An Old-Timer Psychologist's Case for Enthusiastically Embracing AI

How I Stopped Worrying and Learned to Love AI

In my humble opinion, the mental health professions are not handling this well.

The arrival of AI has produced more panic than curiosity, more defensiveness than openness. I want to make the opposite case: that this is an opportunity to look hard at what we actually do that can’t be done by a robot.

To do that, I start with a question the profession has never fully answered. Bear with me — it begins with orthopedics.

If a bone breaks, the medical problem is mostly mechanical. Healing depends on bringing the pieces back into position and keeping them stable long enough for the body to fuse them. There are variations in how this is done, but they all serve the same function. The mechanism of change and the intervention that supports it line up in a clear, self-evident way. There aren’t multiple, competing schools of thought about how to set a fracture.

Psychological work is different.

There are many well-established methods and models of psychological change. They don’t just differ in emphasis; they make different claims about what actually produces change.

Sometimes, when a model is followed closely, change occurs in the way the model would predict.

But often it doesn’t. The work proceeds exactly as it’s supposed to, and nothing changes.

And then there are moments that don’t fit the clinical model at all. The conversation drifts, the structure loosens, the clinician departs from what they were trained to do—and something important happens anyway. Some of the most impactful moments show up when the model isn’t really operating at all.

Taken together, this puts us in an odd position. Unlike in medicine, this isn’t the exception: following a model doesn’t reliably lead to change, and change regularly occurs without following one. This is widely recognized among practitioners. Psychotherapy outcomes research confirms it.

In 1997, Bruce Wampold and his colleagues published a landmark meta-analysis of psychotherapy outcome studies. The finding was counterintuitive: it didn’t matter which model guided psychotherapeutic treatment. Cognitive behavioral therapy, psychodynamic therapy, and many other approaches produced broadly similar results across hundreds of studies and decades of research. Whatever was producing the change, it wasn’t the model. Something else was doing the work.

Critics responded immediately, and the finding has been debated ever since. But in subsequent studies, the core result has held up again and again.

The profession didn’t ignore this. We cited it, taught it—and then went on to build more and more models.

The most common response was to make a gesture toward the relationship — the therapeutic alliance. As if that settled the matter. But what about the relationship?

There still isn’t a clear, widely accepted account of what actually produces change. We know what it isn’t; we just don’t know what it is. Something is missing.

Clients present themselves in therapy as suffering from some form of distress—diagnosed or not, framed medically or developmentally, located wherever the model you’re working from would place it. Underneath all of that, at the most basic level, they’re all in some sort of state that has become difficult enough that they can no longer pretend everything is fine.

Sometimes that state is acute, and they’re in crisis. More often, it’s persistent—low-grade, familiar, and exhausting. The particular form it takes matters less than what happens next. What matters is who they encounter when they walk into the room.

Some providers offer something the client can feel almost immediately, even if they can’t quite say what it is. Others are competent and credentialed and make a sincere effort: their warmth is palpable and correctly calibrated; they’re saying the right things and going through the right motions. And yet the client somehow knows that the person across from them isn’t quite there. The client leaves feeling handled rather than met, and — rarely, but not never — leaves the session feeling worse than when they arrived.

What accounts for this variation? The professional apparatus doesn’t explain it. The “therapeutic alliance” likely points in the right direction. But what actually happens between people when one of them is in difficulty — and not just in the clinical setting, but in everyday interactions?

What does a person in a state of distress actually need from a relationship with another person?

The clearest answer I know comes not from the consulting room, but from the nursery.

Anyone who has spent more than a few minutes in the presence of a crying infant knows what it does to a room. It’s not like other sounds. The urgency is total—not the kind you can ignore, defer, or fit around whatever you were doing. The infant isn’t expressing distress — it is the distress, its whole being is now organized around nothing else. This presents itself as a non-negotiable summons. Something in you — something you didn’t choose and can’t override — was built to receive it that way. Evolution built it to be impossible to ignore. This is distress at its most basic.

What the infant is expressing isn’t quite an emotion in the adult sense — it isn’t sadness or fear or anger, though it may contain elements of all of these. It’s something far more fundamental: simply a state of being that has exceeded the organism’s capacity to contain itself. The infant is overwhelmed, in the most literal sense of that word. Something — hunger, pain, cold, overstimulation, the sudden absence of a familiar presence — has tipped the system past the point it can manage. The crying is a signal, biological and urgent, that the organism needs something from outside itself in order to come back to itself.

And it doesn’t need an interpretation of its distress, or a name for it, or a treatment plan to address its underlying causes. What it needs is a particular quality of adult presence — someone whose own nervous system can remain steady enough, in the face of the infant’s distress. When that steady presence can hold its ground, something can shift that the infant could not easily have shifted by itself. The crying recedes. The crisis passes.

