Change Leadership Series. A companion to The Readiness Illusion from Ideas and Innovations.

In The Readiness Illusion, I named a pattern that I think deserves more scrutiny than the market is giving it. That pattern rests on the assumption that capability and readiness are the same thing. The gap between what AI agents can do in a controlled demonstration and what they can responsibly do inside real organizations with real consequences. The chronic over promise dynamic that markets have run many times before with new technology, and that we are running again now with AI. That piece was diagnostic. It traced causes, situated the AI version of the pattern inside a longer human pattern, and refused the narrative that says the agent era has already arrived. I closed it with a question I deliberately left unanswered, and I want to address that question here.

Beyond the Binary

The diagnostic question, “are we ready,” isn’t the most useful question to organize around, because it presents the situation as a binary the data doesn’t support. Either the technology is ready or it isn’t. Either we adopt aggressively or we procrastinate. That framing suits headlines across media channels and nearly all vendor marketing. It doesn’t suit the leaders who actually have to make decisions inside organizations that contain real people and real consequences.

A more useful question is what posture we should adopt toward the technology while both sides of the relationship continue to mature and evolve. We are in many ways still trying to find a balance, a settled sense of where the technology fits and where we do. I am calling this the partnership posture. The posture replaces “is it ready” with a different set of questions: what does each side bring (AI and humanity), what does each side owe the other, and how do we structure the working relationship so that it produces value while respecting what each side actually is.

That sounds perhaps too simple or too easy. I think it is more significant than it appears, because it changes how we navigate the weeks, months and years ahead. The focus is no longer the model in isolation, and it is no longer the user in isolation. The focus is the working relationship between them. And what you can identify, fund, develop, and measure depends entirely on what you are willing to look at.

I am deliberately calling this a posture rather than a framework or a model, because postures are more durable than frameworks. A posture is the stance from which you approach the work, the assumptions you bring, the questions you ask first. The tooling will keep changing as it rapidly has been. The vendor language will keep overstating. The posture is what the individual brings with them into every new tool and every new claim, and it travels in a way that frameworks pinned to specific products don’t.

What Each Side Actually Brings

If we are going to talk about partnership, we have to be honest about what each partner contributes. Not as a feature comparison, but as a recognition of capacities the relationship depends on.

We bring contextual judgment that is grounded in lived consequence. We don’t only know what is true in a domain. We know what matters in it, why it matters, who it affects, and what happens when it is wrong. That judgment is built from embodied experience, from the residue of previous mistakes, and from a continuous awareness of stakes that shapes what we attend to. We bring accountability that has a name attached to it, a person who can be held to the consequence of a decision. We bring the ability to ask whether the question being asked is the right question. We bring ethical reasoning that comes from commitment to specific people whose welfare we care about, not from rules alone. And we bring institutional memory that lives nowhere in any document, the unwritten rules, the political dynamics and the trust networks that determine what is actually possible (or not) inside an organization.

AI systems bring different capacities, and we should consider them with accuracy and reality. They bring pattern recognition at a scale no individual human could match. They bring the ability to draw useful framings from adjacent domains quickly. They bring a near zero cost of iteration, the ability to produce a tenth draft without fatigue. They bring linguistic fluency across audiences and tones. They bring a kind of patient availability that does not get tired or bored. And they bring the capacity to surface options the human had not considered and to perform routine analytical and generative work at speeds that genuinely change what is possible for an individual or small team.

None of those capacities should be easily dismissed. The mistake is not in noticing them. The mistake is in treating them as though they are equivalent to the human capacities, or as though their continued improvement will eventually subsume them. The difference is structural. The partnership posture is built around that structural difference, not around the hope that it will disappear.

A partnership posture holds both sets of capacities in view at the same time, reflecting both current and forthcoming reality. We should dismiss the thought that says one side dominates and the other will be absorbed. We should set expectations about what is reasonable to ask of each partner and what isn’t. The partnership isn’t the destination at which the two sides become equal. The partnership is the relationship in which the two sides cooperate on terms that respect what each one is and is not.

I am not suggesting that AI is conscious and an equal part in the relationship. However, there is an interesting default of humanity where we tend to apply personal attributes to pets and material items that play roles in our lives. We “personalize,” we give names, and we apply a perception of personality. Those using AI as a “friend” or “personal counselor” or “sounding board” have likely applied attributes that lead them to treat that AI as an equal part. It is not.

