Deterministic vs Emergent Work

There are at least two kinds of work: deterministic and emergent.

Deterministic work is predictable, repeatable, and clear. Even with extremely complex pieces of work, given the same inputs, and the same processes, if we follow the steps we can confidently reliably predict the outputs and outcomes every time – making a cheese sandwich, launching a rocket for the fiftieth time, manufacturing your 500th snowboard, deploying software for the 100th time in a highly controlled environment.

Emergent work, by contrast, is volatile, uncertain, complex, and ambiguous or “VUCA”. It is relatively unstructured, inconsistent, and unpredictable. The same inputs and processes do not necessarily lead to the same outputs or outcomes. Consider the work of coming up with a new recipe, designing a rocket, riding a snowboard, or releasing a new product.

In the early 1900s, Frederick Winslow Taylor became one of the world’s first management consultants with his theory of Scientific Management. It attempted to analyse, standardise, and accelerate repetitive, manual tasks to increase output, reduce waste, and improve quality. It did these things quite well but was even more effective at completely alienating the workforce resulting in mass walkouts in at least one foundry in 1911.

It operated on the assumption that the work was knowable, that it could be entirely planned in advance, and be done in exactly the same way – every time – requiring no original thought, experimentation, or adjustment, only practice.

Donuts vs donut machines

Some work is actually infinitely repeatable, particularly work which has been done many times before with the aim of producing the same output again. This work cries out for standardisation and automation to clearly establish repeatable, scalable, learnable processes that consistently deliver high quality and minimise waste. Consider the work of installing car doors, algorithmically solving a Rubiks Cube™, using a recipe to make donughts, using software to test, build, and deploy other software in a carefully controlled environment.

One type of work is not necessarily better or worse than the other, and both are required to run most kinds of organisations but knowing the difference and being able to identify the different types is essential to applying the right methods of improvement.

Every organisation and every team should identify which pieces of work are deterministic and which are emergent. Work can even be mapped onto a spectrum, it’s not necessarily binary. Once a team understands where key pieces of work lay on this spectrum, they should consider which pieces of deterministic work may want to shift to be more emergent, or vice versa.

They should create standard operating procedures or even step-by-step work instructions, and automate as much of their deterministic work as possible. This will free their creative energy for the emergent work.

Emergent work can be strengthened by diversity, inclusive facilitation frameworks such as Liberating Structures, design thinking, drawing, reflection & adjustment, prototyping, and hackathons. If you’ve got a great donut recipe, build a machine to make donughts. Don’t reinvet your recipe every time.

For domains which are complex, emergent and non-deterministic – like dreaming up a new donut recipe or designing a new digital service – trying to standardise inputs, outputs, processes, and behaviours can be disastrous. In these environments, you need diverse opinions and experiences, complementary skill sets, and open lines of communication. Most importantly, you need the ability to sense and seek out subtle feelings and stimuli that may indicate emerging showstoppers or game-changing opportunities.

Non-deterministic, emergent work requires tremendous awareness, sensitivity, and constant decision making. This is impossible when one person or process pre-determines what others should do. Consider the importance of volunteering vs assigning when we adopt a “commander” posture vs a “leader” posture or when we foster debate over dialogue. There are dozens of inputs and signals which must be weighed up and considered against previous inputs and decisions to make the best possible decision about where to go next.

Don’t use a screwdriver when you really need a paintbrush.

One of my favourite phrases (I wish I could remember where I heard it) is that designing and maintaining complex digital systems is actually more like gardening than carpentry. We used to assume that when we finished building a new system, it would run forever and could be transitioned to “Business as usual” requiring only basic maintenance. We’ve since learned that most systems require not only pre-determined maintenance but ongoing, continuous improvement in order to stay relevant and useful while people’s needs and tastes evolve.

Artificial intelligence has enabled computers to tackle emergent work with increasing skill & success but it’s still critical to understand the difference. Don’t use a screwdriver when you really need a paintbrush.

What I need now is a donut.

See also, Wardley Mapping which, I believe, does a beautiful job of describing the journey that any piece of innovation takes from starting as emergent, novel work at the beginning to deterministic commoditized work later on.