What just happened? The trigger explains the timing, not the form. What actually needs explaining is the way the distress takes over the whole organism, and the way it changes so dramatically in the presence of another.

Here’s one way to understand it. For the duration of gestation, there was, in essence, no boundary between self and world, no requirement to be a discrete organism, no gap between need and its fulfillment — and then, abruptly, all of that is over. The infant now exists in a condition of separateness.

Of course nobody remembers this. What remains isn’t memory, but something else — a trace of that condition that shapes experience without ever surfacing as recollection, part of everyone’s history. If that’s right, then the separation that began at birth becomes part of the background of experience.

That condition doesn’t simply disappear in adulthood. It gets socialized, defended against, managed through layers of cognition and self-narrative — but something of it remains. What was raw and total in the infant becomes, in the adult, something more subtle: a chronic mild vigilance, a longing not exactly for other people but for what other people can sometimes restore — a felt sense of wholeness, of all-right-ness, that the soothed infant knows in its caregiver’s arms. More of what drives human life traces back to that longing than we usually acknowledge.

That underlying condition can remain hidden for years, even decades, managed well enough by the strategies built up around it. Until something happens. A loss, a rupture, an accumulated weight that finally exceeds the system’s capacity to contain it. And suddenly the original experience of separation is much closer to the surface.

So the adult who walks into a consulting room still, underneath everything, carries that infant. The particular history varies enormously — what happened, what didn’t, what was lost, what was never available. But something in the underlying condition is still present. Nothing seems to fully resolve that separateness, but what close relationships provide — in infancy, in friendship, in love, in the consulting room — is an approximation of what was lost: another nervous system, offered steadily, that makes the condition of being separate feel, at least for a moment, less complete.

Tronick, Stern, Trevarthen and others have been documenting this for decades — the mutual regulation that can occur in two-person interaction. Earlier psychoanalytic thinkers, particularly those associated with the British School of Object Relations, were describing related phenomena decades ago in different language.

The client brings history, a constructed self, a well-developed set of strategies for managing whatever they’re carrying. The provider brings training, a theoretical framework, a repertoire of techniques. But underneath all of it, something simpler is occurring. One nervous system is providing something the other can use.

Some people arrive in obvious distress. Many don’t: They’re composed, matter-of-fact, agenda in hand. And yet underneath that organization—sometimes just barely underneath, sometimes buried much deeper—is still, always, that infant.

For as long as I’ve been doing this work, when someone walks into my consulting room, I find myself trying to sense the infant beneath the adult presentation. Not to treat them as an infant, or to project distress onto them, but to remain in contact with the level at which I believe change actually occurs. Their adult story is real and important, and it guides the direction of the work. The infant underneath it is where the work begins.

What they encounter, when the encounter actually makes a difference, is a presence that can stay with whatever is in the room without being destabilized by it. Not a presence that remains unmoved — that would be distance, not contact — but one that’s moved without being swept away. Not just emotionally attuned, but structurally steady under load: staying close to the full weight of what’s there, without flinching or subtly communicating that the weight needs to be lighter, different, or more manageable than it is.

That subtle signal — that the distress is too much, that it needs to be shaped into something easier to work with — is a common feature of failed encounters. It can take the form of a too-quick reassurance, a reframe that arrives before the feeling has been fully met, an interpretation that is accurate and yet beside the point, or a warmth that is genuine but organized around the provider’s need for things to go a certain way.

The client often senses this without being able to name it. So they adjust. They present a version of themselves that’s easier for the provider. The work proceeds on that version, and the thing that actually needed contact doesn’t get it.

By contrast, a provider’s steadiness — their capacity to remain organized in contact with what’s really happening for the client — can become something the client’s system is able to match. Offered in relationship, it can be borrowed until the person’s own system finds its footing.

We can call this organizing contact. It doesn’t interpret, fix, reframe, or validate. It organizes experience: not conceptually, but physiologically, in real time. A nervous system that has lost its footing finds, in interaction with one that hasn’t, something it can use to come back to itself. It isn’t the whole of the work, but it may be what makes the rest of it possible — the conditions under which interpretation, technique, and insight can do their work. Methods matter, but their effectiveness may depend on this.

Organizing contact provides an approximation of being held and contained by something larger than yourself, not having to manage it alone. Not unity itself, but something close enough, steady enough, that separateness becomes more tolerable for the duration of the encounter. The person leaves not just steadied, but having briefly inhabited a different relationship to their own experience.

One way to understand why this matters as much as it does is to see it as reaching beyond the particulars of any single history. In many forms of distress, something in the experience doesn’t feel entirely specific to the situation at hand — something more general in the way the system has lost its footing. The pull toward connection in those moments isn’t only about actual losses or relational failures or unmet needs, though it includes all of these. It can also be understood as a response to a more basic condition: that being a separate organism is, at times, difficult to sustain.

