The Personal Contract
Essay 11 of the AI Contract Series
What to protect when AI is everywhere and how to know when you are drifting
There is a version of the AI conversation that stays abstract.
Civilizational stakes. Social contracts. Monitoring frameworks. Failure archetypes at scale. These things matter. This series has made the case for why they matter. But somewhere in the architecture of the argument, the most personal version of the question can get lost.
You. Specifically. The particular human reading this, with a specific body of work, a specific set of relationships, a specific way of thinking that is yours and not anyone else’s. What does AI do to you? What are you protecting when you protect your Agency? What does drift look like in your life, at the level of actual daily decisions? How do you know when it is happening?
That is what this essay is about.
The drift problem
Drift is the right word because it describes the mechanism accurately.
You do not decide to become dependent on AI. You do not wake up one morning and choose to let your thinking atrophy, your creative synthesis weaken, your tolerance for productive difficulty diminish. You make a series of small, individually reasonable decisions that collectively move you somewhere you did not choose to go.
Today you use AI to draft an email you were going to write yourself because you are tired and the deadline is real. Reasonable. Today you use it to summarize a report you would have read yourself because there are six more reports in the queue and you are behind. Reasonable. Today you use it to generate the options for a decision you would have thought through yourself because the meeting is in forty minutes and you need something to bring. Reasonable.
Each decision is defensible. The accumulation is what matters.
Drift is invisible at the transaction level. It only becomes visible at the trajectory level — when you notice, months later, that you feel less confident in your own thinking, that the gap between what you can do with AI and what you can do without it has widened, that you have not written a piece of sustained original thinking in longer than you can easily account for.
By then, the drift has been underway for a long time. The decisions that produced it felt reasonable in every moment. None of them felt like a choice to become less capable.
That is the drift problem. It requires a personal contract — a set of commitments about what you will protect, made when you are clear-headed, that holds against the accumulating pressure of individual reasonable decisions.
What is actually yours
Before writing the personal contract, you need to know what you are protecting. Not in the abstract but in your specific life and work.
Three categories of capacity are worth identifying.
The work that requires your specific perspective. Not tasks that happen to be assigned to you, but work where what makes it valuable is that you did it — where your accumulated experience, your particular way of seeing, your specific aesthetic sensibilities or analytical frameworks or domain knowledge is what the output depends on. The thought leadership that only you can write. The judgment calls that require the specific pattern recognition you have built over years. The creative work that carries your voice and not a plausible approximation of it.
This is the category most at risk from drift. It is also the category where AI assistance is most seductive, because the output from AI-plus-you often looks similar to the output from you-alone. The difference is in what is being built — or not — in the process of producing it.
The capacity you are currently developing. Not skills you have already built to fluency, but the ones you are in the middle of building. The kind of writing you are not yet good at but are trying to get better at. The analytical domain you are expanding into. The creative form you are learning. The judgment you are developing through the accumulation of decisions and their consequences.
Development requires difficulty calibrated to your current capability. It requires the encounter with the gap between where you are and where you are trying to get. AI assistance that removes that gap removes the development mechanism. The capacity you are trying to build does not build.
The relationships that depend on your full presence. Not all relationships are in this category. But some — the collaborations that require genuine co-thinking, the mentorships that require you to actually listen and respond to what is in front of you, the creative partnerships where the other person needs to encounter your real thinking — are ones where AI-mediated presence is not a substitute for actual presence.
This category is the easiest to compromise invisibly. Using AI to prepare responses, to generate talking points, to pre-synthesize what you will say: these practices can produce the appearance of engagement while replacing its substance. The other person may not notice. The relationship is still affected.
The personal contract: five commitments
The following five commitments are not universal rules. They are a framework from which to build your own — adjusted to your specific work, your specific development priorities, your specific relationships. What matters is that you make them explicitly, before you are tired and behind on a deadline, and that you hold them against the pressure of individually reasonable exceptions.
Commitment 1: I will not outsource the work that is mine specifically.
Identify the work — specifically, by name, by type — that depends on your particular perspective, experience, or voice. Put it in writing. This is the work you will not hand to AI at the generative or synthetic stage, regardless of time pressure, regardless of how good the AI output would be, regardless of how reasonable the exception seems in the moment.
This commitment is not about using AI for nothing in your signature work. It is about protecting the parts of the process where your specific contribution lives. The research aggregation may be fine to delegate. The synthesis is yours. The editing support may be fine to use. The finding of the argument through the act of writing is yours.
The commitment is made at the level of the phase, not the task.
Commitment 2: I will protect the development processes that are building me.
Identify what you are currently in the middle of building — the skills, the judgment, the expertise at its growing edge. For each of these, identify specifically where the productive difficulty lives: the part of the process that is hard and uncomfortable and that is exactly where the growth is occurring.
