A matrix for telling whether your productivity system still works

This post is part of a series on frameworks that combine human-centric organization with AI assistance. It comes out of my own setup, so read it as working notes rather than advice.

When a good framework goes stale

I know the feeling of looking at a carefully built Notion workspace, with the Drive folders in order and the task manager colour-coded, and still dropping things. The system that reorganized my work six months ago now feels like a constraint; the method that brought clarity last year now feels like busywork; the tools that saved me time at first now cost more upkeep than they return.

I do not think that is a failure of the frameworks. Needs change, work changes, and the technology underneath changes faster than either. The question worth answering is narrower: when is a system asking for small adjustments, and when is it asking to be rebuilt? That is what the Framework Evaluation Matrix is for.

The lifecycle

Every framework I have used has moved through the same five stages, and knowing which one you are in tells you a lot before you assess anything.

Honeymoon. The framework feels revelatory. You roll it out across your whole digital life, everything has a place, and you recommend it to anyone who asks.

Stabilization. The excitement settles into routine. The framework becomes second nature and its limits start to show alongside its strengths, so you make small adjustments.

Adaptation. The pain points sharpen. You build workarounds and extensions that keep things working and quietly add complexity while they do it.

Complexity creep. The original elegance disappears under the accumulated adaptations, upkeep grows, and you spend more time running the system than working inside it.

Breakdown or reboot. The framework no longer does its job. You abandon it, strip it back to something simpler, or evolve it into something that fits the work you have now.

I have gone around this loop several times. Tiago Forte's PARA reorganized how I kept information when I found it, and years later, with my work changed and AI tools in the middle of it, I was squarely in complexity creep: managing the system instead of using it. That is what led to OPERA, which extends PARA rather than replacing it.

What AI changes

The current shift is larger than a new tool, and I think it changes four things about what a framework has to do.

Knowledge stops being static. Traditional frameworks treat information as files in folders and notes in notebooks. AI tools work better against structures where context and relationships carry as much weight as the category something was filed under.

Organizing stops being manual. The assumption that a person does all the classifying, tagging and prioritizing is going away. Software can do a growing share of it, and a framework that has no place for that leaves the capability on the table.

Collaboration includes non-humans. A framework now has to accommodate an agent with a different set of strengths and failure modes than mine, which mostly means saying explicitly what it may decide and where it must stop.

Workflows stop respecting tool boundaries. Information moves across platforms constantly, and a framework that assumes one application becomes a point of friction.

This is roughly the transition from paper to digital that Getting Things Done had to make, when its physical folders became digital task managers. My impression is that this one is moving faster, though I would not put a number on it.

The matrix

The Framework Evaluation Matrix scores seven dimensions of a working system. It is deliberately diagnostic rather than a verdict: the output is the two or three dimensions that need attention, not a judgment that a framework is good or bad, because frameworks are personal enough that the same one genuinely works for one person and fails for another.

It takes about fifteen minutes of honest reflection. Longer than that and you are probably rationalizing.

The seven dimensions

1. AI integration capability

Frameworks used to assume a human workflow with technology in support. AI tools now suggest actions, generate content, analyze information and make decisions inside whatever parameters you set, which makes their place in the system a design question rather than an afterthought.

There are four kinds of integration worth separating. Automation covers routine work like email sorting and basic data processing, with little human involvement. Augmentation covers research assistance, editing and analysis, where the output still comes back to you. Orchestration covers coordinating a workflow across several tools. Collaboration treats an agent as something closer to a team member with a defined role.

Integration is weak when AI tools sit outside your main workflow and you are the one carrying context between them, when generated output has no structured place to land, and when you reach for AI reactively rather than as part of how the work runs. It is strong when agents can read from and write to your knowledge base directly, when there are stated rules for what an agent decides on its own, and when your framework names the roles and boundaries for AI assistance.

The quickest improvement is a dedicated input and output zone where AI tools can pick up context and drop results without disturbing anything else. This is the dimension traditional frameworks score worst on, for the obvious reason that none of them were designed with machine collaborators in mind.

2. Knowledge accessibility

Beautifully organized information is worth nothing if you cannot get it back, so a system should be tuned for retrieval rather than for storage.

Accessibility is poor when you recreate things you already have because finding them is harder than redoing them, when you have a vague sense that something useful exists somewhere in the system, and when retrieval depends on remembering the filing decision you made at the time. It is strong when there are several routes to the same thing (search, browsing, following a relationship), when context travels with the content, and when retrieval works the same way across tools.

The quickest improvement is consistent tagging across platforms plus a standard order of search methods you follow every time you go looking.

3. Maintenance overhead

Every system needs upkeep, and the ones that collapse are the ones where upkeep grows faster than value.

