Two roads, one infinite game, and the case for taking SAFe Principle #8 seriously
When a new technology arrives, the same four reactions appear in the same four places.
On the worker’s side of the table, some fear the technology will end their livelihoods, while others hope it will enable them to do work they could not do before. On the leadership side, some see a way to reduce reliance on people, while others see a way to redirect their people toward work that demands more of them.
I have watched all four reactions unfold in ITSM classrooms for fourteen years. Students who arrive convinced that automation is coming for them often ask, often quietly during a break, whether their role has a future. Students who arrive convinced that automation is an opportunity come back with sharper questions about how their team could absorb the time it might free up. The managers in the room split along the same lines. Some calculate cost savings before the morning coffee. Others sketch new operating models on the whiteboard during lunch.
Each of these reactions has a defensible core. None of them is wrong about everything. They yield very different organizations.
That is the first observation worth sitting with: AI is not handing us a destiny. It is handing us a choice we have long avoided.
Two roads in a yellow wood
Robert Frost’s “The Road Not Taken” is widely misquoted as a celebration of choosing the unconventional path. Read carefully, the poem says something quieter and more honest. The two roads, the speaker tells us, are worn “really about the same.” He takes one, knowing he probably will not come back to try the other. The famous closing line, “I took the one less traveled by, and that has made all the difference,” is offered in advance as the story he will tell ages hence, not as a verdict he has already reached.
This is what technology choices look like in real organizations. They are made in meetings that do not feel historic. The two paths look similar in the moment. Decision-makers will tell a clean story about it later, after the difference becomes clear. Leaders making decisions about AI are standing at this kind of fork now. One path uses AI to do less with people. The other uses AI to ask more of them. Both are available. Both will produce a story. Only one of them produces an organization worth being part of in ten years.
A pattern older than computing
The Luddites, often maligned and misunderstood, were not opposed to machinery in the abstract. They opposed mill owners’ use of machinery to suppress wages and undermine skilled workers’ bargaining position. Their fear was not of looms but of what looms made possible in the hands of leaders who saw them primarily as a means to do without skilled workers.
We are not Luddites. The technology before us differs, and so does the surrounding economy. The pattern of reactions, however, remains the same. Workers who have spent years honing a craft watch a machine begin to perform part of that craft and reach for plausible interpretations of what it means. Leaders responsible for their business’s financial performance watch the same machine and reach for plausible interpretations of what it could mean for their margin.
What is new is the breadth of work the current generation of tools can handle. Previous waves of automation focused on physical labor and the most repetitive cognitive tasks. The current wave extends to reading, writing, analysis, synthesis, and decision support. In other words, it reaches the heart of what most knowledge workers do for a living.
This is what makes the question harder than it was last time and what makes the answer more interesting.
What SAFe says about knowledge workers
The Scaled Agile Framework outlines ten guiding principles. The eighth is to unlock the intrinsic motivation of knowledge workers, and it is the one I reach for most often in these conversations.
The principle rests on an empirical claim that has been rigorously tested for decades and popularized by Daniel Pink in Drive. The claim is that for complex, non-routine work, traditional management practices (specifying the task, monitoring compliance, paying for output) do not produce the results we want. They suppress the kind of thinking the work requires. Pink summarized decades of behavioral research showing that contingent rewards, when applied to creative work, reliably reduce both quality and intrinsic interest. This finding has held up across cultures, age groups, and task types and is among the most replicated results in organizational psychology.
The principle proposes a different stance. Knowledge workers respond to three things in particular. Autonomy: the discretion to manage their own work and decide how it is done. Mastery: the opportunity to improve their skills over time. Purpose: a clear connection between their daily work and something larger than themselves.
When organizations provide these three, the work tends to be both better and less expensive than the alternative. When organizations suppress them by over-specifying the task, micromanaging execution, and tying rewards narrowly to short-term output, the work tends to be both worse and more expensive. Deming made the same observation forty years ago, using different vocabulary. The intuition that we get what we pay for is correct, though not in the way the speaker intended. Pay narrowly, get narrowly.
