Failing Faster? What AI Can(not) Do for Your Organisational Structure

Source: https://www.team17.com/games/overcooked

I have had numerous conversations recently on the topic of AI and how companies should approach it. What they all have in common is the immense pressure, from both inside and outside, to do everything even faster and, above all, even better right now.

Faster, higher, further, or something like that... yet all too often, people forget: missing structures, a lack of focus, and flawed communication are not established or solved with AI. Quite the opposite. AI ensures that our past oversights come back to bite us in real time.

From Agile Dream to Structural Challenge

An example: an international software scale-up has been struggling for some time with the challenge of wanting to grow both economically and in head count without having established the necessary structures to do so.

At the start of the growth phase, the pockets of money and motivation are bulging. People want to get serious about organisation and structure now. After all, they are no longer a 5-person start-up, but a multinational company doing business with the big players. So far, so ambitious.

What is immediately noticeable, however: a colourful bunch of specialised individualists is at work here. Everyone works fully remote, self-organised, and often "async-first". This setup works well for many software companies and is (understandably) popular with just as many employees. The big sticking point, however, is: what happens when this setup no longer fits the company strategy?

The Quiet, Creeping Resistance

You might think there would be an outcry when the initiative to establish shared structures and processes is kicked off. But precisely because employees are scattered around the world, the big outcry never happens. The forum for it (for example, at the coffee machine or in the remote daily stand-up) simply does not exist. Nor are there any moves to change that. Resistance to building official rules, processes, and structures therefore often takes place quietly and insidiously.

Meetings scheduled to collaboratively develop processes and guidelines regularly turn into endless discussion rounds with no binding final result, let alone real commitment. As soon as a shared online room is closed, everyone goes back to doing their own thing. Management included. Individuals tilt at windmills while the leadership team hopes that employees will automatically adhere to processes that were once discussed.

Spoiler: Forget that! Change takes time, active engagement, and must be consistently modelled and supported by management. With everything that entails.

Authenticity: Often Scorned, but Highly Effective

Management, too, must adapt, rethink existing ways of working, and realign itself. Open, clear, and authentic communication is, unfortunately, an all too often scorned tool in this process.

We frequently underestimate how authentic words affect others and what they do to ourselves. Suddenly, change is no longer dictated from above, but is experienced by that very same "above". Authentic communication makes us approachable and human. It is a pity that authenticity and humanity are still so often equated with touchy-feely fluff.

I am convinced that this is precisely what sets us apart from machines and will keep us relevant in the future – as long as we allow it.

Growth at All Costs: The Direct Route to Chaos

Back to our scale-up: change is kicked off, and people expect new structures to establish themselves permanently on their own. In parallel, the development team is doubled overnight just like that – after all, the company wants to grow. On top of that, key individuals leave the company, partly due to the non-existent structure pushing them to the brink of burnout. The shock is immense. The frantic activism even greater.

With double the team size, product topics are tackled synchronously and without structure. After all, growth is imperative. More features = more customers = more revenue. Right?

Market fit is also short-circuited: whoever screams the loudest gets what they want. What a strategy!

Before the team can get used to any of this, let alone reform, missing key roles are replaced in a downright hiring sprint.

The chaos is complete, and the company has lost vital months and resources (money, time, personal dedication, commitment) in the meantime. The team is thoroughly frustrated. The tentative beginnings of a genuine team have turned back into specialised lone wolves. Working on the system is swapped for constant operational firefighting.

And then comes AI... and everything gets better... not!

AI as an Accelerant

AI, disguised as a knight in shining armour, divides the fragile team even further. The trenches that were painstakingly filled in split open deeper than before. Suddenly, the sentiment is: "Either you are for AI, or you are against us."

A group of AI enthusiasts among the developers charges ahead regardless of the consequences, pinning to their banner the promise to deliver new features faster and better. The AI sceptics are perceived as backwards and obstructive. In no time at all, old waterfall dynamics return at an unprecedented pace.

Product Management, Business Analysis, and UX are under massive pressure to define new requirements – 100% complete, correct, and pixel-perfect, if you please. Otherwise, the AI is all too happy to hallucinate complete absurdities into the slightest leeway for interpretation.

QA experts no longer have any idea what a feature is supposed to do – or not do. Documentation that was already neglected is diluted even further by vibe coding. Instead of having deficient information as before AI, QAs now have to decipher from a pile of AI slop what – if anything – actually applies.

More AI, More Stress: The Automation Trap

The apparent solution is obvious: PM, QA, and basically everyone just need to use more AI. That way, everyone gets faster – problem solved. Seriously? No, the problem is not solved at all.

Now, even more people across all disciplines are not only constantly stressed and frustrated, but also endlessly dissatisfied with themselves. They are perpetually too slow, too inaccurate, a blocking element in the fast lane to the promised AI land.

Who wouldn't ask themselves: "What am I actually doing here? What good am I as a human being anyway? I can't even manage to tell a machine what it should do. Of course that very machine is going to take my job away in no time."

Developers wrestle with AI-generated spaghetti code. Many a complete beginner places more value on maintainability and security than our highly praised AI colleague. Every tiny detail has to be explained to our little AI helper. Code reviews are either not conducted at all or carried out fully automatically.

Thanks to vibe coding and AI code reviews, a corresponding amount of output lands on the QAs' plates. The QAs, in turn, generate and automate even more test scenarios with AI assistance. But are all these tests even necessary? Are they the right tests? One hardly dares to ask these questions out loud. More is more, isn't it?

And soon enough, nobody is doing what brings them joy anymore; everyone is just chasing after AI.

Shit In, Shit Out: The Return to Real Strategy

At the beginning of this new, familiar waterfall, product management still stands (optionally supported by PO, BA, RE, UX, etc.). Shouldn't they be capable of preventing this chaos? Theoretically and in part, yes.

The fact is: without clear prioritisation and focus, there is no real product strategy.

A product strategy enables (product) management to say no. Because if everything is important, nothing is really important. In the end, you are left with a bunch of loose ends, and before you know it, you are out of the market.

And so, after a few months of AI hype, our scale-up stands at a crossroads. Now, focus, clear communication, and structure are required. What's more, expectations around speed and scope of delivery have not diminished.

In product management, it quickly becomes clear: it cannot go on like this. Shit in, shit out!

Every AI-generated requirement only generates more slop in the process and comes at the expense of the staff.

Quality Remains a Team Sport

Across all levels, it becomes clear: defining viable and sustainable strategies must come from within, from the organisation itself. AI makes working systems more efficient – and chaotic systems simply chaotic even faster. Anyone hoping that an algorithm will smooth over a missing organisational foundation is building a house of cards in a storm.

What is needed now is no further frantic activism, nor a new tool. It takes the courage to pause, the courage to prioritise, honest communication at eye level, and clear, lived structures that provide orientation instead of creating bureaucracy. Technology can relieve us of routines, but never of the responsibility for a viable togetherness and a clear direction.

In the end, quality – both organisationally and technically – remains a team sport that we as humans must shape.