Overhead is high when you skip organizing because it takes too long, when "clean up the system" is a recurring task on your list, and when parts of the system have fallen apart and you feel bad about it. It is low when organizing happens as a side effect of working, when the system survives a few weeks of neglect intact, and when basic filing takes no real decision.

The quickest improvement is to find the organizational decisions that cost you the most friction and write a default rule that covers around 80% of cases.

4. Adaptability

Fixed frameworks break when the work changes shape; adaptive ones bend.

Adaptability is poor when new kinds of work produce orphaned information with no home, when you keep parallel systems for different roles, and when adding a tool means restructuring. It is strong when new categories fit without a rebuild, when principles rather than rigid rules decide where things go, and when the same structure handles a one-hour task and a six-month project.

The quickest improvement is an explicit space for experiments, so you can try a new workflow without touching the parts that already work.

5. Cross-platform consistency

Most of us work across several tools, and the framework has to hold across them.

Consistency is poor when each tool has its own organizing scheme, when information ends up trapped in one application, and when moving between platforms costs you a mental reset. It is strong when the same organizing principles apply everywhere, when information moves between platforms without manual reformatting, and when you can reach what you need whichever tool you started in.

The quickest improvement is a naming convention that is independent of any tool and works as a consistent address for the same thing everywhere.

6. Information processing capacity

Overload is a question of processing rather than volume, so a framework should help you triage, prioritize and pull value out of what arrives.

Capacity is low when unprocessed material piles up at your collection points, when you save more than you can ever read, and when the valuable items are buried under the merely interesting. It is high when different kinds of incoming information have different handling rules, when the system separates what needs attention now from what is reference, and when routine flows are handled automatically.

The quickest improvement is a triage rule based on two questions: can I act on this, and is it worth keeping if I cannot.

7. Operational orchestration

Beyond filing, a framework has to coordinate work that spans tools, platforms and automated steps.

Orchestration is poor when tasks and their supporting information live in different systems, when workflows stall at the handover between tools, and when you are the one tracking dependencies. It is strong when a task carries its context with it, when transitions between parts of the system have defined handling, and when the framework coordinates automated actions alongside your own.

The quickest improvement is to write down the two or three workflows you run most often, including where information moves and who or what moves it.

What to do with the result

Once you know which dimensions are weak, there are four sensible responses, in increasing order of cost.

Targeted improvements. If the framework scores well apart from one or two dimensions, add a component or change a process and leave the rest alone.

Hybridization. Take the organizing structure you already have and borrow a specific technique, usually orchestration, from another framework.

Transition. When several dimensions are weak, or when an assumption the framework rests on no longer holds, move deliberately to a different framework and carry the valuable content across.

Redesign. Rarely, when the work itself has changed or a new capability makes something possible that was not before, design from first principles.

Which one fits depends on how many dimensions need work, whether the framework's founding assumptions still hold, how much content is invested in the current system, how much change you can absorb right now, and how stable your workflows are.

OPERA as a worked example

OPERA came out of running this assessment on my own PARA setup. Two dimensions scored badly, AI integration and operational orchestration, and the rest held up, which pointed at extension rather than replacement. The Operations layer addresses those two specifically: defined agent roles and workflows for the first, automated organization and curation for the maintenance cost, and workflow templates for the third. PARA's adaptability and cross-platform consistency stay as they were, because there was nothing wrong with them.

Forte has reached a similar conclusion from his own side. In his piece on AI and Building a Second Brain, he argues that AI will not remove the need for a second brain but will change how it is used, and that it is strongest at the Organize and Distill phases of his CODE process, automating what he calls low-value housekeeping while a person still does the capturing and the expressing. That matches what I am trying to do with OPERA: keep the part of the existing framework that earns its place, and hand the mechanical work to software.

I would extend a framework before replacing one, in most cases. The content you have already invested stays where it is, and you can adopt the change a dimension at a time instead of rebuilding everything at once.

Keeping it up

Framework evolution does not finish. Technology moves, the work moves, and what you need from a system moves with them, so I would rather assess regularly than hunt for a perfect framework. Cal Newport's argument for deliberate system design over chasing every new productivity trend applies here: regular review keeps the friction from accumulating to the point where only a rebuild will do.

The matrix is a maintenance tool more than a crisis tool. A quarterly pass catches gaps while they are still small, and small regular adjustments are cheaper than an overhaul.

It also works well on a combination. There is no reason not to keep GTD's capture discipline, Pomodoro's focus intervals and PARA's knowledge organization at the same time, and add AI to whichever of them is costing you the most manual effort.

Using it

Open the Framework Evaluation Matrix, give it fifteen minutes, score the seven dimensions, pick your response from the four above, and write a concrete plan for the one or two dimensions that scored lowest. A perfect score is not the goal and is probably a sign you scored yourself generously.

Later posts in this series go through the dimensions one at a time, and further into how OPERA's Operations layer works in practice.