I keep returning to this principle because it gives the AI conversation a frame that does not require speculation about future capabilities. We do not need to know whether AI will eventually do everything. We need to know how to use the work it can already do. SAFe’s answer is clear: hand back the parts of the job that suppress intrinsic motivation and give people more room to exercise the parts that depend on it.
What AI gives back
Consider the parts of typical knowledge work that AI is already reasonably good at. Drafting first versions of routine documents. Summarizing long materials. Reformatting data. Translating between formats and registers. Producing boilerplate code. Answering factual questions with well-documented answers. Generating variations on a structured artifact.
None of these tasks build mastery in a meaningful way. They consume the time that mastery requires. They offer no autonomy worth the name; they are mostly compliance with someone else’s request. They sit at the far end of the work from purpose; no one chose their career to write more status reports.
If we are honest about it, these are exactly the parts of knowledge work that workers complain about and that managers reluctantly tolerate. They are the parts that crowd out the work that pays the rent for the discipline.
AI is increasingly able to take that work off the plate. Whether it actually does so depends entirely on what we put in its place.
A worker who is freed from four hours a week of previously tedious work can spend those hours on three different things. They can do more of the same work in the same time, which yields no improvement in motivation or mastery. They can be asked by a manager who has decided the time should go to throughput to do four more hours of something equally tedious. Or they can spend their time on work that builds capability, deepens judgment, and connects more clearly to the organization’s larger purpose.
Only the last option delivers the value SAFe Principle #8 promises. The other two waste it.
ITIL’s parallel guidance
I have spent most of my career teaching ITIL, a different tradition with similar instincts on this question. ITIL’s guiding principles include “Optimize and automate” and “Focus on value.” Taken together, they describe an organization that should be enthusiastic about AI, because AI enables the organization to do more of what those principles call for.
“Optimize and automate” advises us to automate everything technology can handle. The qualifier matters. It does not say to automate everything that can be priced. It says to automate what the technology can do well and to use human attention where it adds the most value. The principle is careful about sequence as well: optimize first, automate second. Automating a broken process produces a faster broken process. The qualifier rules out one of the more common failure modes I have seen in the field.
“Focus on value” advises eliminating anything that does not link back to stakeholder value. This principle gives organizations permission to stop doing work that has accumulated by inertia and no longer serves a purpose. It also gives organizations permission to stop measuring people on activities that do not connect to value.
Taken together, these principles describe a form of organizational hygiene. Automate the repeatable, eliminate the wasteful, and let people work on what remains. SAFe Principle #8 supplies the missing piece: what remains depends on autonomy, mastery, and purpose. The two frameworks arrive at the same place by different routes, which is one reason I trust the conclusion. When two disciplines converge from different angles, the answer is usually worth taking seriously.
What the positive version looks like in practice
The positive version requires a few specific commitments.
It requires leaders who treat AI as a means to eliminate drudgery rather than people. Accounting can support either choice, and the choice is strategic rather than financial. Organizations that eliminate drudgery tend, over time, to attract people interested in more challenging work. Organizations that eliminate people tend, over time, to attract people interested in the next job. The compounding effects diverge quickly.
It requires investment in mastery. The half-life of any particular skill is shrinking. The underlying capability for learning is the durable asset, and that capability requires time, attention, and tolerance for the early stages of being bad at something. Organizations that protect this time will get the practitioners they need. Those that do not will not.
It requires articulating the purpose clearly enough that people can make sound local decisions when no script applies. As AI handles more of the scripted work, the work that remains is the work where scripts do not exist or do not apply. People will do that work well only if they understand what the organization is trying to achieve and why it matters.
It requires granting teams genuine autonomy over their working methods. The tools are changing faster than any central function can keep up. The people closest to the work are the only ones with the information needed to redesign it effectively. Centralizing those decisions is a sure way to fall behind quickly.
