Photo credit: Mohammed Nohassi via Unsplash
You can’t get there from here seems to be the motto of large scale digital transformations.
If you have worked on such a transformation then you will know that asking people how to improve a service usually yields a set of incremental changes, not the big shifts you are looking for.
This phenomenon is sometimes illustrated with the quote, attributed to Henry Ford, that if you ask people what they want, they will ask for faster horses rather than cars. Ford never actually said those words, but they capture a useful idea: that you need to discard some of your context in order to imagine what could truly be different. That’s why digital transformation projects include some kind of ‘reimagining’ process, often involving brainstorming sessions, whiteboards, post-it notes and user research.
Prior to the emergence of the current wave of AI products, this reimagining was driven by the potential of established digital technology: the ability to put experience into the palms of people’s hands, the ability to use data to understand their needs, and the ability to use computing to fulfil those needs swiftly and reliably.
Today, such exercises must explore the potential of AI, particularly AI agents. This means that they must think harder and deeper than previously, because AI agents offer a capability which is alien to the ways in which most large organisations work: autonomy.
It may sound strange to say autonomy is an alien concept to large organisations. Management theory has promoted concepts such as empowerment and purpose for decades. In theory, an empowered, purpose-driven workforce is happier, more productive, more efficient and effective than one which is simply following the process and going through the motions.
And yet, despite decades of talking about empowerment, most organisations are institutionally allergic to freedom, to a degree that they may not even be aware of: it is built into their constitution, and functions like an immune system.
Consider, for example, the process of introducing a new piece of software into an organisation. Such a decision carries risk and must be taken thoughtfully. Most organisations will respond to this risk and orchestrate the decision in two ways. Firstly, they will hire a smart expert and give them authority and accountability for taking the decision. Secondly, they will send the decision through multiple approval steps and governance forums, at which it will be subject to scrutiny and challenge, by people with less expertise, authority and accountability. They will do both of these things at the same time, and consider them to be part of a coherent governance process, even when the parade of approvals undermines the notion of accountability, and the application of expertise negates the value of extra approvals.
I’ve picked this example because I am most familiar with it (and because I have serially been guilty of implementing, operating and perpetuating bad governance structures), but it’s not isolated to technology: indeed, people in the technology team probably get rather more autonomy than other parts of the organisation because others don’t always understand (or care) what they are up to. Throughout traditional organisations, it is normal to see people employed to do a job, following processes which prevent them from exercising any discretion or judgement. If you examine those processes, you can usually see traces of everything that has ever gone wrong, and attempts to prevent them from going wrong again.
The tendency of organisations to grow processes that restrict autonomy does more than create friction: it skews their operating model and working rhythm. Action follows the tick of the governance clock, and teams are built to smooth the passage through (and occasionally around) formal processes.
This means that, if we are to introduce autonomous agents into our organisations, we cannot simply reimagine our services and processes as we have previously done for digital transformation. We must think about trust, control and the conditions for autonomy: what does it take for us to believe that any entity has the authority to take action, which controls are meaningful and worthwhile, and which are purely performative? What prevents us from sustaining that trust, and causes us to clog up the system with approvals?
I don’t pretend to have answers to those questions: in every organisation I have ever worked for, processes have reasserted themselves, despite best intentions. But if we think hard enough about them, perhaps we can do more than reshape our enterprises for autonomous agents: we can figure out how to make good on the promises of empowerment and purpose for humans too.
