Agile is Obsolete

Are you old enough to remember Kodak?  Some may not know that they dominated the photography market starting around1896.  They maintained market dominance through their monopoly in film, which drove market control over the devices that used the film.  But overnight in January 9, 2007, when Apple introduced the iPhone, the writing was on the wall with Kodak in the face of the new digital camera era.  This new technology disruption of device integration, centered around discontinuous technological jumps with the internet and mobile devices, transformed how photography would be performed forever. 

The same is now becoming abundantly apparent for software development and how it has been largely performed since the inception of the computer.  This article is not a false dichotomy discussion about the “waterfall” versus “agile” – even though the negligence of leveraging a misinterpreted 1970 paper by Winston Royce continues to this day and mostly for nefarious reasons.  I am talking about modern software development approaches dominated under the “Agile” label and branding. 

The introduction of AI, specifically Generative AI in combination with Declarative AI has now rendered Agile mostly obsolete.  I can say this because I have lived the first line of the Agile Manifesto for over 15 years specifically in relation to AI; it reads “We are discovering better ways of developing software by DOING IT and teaching others to DO IT”.  When you look at all the frameworks and stuff peddled out there in the Agile echo chamber, an extremely large percentage of it is not reflective of “doing it”, or at the very least is based on either confirmatory bias or is not created in the spirit of addressing Muda Type 1, whereby currently necessary waste can be removed by changing ones’ way of working.  The industry has been infected with a mind virus related to a Scrum-oriented bias for many years now. 

To support these rather provocative assertions, let me gently dissect the beloved Agile Manifesto that has become rather dogmatic broadly in industry.  The reverence that is held for the 17 dudes that dropped this political movement is astonishing, and the obstinance and willful ignorance that has ensued is equally damaging. 

Individuals and interactions over processes and tools

This declaration of bias indicates software development is solely reliant on Human Intelligence.  Enter Artificial Intelligence and machines (in the form of tools), and it becomes confusing.  Add the fact that these tools can make individuals more productive by a factor of 100X or more, and the bias seems a tad bit misplaced.  Facilitating the orchestration of the Human Intelligence and the Artificial Intelligence is necessarily a process so that part kind of gets destroyed as well.  The disintermediation of the supply-chain / process is only enabled by the disruption created by AI and, with AI being a tool/technology.  I note that much of the mojo surrounding the Agile movement stemmed from “Humans and Technology”.

Working software over comprehensive documentation

This bias, while consistent with the spirit of the first line of the manifesto, needs to be an AND as the technology landscape has changed and documentation processes are largely automated and crucial at the API and architecture level.  The original intent was to stop all the waste created from those furthest from the craft of software construction and programming, but a lack of good documentation on cloud APIs that are in constant flux can waste a great deal of time.  It should be ‘don’t have humans manually create documents for knowledge capture (AI can do a lot of this now), and definitely don’t insert delays in the feedback loop (which is accelerating with AI) because tacit knowledge emerges and is not prescriptive (a small portion of knowledge)”.

Customer collaboration over contract negotiation

This bias has always been laced with political wokeness and trade-unionist sentiment.  Obviously, collaboration with customers to extract the tacit knowledge they possess to hit the mark with software is necessary, and that extraction requires trust.  While a contract is a formalized trust relationship, it also reflects the realities of the commercial environment where software is created and the common law realities of Tort Law.   When you reduce the cost of software substantially due to the degree of AI automation now available, the damages that flow naturally from breach are greatly reduced, so emphasizing these issues in a false dichotomy is counterproductive.  You need pragmatic invocations of both concepts in the context of AI powered software delivery.

Responding to change over following a plan

This declaration needs a facelift.  The time horizon for which AI supported software development is compressed, so the issue of adaptability and dynamism in software is focused more on extraction of tacit knowledge or sparking innovation.  The frequencies of iteration that are involved are so high that plans are morphing into a focus on deployment plans when software is harvested from the “innovation flywheel". Planning on when the stability of the emergence from the software factory is a fool's errand, as stakeholders and fiduciaries are observers to the "theatrical production". These plans are automated so following them is innate with deployment automation.

I could continue on to the principles which need to be tempered to the realities of the new AI-driven paradigm, but perhaps I have made my point well enough already that the software engineering “standard-of-care” needs rearticulating. 

A final point related to architecture.  There is something called Conway’s law, named after Mel Conway which can be stated as:

The structure of any system designed by an organization is isomorphic to the structure of the organization.

The implications here are that when you fundamentally change the way of working or the technology that supports how software is developed, they are co-dependent and co-emerge.  With HyperAgility™, which articulates the new paradigm of software development with the integration of Generative and Declarative AI, you can quickly apply Conway’s Law to understand the implications to the status quo tooling environments in the Agile world.  These tools are about to change drastically.

Obviously, there will be those with a vested interest in the status quo that will violently react to this wakeup call.