And it requires patience. None of these commitments pay off in the same quarter they are made. The organizations that get the most from AI over the next five years will be those that resist the temptation to claim savings in year one.
The honest fear, the honest hope
The fearful reading of AI is not paranoid. Organizations have, more than once in living memory, used new technology primarily to reduce headcount, leaving operations fragile, demoralized, and struggling to adapt. The 1990s ERP wave produced a long list of cautionary cases. The early 2000s outsourcing wave produced more. The pattern repeats because short-term incentives reliably point in that direction, even when long-term consequences argue otherwise.
The hopeful reading is also available, and it is the one worth working for. It depends on leaders making a different choice than the short-term incentives would favor. It depends on workers being ready to fill the space that opens up when drudgery is removed. And it depends on both sides understanding that the new arrangement asks more of each of them than the old one did.
That last point bears emphasis. The hopeful reading is not a soft option. Autonomy is harder than being told what to do. Mastery is harder than coasting on yesterday’s skills. Purpose is harder than going through the motions. The bargain described in SAFe Principle #8 is demanding. It rewards both sides, but it asks something of each side first.
The infinite game
Simon Sinek’s book “The Infinite Game” frames the key difference. In his terms, finite players play within fixed rules to achieve a defined victory. Infinite players play to keep the game going and to improve the conditions under which it can be played. Business, like life, is an infinite game. The leaders who use AI to win the current quarter are using a powerful tool to play the wrong game. The leaders who use AI to extend the conditions under which their organization can continue meaningful work are using the same tool to play the right one.
The two stances look similar in year one. They look very different in year five. By year ten, only one of them is still in business. Compounding does not favor the finite player, because the finite player keeps having to find new staff, new processes, and new customers each time the previous round’s optimization has hollowed out the next round’s capacity. The infinite player accumulates capability, judgment, and trust, and uses each successive wave of technology to extend rather than exhaust those assets.
SAFe Principle #8 is, in this sense, infinite-game advice. It tells leaders to invest in the things that compound over rounds: autonomy that builds local judgment, mastery that builds durable skill, and purpose that builds shared direction. None of these can be acquired in a single quarter. All of them, once present, make every subsequent round of the game easier to play well.
The choice in front of us
Leaders will choose. The choice is whether to use AI primarily to reduce reliance on people or to redirect people toward work that depends on autonomy, mastery, and purpose. Both options are defensible in the short term. Only the second option will pay dividends with compound interest.
Workers will also choose. The choice is whether to treat AI as a threat to resist or as a tool to claim. Resistance is reasonable, but it is also a losing position; the technology is not waiting for permission. The more useful response is to build mastery in the parts of the work AI cannot do, to ask for the autonomy needed to do that work well, and to hold employers accountable for the purposes they claim to serve.
Neither party can fully act on its own initiative without the other. Leaders who try to unlock intrinsic motivation in a workforce trained to wait for instructions will find the process slower than expected. Workers who try to claim autonomy from leaders unwilling to grant it will find the process more punishing than hoped. The work is mutual.
Closing
The honest reading of the moment includes both fear and hope, and it would be dishonest to pretend that one is more reasonable than the other. Fear has historical precedent, and hope has good arguments behind it.
SAFe Principle #8 contributes to the conversation by offering a clearer description of what the hope can be about. It is not the hope that the technology will produce a utopia. It is the hope that we will use the technology to give knowledge workers the conditions under which they do their best work. Those conditions have a name. They are autonomy, mastery, and purpose. AI is offering, more clearly than any prior technology, to hand them back.
Whether we take them is a separate question, and it is the one worth addressing. Frost’s traveler looked down both roads as far as he could. We can see further if we choose to look. The fork is real, the difference compounds, and the story we tell ages hence will depend on which road we take now.
This article was first published on Substack on May 25, 2026.