and skills development are still rather stuck in the era of courses, certification, one-size-fits-all training and learning as something that happens away from the workplace. This is one of the most important foundational elements that can be tackled today for companies that want to bring their workers with them on the change journey.
Josh Bersin’s insightful report on how AI will revolutionise the Learning & Development function is a good place to start thinking about the art of the possible today. Diginomica’s review of the report is a good summary, and highlights what I think are two key points:
[Large organisations] … have these giant, massive intellectual property databases in people’s heads. A lot of it’s not even written down. For example, what Rolls Royce is doing is going topic by topic, and interviewing people who are getting ready to retire, to get their information into some form, and then bringing it into a platform like this. So they’re going to have a bunch of subject matter expert systems. The AI platform becomes almost a digital expert. So if you choose to just ask it questions, you can do that. But it can also teach you things.
I think a lot more of the learning that people need is going to happen in the flow of work while they’re doing something else. You are not going to stop what you’re doing, go learn something and come back. You’re going to be processing a claim, or fixing something, and you’re going to say, ‘How do I do this?’. And this kind of a system could be injected right into workflows.
This topic, and the need to re-imagine the delivery of learning within the flow of work and as part of organisational development, is something we will return to in the coming weeks.