Will AI make management frameworks obsolete?

Every management framework makes the same promise to leadership, that if teams set goals in a shared format, meet in a fixed rhythm, and report progress in a standard way, the organization will move in one direction. Companies have built entire operating models around that promise, whether they call their version SAFe, Scrum, a quarterly planning cycle, or a homegrown mix that grew over the years, and many have invested heavily in training people to follow it.

AI agents now put that investment under pressure. They draft priorities, summarize progress, prepare meetings, and deliver work in a fraction of the time a team needed only a year ago, which means a growing share of the work a framework organizes no longer needs a person to do it. Anyone who owns a framework has to ask whether it still earns its place, or whether it has turned into a layer of ritual that slows the organization down.

Johannes Etzelmüller, founder and CEO of Workpath, explored that question in a recent webinar with Marco Alberti, founder of Murakamy, who has worked with goal-setting frameworks for more than fifteen years. Their conversation suggests a more precise way to frame it. Most parts of a management framework exist to make up for specific human limits, and AI removes some of those limits while it leaves others in place. The useful question is therefore which parts of a framework agents can absorb, and which parts carry the judgment that leaders still have to supply.

Management frameworks exist to make up for human limits

It's easy to treat a framework as a set of rules someone invented, but most of its rituals and artifacts trace back to the constraints of the people who use it. The clearest one is processing capacity. A leadership team can't read every project update or follow every metric, so the framework creates formats that compress information into something a person can absorb in a meeting. Status reports, scorecards, and the slide deck for the quarterly business review all exist because human attention runs out long before the information does.

Compression creates the second problem. Every time information moves up a level in the hierarchy, someone condenses it, and every condensation drops detail and adds a bit of interpretation. By the time a project update reaches the executive team, it has passed through several people who each decided what mattered, which is why frameworks add fixed reporting lines and review meetings that try to keep the story consistent from the team to the board.

The third constraint is speed. Annual plans and quarterly cycles match the pace at which human teams used to deliver, because a quarter gave teams enough time to produce something worth reviewing and gave leadership enough time to react before the next round of decisions. The rhythm of most frameworks reflects how long work used to take.

The fourth constraint is memory. People forget what they learned last cycle, and different leaders ask different questions each time they plan. Marco sees this as the core of every framework, which he describes as a checklist that makes sure the organization asks the right questions and learns from its earlier mistakes. He compares it to the skills people write for AI agents today, which work much like the standard operating procedures companies have used to guide their teams for decades.

Mapping a framework against these four constraints reveals something useful. Most of the visible ceremony, meaning the reports, the review meetings, and the fixed cadence, serves the first three constraints, and those are exactly the ones AI changes.

AI absorbs most of the work that compensates for those limits

The first piece of compensating work to go is drafting. Building a set of priorities used to take weeks of workshops, alignment loops, and edits, mostly because no single person could hold all the relevant information at once. With enough context about financial targets, team structure, and past performance, AI can already propose priorities along the whole impact chain that a leadership team would have reached on its own, and often faster. Johannes sees context as the main bottleneck, since most teams give the AI a fraction of what it needs and then conclude from the weak results that humans still do this work better.:

If the context is there, I think that in one or two years we won't need humans to draft these priorities.

Reporting follows the same path. When an agent can read every update in every system, the reason to compress information into slides gets much weaker, because the agent can pull the current state of any initiative straight from the place where the work happens. Marco takes this to its logical end and would remove PowerPoint from the organization altogether in favor of working from entries in a database, since every slide only condenses information that already exists somewhere else. Even companies that stop short of that step can see how much of their reporting effort goes into reformatting data for human readers.

Meetings are where this shift becomes most visible. A large share of recurring meetings exist to move information between people, and once agents handle that movement, those meetings lose their purpose. In a live poll during the webinar, most attendees estimated that their organization could drop more than half of its meetings. Johannes warns against cutting meetings first, though, and suggests starting with automation. The idea is to look at what happens before, during, and after each recurring meeting and hand every preparation and follow-up task to agents, from building the agenda to tracking decisions. That exercise shows which data the organization needs in structured form, and once that data flows on its own, the meetings that only existed to share it become easy to spot.

The deepest change hits the rhythm itself. An agentic workforce can deliver in hours or days what a human team used to plan for a quarter, which means the work now waits on leadership feedback. Workpath's own product engineering team already feels this shift, because working with Claude Code has made the company's quarterly rhythm a poor fit and pushed the team toward two- to four-week horizons, while other teams still plan by the quarter. That split will become common in many companies, and it makes a single company-wide operating rhythm harder to defend as delivery speed starts to vary from team to team.