The Conditions That Make Partnership Work

For the partnership posture to function inside organizations, certain conditions have to be in place. I want to name three that I consider essential, and they extend the trust components I described in The Readiness Illusion into something more constructive.

The first is shared vocabulary. Partnerships break down when the two sides do not mean the same thing by the same words. Breakdowns occur across cultures, teams, geographic boundaries and even across dialects. So it’s natural that vocabulary plays a role in what we are considering here.

In human to AI collaboration, we close that gap through prompting, through evaluating outputs, through naming what we mean explicitly, and through teaching the system in each interaction what it is being asked to do. Most users don’t yet treat this as a discipline. They treat it as a chore, or as evidence that the technology is inadequate. The partnership posture treats it as the work itself. The construction of the brief is not preparation for the work. It is part of the work, and the depth of the brief is the largest single variable that determines the depth of the output.

The second is clarity of decision rights. In any partnership, the question of who decides what has to be answered before the work begins, not after a disagreement surfaces. With AI, this means deciding which decisions the system can make, which it can recommend, and which remain with the human. The decisions that remain with the human are not the residual category (deference to the output without question). They are the consequential category (the application of human critical thinking and assessment skills). And the choice of where to place that boundary is itself a leadership act. Most organizations have allowed the boundary to be set implicitly by the limits of the tooling, which means it moves every time the tooling changes, which means it isn’t a stable basis for trust.

The third is error recovery. Partnerships are tested by failure, not by success. With AI, this means designing the workflow around the assumption of error rather than around the hope of accuracy. It means making verification natural, rewarded, and difficult to skip. It means building the audit trail before it is needed. Most organizations are deploying agents with recovery plans that exist only on paper and have never been tested under real conditions. That is not partnership. That is exposure dressed up as confidence.

The Practitioner’s Disciplines

Beyond the organizational conditions, the partnership posture requires specific disciplines that individuals and organizations have to develop deliberately. Here are three that I have come to consider essential in my own work.

The first is briefing as professional craft. The prompt is not a query. It is a brief, in the sense that a senior practitioner briefs a junior colleague before delegating a piece of work. It includes the goal, the audience, the constraints, the success criteria, the prior context, and the specific risks the briefer wants the recipient to attend to. Most users treat the prompt as a search query and are then surprised when the output reflects search query depth. Picture the difference between typing “write a client update” and writing a short brief that names the client’s situation, the three points that must land, the tone to avoid, and the pressure behind the deadline. Same tool, very different result, because the second person did the work of briefing rather than searching. The partnership posture treats the prompt as the professional act it actually is and invests in it accordingly. Might it be time consuming? Yes, often in perception and sometimes in reality, but the exercise is important to the expected outcomes.

The second is treating correction as signal rather than frustration. When the output is wrong, the wrong is information about how the system understood the task and where the gap was. Refining the brief in response, and noticing the patterns of error across many interactions, is how one builds partnership skill over time. Treating the wrong output as evidence that the technology does not work is a way of skipping the learning that the partnership demands, and it produces a flat trajectory of skill rather than a rising one.

The third is treating every output as a draft. No output should leave one’s hands as the system produced it. This is not a concession to the limits of the technology. It is the same discipline a senior practitioner applies to any draft from any source. The act of editing is itself a thinking act. It is where the practitioner’s judgment, taste, and accountability enter the work. Practitioners who skip this step are not gaining productivity from the partnership. They are exposing themselves and their organizations to risk dressed up as efficiency.

The Leadership Work

The leadership work that follows from the partnership posture is significant, and most of it is underdeveloped in current organizational practice.

Leaders have to redesign roles around the partnership rather than around substitution. Substitution thinking asks what an agent can replace. Partnership thinking asks how the human role changes when an agent participates, what new capabilities the human now needs, and how the development path for that human is structured given that some of the routine work that used to build their judgment is now performed by the system. That is a different design question, and it produces a different role architecture. The organizations that do this work will have professional capacity in five years that the others won’t. I have explored this identity and capacity role shifting in my presentations, podcast episodes and newsletter. It is a critical exercise that most individuals must face and consider as AI re-orients the definition of the professional.