Commit to protecting those specific difficulties. Not all difficulty in your work. The developmental ones. The ones where the struggle is the mechanism, not an obstacle to it.
This commitment requires honesty about the difference between productive difficulty and unproductive friction. Productive difficulty stretches your current capability. Unproductive friction wastes time without building anything — formatting, administrative tasks, searching for information you know exists. AI is excellent at removing unproductive friction. The commitment is to distinguish clearly between the two and protect only the first.
Commitment 3: I will maintain a body of work I produce without AI.
Separate from whether you use AI heavily in your professional work — maintain a practice, however small, of producing work entirely without it. Writing, thinking, creating, deciding. Something that is fully yours, unassisted, produced through your own generative and synthetic capacity.
This commitment serves two purposes. First, it keeps your unassisted capacity from atrophying entirely. Like any capacity, if it is never exercised it diminishes. Second, it gives you an honest benchmark — a way to know what you can actually do, so that the gap between your AI-assisted capability and your unassisted capability remains visible and by choice rather than invisible and by drift.
The size of this practice matters less than its consistency.
Commitment 4: I will be present in the relationships that require it.
Identify the relationships in your life where your genuine, unmediated thinking is what the other person needs. Then commit to being fully present in those relationships — which means not using AI to pre-generate your responses, not letting AI synthesis replace your actual listening, not substituting the appearance of engagement for the real thing.
This commitment has nothing to do with whether AI is present elsewhere in your life. It is about maintaining the distinction between relationships where mediated presence is acceptable and relationships where it is not. Making that distinction explicit — naming the relationships in the second category — is what allows you to protect them.
Commitment 5: I will audit myself honestly on a regular cadence.
The drift problem is a visibility problem. Drift is invisible at the transaction level. It only becomes visible at the trajectory level. Which means you need a regular practice of looking at the trajectory.
Set a cadence — monthly is reasonable, quarterly is the minimum — and ask yourself honestly: where have I drifted? What work that should have been mine did I hand over? What development process did I shortcut? What relationship did I mediate when it needed me directly?
The audit is not punitive. It is diagnostic. The point is not to feel bad about drift. It is to see it clearly enough to recalibrate before it compounds to somewhere you genuinely did not choose to go.
How to know when you are drifting
The audit cadence is planned. But drift also produces signals in real time, if you know what to look for.
You feel less confident in your own judgment. Not humble — specifically less certain that you can trust your own thinking without an AI check. If the default before any decision is “let me see what AI says first” rather than “let me think about this and then check my thinking,” the balance has shifted past the point where AI is supporting your judgment and into the territory where AI is substituting for it.
The gap between what you can do with AI and without it is widening. This one requires the unassisted practice from Commitment 3 to be visible. If you have maintained it, you will know. If you have not, this signal is invisible — which is itself a form of the signal.
You are avoiding the hard parts. Not sometimes — as a pattern. The places in your work where the productive difficulty lives, you are finding ways to route around them consistently. The synthesis you are handing to AI. The draft you are letting AI start because facing the blank page is uncomfortable. The decision you are outsourcing because sitting with the uncertainty is hard. When the avoidance of productive difficulty becomes a pattern rather than an occasional exception, drift is active.
Your original voice has gone quiet. This one is subtle and usually the last to be noticed. If you cannot remember the last time you wrote something that surprised you — something where you did not know what you thought until you wrote it — that silence is a signal. Original voice develops through the act of finding it. If you have not been doing the finding, it has not been developing.
The personal contract is not a sacrifice
It is worth being direct about what this essay is not arguing.
It is not arguing that you should use AI less. It is not arguing that the costs of AI use outweigh the benefits. It is not arguing for a Luddite relationship with the most powerful productivity tool most people have ever had access to.
It is arguing for consciousness. For the explicit decision, rather than the accumulated drift. For knowing what you are protecting and why, so that the decision to use AI — made constantly, in every interaction — is a real decision rather than a reflex.
The personal contract is not a restriction. It is the structure that lets you use AI extensively without losing the things that make you specifically valuable: your particular perspective, your developing judgment, your original voice, your capacity for genuine relationship.
Without the contract, AI makes you efficient. With it, AI makes you more capable.
The difference is not in the tool. It is in the decision you make about what the tool is for — and what it is not for.
That decision is the most agentic thing you can do in a world where AI is everywhere. Making it explicitly, holding it consciously, and revisiting it honestly is the personal version of everything this series has been arguing.
Essay 12: The social contract — moving from the personal to the structural. What humans as a category have a right to expect from AI systems as a category, and why the silence on this question is no longer acceptable.