Judgment becomes the part of the framework that matters most

If AI absorbs most of the compensating work, it would be tempting to conclude that frameworks are on their way out. The fourth constraint points the other way. Agents need structure as much as people do, because a clear checklist saves an agent tokens in the same way it saves a person energy, so the part of a framework that defines which questions to ask gains value as more work moves to AI. What changes is who works through the checklist and where the human contribution sits.

That contribution starts with intent. An agent can draft priorities once it knows where the company wants to go, why it wants to go there, and which trade-offs it accepts on the way. In many companies, that intent lives in the heads of a few executives and in slide decks from the last strategy offsite, which leaves agents to fill the gaps with plausible guesses. Johannes recommends that leadership teams test how clearly their organization defines and documents its intent, and whether the AI tools their teams use can access that documentation.

Intent also shapes how a company uses its new capacity. When delivery gets faster, the obvious move is to raise targets and expect more of the same work. Marco argues for a different use of that capacity, where teams place more bets on what creates customer value and treat each one as a hypothesis to test. That approach multiplies the number of decisions leadership has to make, because someone has to choose which bets deserve resources and which ones to stop when the evidence comes in.

Those decisions need clear owners. Marco points to Jack Dorsey's idea of moving "from hierarchy to intelligence," where small teams with clear ownership make decisions and an AI layer in the middle handles the coordination between them. The model sounds simple, but it depends on ownership people can exercise, and many organizations fall short here. A common example is a product owner who holds the title and still has no say over the product. Agents make that gap more costly, because they move fast enough to expose every decision that has no clear owner and pile up work in front of it.

The last piece of judgment is the least obvious, because it concerns judgment itself. Once AI does the drafting, people tend to stop forming their own view. According to Johannes, people accept more than half of the suggestions from Workpath's own AI Companion with a simple yes. That habit saves time in the short term, and over time it erodes the ability that makes leaders useful, since a leader who no longer forms an opinion also struggles to recognize a weak suggestion. Marco's answer is to change the order of the decision.

We have to turn the decision-making process around, so people form their own view first and use the model's answer to test it.

In practice, a leadership team writes down its own priorities before it looks at the AI draft and then spends its discussion on the differences between the two. The AI still does most of the work, and the team keeps exercising the judgment it needs for the decisions that matter.

Two mistakes that make frameworks worse with AI

The shift toward judgment only pays off if companies avoid two traps that AI makes easy to fall into. The first one concerns information. AI can make the old problem of information loss worse when people use it on both ends of a report. One person asks AI to expand a few bullet points into a polished report, the recipient asks AI to summarize that report again, and the content degrades with each pass until it says more about the model than about the work. The fix lies in the source data, since an intelligence layer that works with the full, uncompressed information from the systems where work happens gives decision-makers far better material than any chain of summaries.

The second trap is more tempting, because it looks like progress. Companies add agents to every existing meeting, report, and planning cycle, the old structure runs faster, and the organization ends up with more output and more noise while the underlying decisions move as slowly as before. Johannes sees this as the biggest risk of the coming years.

The biggest mistake companies can make in the next five years is to keep everything as it is, how they meet and how often they decide, and simply add AI agents on top.

How to rebuild your framework for agents

The four constraints give leadership teams a practical way to review their framework. The starting point is a full inventory of every recurring meeting, report, template, and planning cycle, including the informal ones that everyone attends and nobody owns. For each item, the team asks which human limit it compensates for and which decisions it supports. The second question matters most, because it separates the parts of the framework that move information from the parts that produce decisions.

A monthly business review makes a good example. It usually compensates for limited processing capacity and information loss, since teams spend days preparing slides that condense their numbers for leadership. The decisions it supports, such as shifting budget between initiatives or stopping a project that underperforms, are what make it valuable. Following Johannes's approach, the team first hands the preparation, the reporting, and the follow-up to agents and runs the review that way for a cycle or two. Once the numbers flow on their own, the team can see whether the meeting still needs its full length, or whether a shorter decision session held whenever the data calls for it serves the business better.

The same review should then extend to the rhythm itself. Teams that deliver in weeks need a cadence that matches their speed, and teams with longer delivery cycles can keep a slower one, as long as leadership can still connect both to the same intent. That connection depends on intent that leadership has written down, made clear about trade-offs, and opened up to the agents every team uses. The final step is more of a habit than a process, where leaders form their own view before they look at an AI suggestion and use the differences as the starting point for the discussion.

So, will AI make management frameworks obsolete?

Frameworks will stay, because humans and agents both work better with a shared checklist of the right questions. The balance inside them will change, though. The compensating layer of reports, status meetings, and fixed quarterly cadences will shrink as agents take over the work it existed to handle. The judgment layer of intent, trade-offs, ownership, and the choice of bets will carry more weight, because every agent in the organization depends on it. Companies that rebuild their framework around that judgment will get far more out of AI than companies that simply run their old processes faster.