Will We See the First Programmable Organisations In 2025?
Looking ahead to the big prize in Enterprise AI and how to pursue it
This long-form piece is an example of the content we share in the premium edition, which we are making available to all subscribers - HNY!What Lies Ahead for Enterprise AI?
2024 was an extraordinary year for ‘AI’ - from the rapid improvement of LLMs to new conversational interfaces and other ways to interact with GenAI. OpenAI started looking like the new Netscape. Pro- and anti-AI positions took on semi-religious, often apocalyptic overtones, in a debate fuelled by heretics, sceptics, believers and (inevitably) grifters launching countless LLM wrapper apps to soak up some of the excess energy and cash.
There are many technical benchmarks that quantify the progress of LLM intelligence in 2024, but most notably,. AI finally managed to create a credible video of somebody eating spaghetti:
Discussion about the imminence and significance of AGI or even ASI will continue, and yet we already have enough new affordances emerging from AI technology to elevate and augment vast amounts of underused human creative potential.
But will that happen, really?
We have already seen solid evidence of good personal productivity use cases, and the outlines of agentic AI are emerging as the next stage of technical development. But the real world-changing impact is expected to be in the enterprise space. Shaping and running 'smart' AI-enhanced organisations that optimise the balance of automation and agentic management combined with human creativity is the exciting prize, and one that - although catalysed by AI - also requires lots of other challenges to be overcome in addition to the technology.
In 2025, we can expect to be distracted by a lot of noise and magical solutions, so it is worth reflecting for a moment on where we want to navigate to with enterprise AI, and try to focus on what is needed to get there.
Towards Smarter Organisations
The world of work is changing rapidly, but the way we aggregate and coordinate work within organisations has remained largely unchanged.
For various reasons, traditional organisations are not great talent attractors right now. Longevity means senior people are hanging on to well-paid roles, blocking career paths for younger workers. In many entry-level jobs, living costs have outstripped salaries. These factors and others, such as poor management, are leading some of most creative people to eschew traditional organisations and conventional roles. Side hustles, micro-startups and digital nomad living are hard to compete with for corporate recruiters.
But to develop our societies and economies, we still need industry, transportation, utilities and other areas of activity that operate at a collective scale, which means we still need companies and other kinds of organisation to keep things running.
In Europe and the United States, there is a growing consensus that we need to get back to building things that go beyond the virtual influencer economy, for example in areas like climate solutions, electrification and automation, transportation, medicine, food science and defence.
We are already seeing smart technology deployed in these areas, from solar and renewables and automated vertical farms to cheap, smart drones that are transforming warfare.
But in most AI-enhanced science labs or smart factories, if you wander off into the offices that house the management functions of the organisation, you will probably still find legions of people using fifty-year-old technology like emails and virtual meetings to run the company in painfully manual ways. When technology first entered these offices, the ‘shop floor’ lagged behind, but since then the situation has reversed.
The expected arrival of enterprise AI at scale in 2025 presents a once-in-a-generation opportunity to reinvent management and work coordination in ways that could substantially reduce operating costs, whilst making organisations more agile, adaptive and automated.
Ideally, the goal should not be to replace people per se, but to stop using people like automatons in a process factory and allow them to do what they are uniquely capable of doing: innovating, connecting, sensemaking and creating value for each other in so many new ways.
Organisations can be more automated and machine-like at the back-end platform level in order to be more human and connected at the front end. It will lead to an unbundling and re-bundling of 'jobs' and roles; many will no longer be needed, others will become more interesting and valued, and new roles and opportunities will emerge for those willing to engage with change.
We are now routinely using the term 'programmable organisation' to loosely describe where we think enterprise AI is headed and, for now at least, 'programmable' implies human control and oversight.
The Immediate Opportunity & Challenges
In our work on technology-enabled organisational improvement, we have seen successive waves of innovation crash against the sea wall of management bureaucracy, causing some erosion but not a breakthrough.
So why would it be different this time?
AI looks like a catalyst that will combine with other new ways of working, because these are necessary to its success, to create a wave of change that finally breaks through the ancient coastal defences of the C20th corporation.
We think this will lead to a new era of organisational design.
In terms of immediate predictions, I think we will see the following signals over the next few years:
Large organisations deploy agentic AI to reduce their cost base considerably, whilst increasing productivity and speed to market.
A widening gap in business agility between companies that embrace Centaur service teams and human-AI hybrid ways of working, and those that do not.
Billion-dollar revenues generated by small teams using AI augmented ways of working, without the need for excess management layers.
New startups and scale-ups bake in intelligent management and automation from the start, bypassing the traditional separation into functions and divisions entirely.
We have written about the various AI readiness challenges that organisations need to grapple with, such as connected data and knowledge stores, but there are also major structural blockers in the existing operating system that need to be addressed - for example:
Internal functions and processes need to be re-imagined as services that are partly or even largely automated and composable, so that teams can access what they need (or are required to use) via internal platforms. And the people and teams within these functions should develop more of a product owner mindset, seeking to improve and promote their function's offering, whilst focusing their face time on internal customer experience and exception handling.
Teams that create value for customers should be as autonomous as possible, working on internal platforms that encode the company's policies and compliance considerations, and using an appropriate level of AI augmentation and automation, like a kind of team exoskeleton that is human-guided, but technology-powered.
Managers need to specialise. Instead of occupying generic management layers that oversee branches of the old hierarchy, managers might specialise in people (coaching, supporting, developing talent), work coordination and optimisation (using tech and data to keep an eye on things, like organisational sysadmins), or areas such as strategic leadership and innovation. With the support of smart tech and more autonomous ways of working, only a fraction of today's management layer is likely to be needed in future, but their roles would become more important, interesting and valued.