Leaders have to invest in the practitioner disciplines I named above. Most current AI training is tooling training. It teaches users how to operate the interface. Companies are holding end-of-day prompt reviews, output summaries and the like. Those are helpful activities, but are likely teaching the wrong practices.

The partnership disciplines are not tooling skills. They are professional skills that require practice, feedback, and development over time. Organizations that treat practitioner development as a budget line will outperform organizations that treat it as a self-serve responsibility, and the difference will be measurable inside three years.

Leaders have to set expectations honestly. The partnership posture requires acknowledging the limits of the technology to the people being asked to use it, rather than performing certainty for the sake of momentum. People can absorb the truth that the technology is genuinely useful and not yet autonomous. They can’t absorb the contradiction between marketing certainty and their lived experience without losing trust in their leaders. The partnership posture is, among other things, an honesty posture.

And leaders have to design measurement around the partnership rather than around the agent in isolation. I named this in The Readiness Illusion: measuring agent activity without measuring correction load, recovery cost, and the trajectory of human capability over time is not measurement. It is reporting. The two are not the same.

Where Partnership Goes Wrong

The risks of getting this wrong are worth calling out, because they tend to be invisible until the moment they aren’t, the way so many shifts across history revealed their unintended consequences only in hindsight.

The first risk is atrophy. If the partnership is structured so that the human side gradually stops exercising the judgment, the contextual awareness, and the ethical reasoning that the partnership requires from them, the partnership stops working even if the agent keeps producing output. The output looks fine until the moment the agent encounters a case it cannot handle, and at that point the human capability that should have caught the problem has eroded. This is not a new observation, but it is a real risk with severe consequences. Lisanne Bainbridge named it in her 1983 paper on the ironies of automation. When automation takes over the routine work, the operator stops practicing the skills that the rare and critical intervention later demands and is least prepared at the moment preparation matters most. We remove ourselves from the driver’s seat and abdicate responsibility. The atrophy is slow, expensive, and largely invisible until the failure that reveals it.

The second risk is automation bias, which I described in The Readiness Illusion. If the practitioner doesn’t maintain the verification discipline, the partnership tilts into deference, and the deference isn’t earned. The system’s apparent competence, combined with the practitioner’s time pressure, produces an under verification pattern that Mosier and Skitka documented in high technology cockpits in the 1990s, the tendency to accept an automated recommendation without checking it against the information that should have contradicted it. Remember, we are evolutionarily programmed to conserve energy and default to taking the back seat too easily, often doing so unconsciously.

The third risk is false equivalence. A partnership in which the two sides are treated as equivalent in capability obscures the differences that the partnership posture exists to respect. The human side has to remember what it brings that the agent doesn’t. The agent side, through its language and through the way its outputs are presented, can sometimes invite the human to forget. Maintaining the appropriate asymmetry is not pessimism. It is precision.

The Partnership Question

If the readiness question was “is it ready,” the partnership question is something different. It’s whether we are structuring the working relationship in a way that produces value on both ends, respects what each side actually is, and develops both sides over time. That question does not resolve into yes or no. It is a question to live inside, to revisit regularly, and to use as the standard against which to evaluate the choices an organization (and really society) is making about how to deploy these tools.

I think that question is the right one for the next phase of where we are heading. It is a harder question than the readiness question, because it doesn’t have a clean answer. It is also more useful, because the work it points toward is the work that will determine whether the agent era produces value for the people inside organizations and the people those organizations serve, or whether it produces another cycle of overpromised capability followed by quiet disappointment. We have historically run that cycle many times. We know what it costs. The partnership posture is one way to refuse to run it again.

Sources and References

Lisanne Bainbridge, “Ironies of Automation,” Automatica, Volume 19, Issue 6, 1983.

Kathleen L. Mosier and Linda J. Skitka, research on automation bias in high technology cockpits originating in the 1990s, including Skitka, Mosier, and Burdick, “Does Automation Bias Decision Making?” International Journal of Human-Computer Studies, 1999.

Kevin Novak, “The Readiness Illusion: Why the AI Agent Era’s Loudest Claims Outrun the Evidence,” Ideas and Innovations, 2026.