70% of projects still fail to meet their goals, even as PMs spend 60% of their time on coordination overhead - chasing updates, running meetings, and manually following up. This post explores the background of software project management, how it's failed, and how AI project management asks a different question: "Can we automate the project management coordination work itself?"
It helps to look back at the history and current state of project management, as that defines what an AI Project Manager can and should do. That history starts with the Broken Promise of Project Management Software.
If you've ever felt like you're drowning in project management tools while simultaneously losing control of your projects, you're not alone.
The global project management software market is worth over $6 billion annually. Organizations use an average of 3-4 different PM tools. Teams have more visibility, more tracking, more dashboards, and more integrations than ever before.
And yet, according to the Project Management Institute, 70%of projects still fail to meet their original goals and business intent.
Here's the uncomfortable truth: we've spent the last two decades building increasingly sophisticated tools for tracking project work, but we haven't fundamentally changed how projects are managed. We've digitized the status report. We've colorized the Gantt chart. We've gamified the task list.
But we're still doing the same manual coordination work we did decades ago.
Every PM software promises to "streamline your workflow" and "keep your team aligned." The reality? Most project managers spend:
Think about that: 85-90% of a project manager's time is spent on coordination overhead, not value creation.
We've built tools that make it easier to see that a task is blocked. But they don't unblock it. We can visualize that a developer is overloaded. But the software doesn't rebalance the workload. We can track that a code review has been sitting for three days. But nothing reminds the reviewer or finds an alternative.
This is where AI project management fundamentally differs from everything that came before it.
Traditional PM software asks: "How can we give managers better visibility into what's happening?"
AI project management asks: "How can we automate the management work itself?"
The difference is transformational and has only become possible recently with the advent of modern large-language models.
In this series we'll explore what an AI project manager actually is, how it works, why current solutions fall short, and what it means for the future of software development. Whether you're a burned-out PM drowning in status meetings, a CTO trying to scale your engineering org, or a developer frustrated by coordination overhead, this guide will show you what's possible when project management becomes proactive instead of reactive.
Let's start by understanding how we got here.
To understand why AI project management is revolutionary, we need to understand what it's replacing. Let's trace the evolution of project management in software development - and why each iteration, despite its improvements, still left PMs drowning in coordination work.
In the beginning, there was Waterfall. And it made sense - sort of. Borrowed from construction and manufacturing, Waterfall project management treated software like building a bridge: extensive up-front planning, sequential phases, comprehensive documentation, and change control processes that treated any deviation from the plan as a failure.
Project managers in this era created
The coordination tax: PMs spent 60-70% of their time maintaining documentation that was obsolete almost immediately. Developers spent 20-30% of their time in meetings explaining why the plan needed to change.
The tools? Microsoft Project, Excel spreadsheets, email, and these days an infinite flood of Slack/Teams messages.
Why it failed: Software isn't like construction - in construction you can’t afford to have shifting requirements, because the cost of construction is high. But in software, requirements change precisely because they can – and so by the time you document everything perfectly, the business problem and software need has transformed.
Then came Agile, and it felt like liberation.
The Agile Manifesto(2001) promised: "Individuals and interactions over processes and tools." Shorter cycles. Continuous feedback. Adaptive planning. Empowered teams - these were the promises.
Scrum gave us
Kanban gave us:
This was genuinely better and teams shipped faster as a result. If we’re honest, teams shipped faster simply because the timelines shortened and required velocity increased. This occurred in a context where IDEs and frameworks improved as well, facilitating the velocity increase.
But here's what nobody warned us about: Agile didn't reduce coordination work - it increased the coordination frequency.
The Coordination Tax Rises with Team Velocity
Under Waterfall, a PM updated the project plan weekly. Under Agile, that PM now:
Let's do the math: That's as much as 8-12 hours per week in formal ceremonies, plus many additional hours of informal coordination. We've gone from 60% coordination overhead to... 60% coordination overhead, with the addition of higher frequency to frazzle the mind.
Agile made communication more frequent, but it didn't automate any of it, leaving more work on the PMs.
The Agile movement spawned an explosion of project management tools, each promising to be the answer:
The Task Management Generation:
The Collaboration Generation:
The Developer-Specific Tools:
Each tool promised to "reduce context switching" and "keep everyone aligned." And in isolation, many of them are quite good.
But here's what happened in practice:
A typical engineering team in 2026 uses:
That's not "reduced context switching." That's context switching as a service.
The Fallacy of "If We Just Track It Better, We'll Execute Better"
Every new tool made the same implicit promise: Better visibility leads to better execution.
The logic seems sound: If we can see all the work in one place, And we can see who's working on what, And we can see what's blocked... Then we can manage projects more effectively!
But visibility isn't the bottleneck. Action is the bottleneck.
Think about the last time you looked at your team's Jira board and saw:
You saw it. The dashboard showed you the problem. But then what? You still have to
The tool showed you the problem. You still have to solve it manually. This is the core limitation of traditional project management software: it's built on a reactive model. It waits for you to look at dashboards, notice problems, and take action.
The Manual-First Reality
Here's what a day-in-the-life of a project managemer looks like with modern tools, with coordination tasks highlighted in bold:
Total coordination time: 6 hours out of an 8-hour day. And this is a good PM having a normal day, with no crisis and no surprises!
The Manual Approach Can't Scale
This model breaks down as organizations grow. One team of 5-8people is manageable for a single PM. Three teams with 15-20 people total is a struggle, but still doable. Beyond five teams, and you just have to hire additional PMs and accept the overhead cost.
If the teams are jointly working on one larger project, the coordination work can grow exponentially with team size, even as PM capacity grows linearly (or not at all).
This is why organizations hit a scaling wall around 20-30 engineers. It's not that developers can't do the work. It's that project managers can't coordinate it all manually anymore.
Can AI Provide an Escape Hatch?
Traditional PM software gives you better visibility, better dashboards, better integration between tools. But it fundamentally assumes that a human will do the coordination.
A human will notice the blocked task and follow up.
A human will see the idle developer and assign work.
A human will realize the code review is delayed and ping the reviewer.
A human will detect the ambiguous requirement and request clarification.
What if that assumption is wrong?
What if the coordination work itself - the follow-ups, the status checks, the work assignments, the blocker escalations - could be automated?
That's not a better project management tool.
That's an AI project manager.