Distributed leadership. The days when a senior management board or a C-suite has enough knowledge of all areas of the business to make all the key decisions are long gone. Working within a clear strategy framework, smart leaders need to make better use of the collective intelligence below them to guide decisionmaking, rather than require all key decisions and options to be pitched to them for a final say.
Process Management → Service Automation
In the past, we took process management for granted and accepted it tends to lead to a gradual accretion of rules and process steps that will slow things down. If we think of the coordination and management of work as a software problem, and apply software design principles to solving it, then it is possible to see how we can create smarter, more connected solutions.
But for something to be programmable, it first needs to be addressable. We need to create the building blocks and components that can be combined and chained together to support automation. For example:
Process steps that can be referenced like software routines or functions
Processes as micro-services that can be called directly or combined with other services
Bundles of processes and workflows as organisational components
Management playbooks as software libraries
AI agents as software workers that run in the background
Customer journeys as 'apps' built on top of these components with their own UX
In the first wave of automation, many companies used Robot Process Automation (RPA) to enable computers to perform manual process actions much faster than people could, which was useful in areas such as procurement and contract management for big companies. Today, this semi-automated approach is even easier to achieve using the 'computer use' features of new AI tools (i.e. you can give an AI the right to operate your computer, navigate systems and perform tasks quickly at scale). But when each process is encoded as a small online service, ideally with an API so that it can be called by other services within the organisation, then things start to get really interesting.
Componentisation and common standards were key enablers in the evolution of software from manually-created code to a system of libraries, frameworks and automated operations. And now, thanks to GenAI, we can invoke all this magic through simple human-readable prompts and incantations. One day soon, we will be able to create and run organisational units in the same way - just by asking questions and issuing instructions, leaving the coordination of underlying process steps to the agents under our control.
Organisations that have pursued a platform architecture model are already further down this path, with online services doing the work that manual process management used to run. But the next step - allowing AI agents to navigate and operate the service and process catalogue - will enable programmable organisations to become a reality.
Organisations as Software?
Too often, change management and organisational development have taken the underlying architecture of their organisational operating systems for granted and worked to fix the symptoms (behaviour, culture, divided functions, lack of alignment, etc) rather than the system that causes them. When we can automate control and monitoring systems, we will have a much clearer view of what needs to be improved and fixed, rather than just blindly tinkering at the edges of the problem.
Software engineers, architects and designers have progressed so rapidly since the days of manual coding and testing. We should expect organisational designers, architects and other roles, such as change management, to learn from this trajectory and take a similar approach to ensuring our organisations are fit for purpose once the organisational structure becomes more programmable.
None of this is an argument for anarchy or lack of control. Quite the opposite - relying on manual enforcement and comms to uphold values, ethics and legal compliance is clearly hit-and-miss at best. And whilst attentive managers might spot problems as they emerge, they cannot compete with open, real-time data flowing around systems that automatically flag out-of-bounds results or signals that indicate action might be needed.
New startups and scale-ups are more likely to bake in automated monitoring and control systems from the beginning, and a mature service platform will have compliance and the rules of the road embedded in their algorithms so non-compliance is harder to achieve (although to be fair, people are already proving adept at persuading LLMs to ignore their own guard rails through clever prompting). These emerging post-AI companies will run pretty lean compared to traditional bureaucracies, but they will also have the chance to set their own rules, culture and norms, so there is no reason to believe that these smaller teams will not be great places to work.
A long time ago, when we were younger and more optimistic about the internet, Larry Lessig wrote a book called Code and Other Laws of Cyberspace, which popularised the phrase 'code is law' to explain that computer code played a similar role to laws and legal precedents in governing behaviour online (for better or worse). He was later prompted to return to the topic when the 'code is law' argument was used by crypto scammers to argue that anything permissible in code (even if it might be a bug) is permissible in law, at least within the Wild West crypto gold rush.
Organisations are, in essence, a form of software operating systems for collective action by people. And they are becoming more so all the time. Some might fear this dehumanising, but in fact, the reverse might be true. From feudalism to factories to offices, organisations have existed mostly for the benefit of those at the top, regardless of talent or work rate. In a world where talent and insight can come from anywhere, and where diversity of thinking and experience are an asset, we need organisational systems that are more objective, more competitive and with governance and control mechanisms that are less performative and more effective than what we have grown up with.
Software is a form of language, and has evolved to become more powerful in its expression over time, just as human languages went from basic communication to poetry and literature. But it could only achieve this by building the services and components that we can now invoke with a single command. If we can do the same for our organisational operating systems, then we can design, build and operate businesses and their value chains just by describing what we want them to do.
This would be a powerful new form of literacy that could enable a revolution in world building and problem solving, whilst freeing many jobs from the straitjacket of conventional management.
Our contribution
In 2025, we will be sharing techniques and approaches that can help leaders, managers, internal function or product leads take practical local steps towards making their areas of responsibility smarter and, over time, more programmable. A key part of this will be helping leaders develop enterprise AI literacy to think like organisational architects and developers, and equipping them with practical techniques, recipes and use cases to get to value quickly, plus tools to oversee and manage the associated change management challenges.
This work involves technology, of course; but mostly it is about how we wrap learning, change and organisational development around the emerging use cases for AI and agentic automation.
We will also be highlighting and showcasing the work of colleagues across all these fields who are advancing towards smarter organisations, including the future of work and jobs, organisations as a platform, leadership in the age of algorithms and other adjacent topics.
So - please - if this mission resonates, take a moment to share this essay with colleagues or practitioners who might be interested, or who might benefit from connecting with us, and share your work or insights in this area with us.