The Second Coming of Waterfall

I spent most of my career as a technical delivery manager, a term I use deliberately because it covers many schools, and the best of us lived in all of them. I earned my PMP in my thirties when it was a source of real pride. Later I learned to keep it in my pocket. In the rooms I worked in, admitting you were a certified project manager marked you as the enemy of everything new. So I became an agilist too. Product Owner, Scrum Master, Agile Coach, certificates from the Scrum Alliance and Scrum.org. I believed in it, and I still do.

Now I hear people saying agentic teams mean Agile is dead. I remember saying that about waterfall. So permit an old man the luxury of his earned experience: not as much has changed as we think.

The Goals Never Moved

Every delivery method ever invented is trying to maximize the same three pillars: speed, function, and quality. We do this by balancing two masters that have been at war since the first project manager drew a Gantt chart on a cave wall. Agility and control. They want different things. They always have.

Waterfall served the second master. We retired it because what it gave us in control, it took from us in agility. And if we are honest with ourselves, which I am trying to be, the control was often false anyway. The baseline plan was a prediction, and predictions rot. By the time the requirements document was signed, stamped, and filed in triplicate, the world it described was already gone. We knew this. We documented it anyway. Then we blamed the team when reality refused to match the paperwork.

I did this. I enforced baselines that I knew were fiction because the process said to enforce them. I am not proud of it, but I am not going to pretend I was always the enlightened coach I later became.

Here is what I notice now, though. Agility has been won so completely that the demands I hear from leaders are for exactly what waterfall once promised. Visibility. Traceability. Artifacts. Gates. The pendulum is not swinging back because agile failed. It is swinging back because agile succeeded, and success at speed creates a hunger for control that we spent twenty years teaching ourselves to distrust.

The Trade That No Longer Holds

The Agile Manifesto told us to value individuals and interactions over processes and tools. That was the right call in 2001. Process was expensive. Every artifact was a person’s afternoon. Every gate was a meeting that could have been an email, except we did not have email culture yet, so it was a meeting that could have been nothing at all. Every status report was time stolen from the work itself.

When the scarce resource is human intelligence, you spend it on the product, not the paperwork. So we did, and we were right.

But does “people over process” apply to agents?

I do not think the authors of the manifesto ever imagined a world where the cost of process approaches zero. They were solving for the expensive thing, and the expensive thing was documentation, coordination, ceremony. All the stuff that slowed us down. All the stuff that made waterfall feel like swimming through concrete.

When an artifact maintains itself, when a gate enforces itself, when a status report assembles itself from the actual work in real time, the trade the manifesto made no longer holds. We traded away control because control was expensive.

It is not expensive anymore.

A Modest Proposal

So here is my thought, offered with the humility of someone who has been wrong about the future of software delivery at least three times. Perhaps we are finally agile enough for waterfall.

Not waterfall’s sequencing. That stays dead, and good riddance. Nobody misses the six-month requirements phase followed by the six-month design phase followed by the panicked realization that the client changed their mind in month two. Sequential phases were a response to expensive change, and change is cheap now. Let them rest.

What can return is waterfall’s discipline. The risk register. The dependency log. The change control board. The toll gates. All the things we threw out because maintaining them cost more than they saved.

Except this time, continuously true instead of baselined. Enforced in real time by agents instead of reviewed quarterly by armies. And here is the part I like best: maintained somewhere humans never have to see them.

The Agentic PMP

Imagine an agentic project manager. It needs no meetings. It holds no baseline versions, because it does not predict. It observes. It keeps every artifact current from the actual work, not from what someone remembered to update last Friday. It enforces the gates the moment they matter, not at the phase review three months later when the problem has already metastasized into six other problems.

Beside it, a risk manager that never sleeps. A security reviewer that reads every change. A QA lead that watches every build. All of it reporting in real time, and none of it costing a single human hour that could have gone to the product.

I know what you are thinking. You are thinking that I have just described a panopticon. You are thinking that surveillance is not the same as governance. You are right. But stay with me for a moment.

The thing that made waterfall oppressive was not the discipline itself. It was who the discipline was aimed at. We documented the work to control the workers. We held gate reviews to catch people, not problems. We treated process as a substitute for trust, and the workers felt it. They resented it. They were right to resent it.

But what if the discipline is aimed at the agents instead?

Process Over Agents

If people over process was the right value when people did the work, then maybe the new mantra, now that agents do so much of it, is this: process over agents.

The agents get the discipline. The humans get the judgment. We finally separate the two things Waterfall always confused. Control of the work and control of the workers. We give the machinery the first and keep the second for ourselves.

I am not saying this is easy. I am not even saying I know how to do it. I left the day-to-day a while ago, and my colleagues who are still in the trenches are working these problems with an intensity I remember from my own years of late nights and whiteboards. It is their turn. They will figure out what I am only guessing at.

But I spent a career watching us swing between the two masters, agility and control, and I will say this from the porch: the time boxes are infinitely shorter now, but the goals are exactly what they were the day I sat for my PMP. Speed, function, quality againts the clasic constraints of scope, time, and cost. Balance the masters. Serve the product.

We did not kill waterfall because its questions were wrong. We killed it because answering them cost too much.

The cost is dropping fast. The questions are back.

I could be wrong about all of this. I have been wrong before, and I will be wrong again. But if I had to bet, I would bet that the kids coming up now will rediscover things I was taught to forget, and they will wonder why we ever threw them away.

The pendulum swings. It always swings. The trick is to stop pretending we can hold it still.