Seen in that light, what the person is reaching toward isn’t just relief from a particular problem, but a shift in the way experience is being held — a temporary easing of that strain. Organizing contact can be understood as operating at that level: not resolving whatever underlies the distress, but making it more manageable for a time.

Even without taking that view, what happens in the room remains the same. The client can be understood, on some level, as a nervous system in search of another nervous system that can help it come back to itself. When it finds one, something shifts. When it doesn’t, time passes and nothing essential changes.

What determines where an encounter lands? The provider’s baseline capacity, and their state in this particular session.

Below a certain threshold of organizing contact, the provider may either add to the dysregulation or offer nothing the other person can use. Above it, something useful becomes available, and the question becomes how much, how consistently. Most encounters fall somewhere in that range, and the position can shift within a single session.

At and just above the threshold of organizing contact, the task is restraint. Just don’t make things worse. Be present and steady, nonreactive. Don’t withdraw when the weight gets heavy, don’t escalate when things get more intense, don’t import your own instability — your uncertainty, inadequacy, fear, confusion, distaste — into what’s already in the room.

Simple, but not easy. And not nothing.

Above the threshold, the provider isn’t just steady — they’re actively working with what’s happening in the moment. They slow things down when the client is getting overwhelmed, keep attention on a feeling that the client would move past too quickly, gently interrupt when the conversation drifts away from something important. When things start to fragment, they help the client to cohere. Their attention isn’t just steady: it’s directed toward what can be engaged, while leaving alone what would make things worse.

For someone who has rarely encountered this — a presence that remains steady, doesn’t flinch, doesn’t need it to be different — something that was braced begins, tentatively, to unbrace.

A provider who reliably holds here — week after week, not destabilizing, not withdrawing — gives the client something valuable to discover: that proximity to another person can be safe. That discovery can extend far beyond the bounds of this particular relationship.

Picture a client trying to articulate something they’ve never said out loud before. A loss, a failure, something they’ve done or had done to them that carries immense shame. They’re watching, as they speak, for a signal they’ve learned to expect: an almost imperceptible withdrawal, a reframe that dilutes the experience. But the signal doesn’t come. The provider stays close to the raw experience. Doesn’t rush to fix or reframe. The supercharged material is held gently in the space between them.

That reception is active, even when it looks like stillness. Something that had been held in a diffuse, unformed way takes shape in the act of being said out loud, in the presence of someone who can hear it without flinching. What shifts, and how much, varies.

Sometimes something releases mid-session: the body settles, the person becomes more present with themselves in a way neither of them fully orchestrated. More often it’s subtler: they leave the room and notice, later, that something has moved. That they can think about something they couldn’t think about before. They may attribute this to the interpretation that was offered, or the question that was asked. Those things were real.

But these explanations can be understood as secondary to something else: the conditions that made them possible in the first place — the steadiness of the other person’s presence, temporarily available to borrow.

Something has changed. The profession is under pressure in a way it wasn’t a decade ago.

AI-assisted mental health tools are proliferating. There are apps that deliver CBT protocols, track mood, provide coping strategies, and help manage recovery from addiction on demand. Some of them are very well designed and highly sophisticated. Some of them are genuinely useful, for some people, for some things. And they raise a question the profession is finding surprisingly difficult to answer: what exactly do you, the practitioner, offer that this doesn’t?

A profession with a clear account of its own mechanism of change would be better able to answer this question. It would know, more precisely, what it provides that a language model can’t, and it would be able to say so. But the profession doesn’t have a confident answer. And it doesn’t have a confident answer because it never really resolved the question Wampold raised thirty years ago.

AI can do a great deal. It can provide information, reflection, validation, psycho-education, structured protocols, and something that resembles empathy closely enough that many people find it comforting. It can be available at 3:00 am. It doesn’t get tired, doesn’t have bad days, doesn’t bring its own unresolved material into the room. For certain purposes, in certain circumstances, it may be more than adequate.

What it cannot do is be a nervous system, which matters if the mechanism of change actually depends on one.

That’s not a metaphor. It’s a precise claim. Organizing contact is one way to describe relationships in which co-regulation can occur reliably. Co-regulation, by definition, requires an actual biological nervous system, with its own regulatory history, its own capacity to be moved without being swept away, and its own embodied steadiness available to be borrowed. Other things may help, but they’re a different mechanism.

With a language model, what you get is words, however sophisticated that process may be. But it doesn’t have a nervous system that can be dysregulated or regulated. It can’t be genuinely present in the way described here. It can’t be moved by what’s in the room in the way a nervous system can. It can’t stay close to the full weight of it. It can’t make steadiness available as a felt experience to another nervous system. It can describe steadiness. It can perform steadiness. It cannot offer steadiness as a biological reality.

This is what the profession has that AI doesn’t. Not so much the models or the techniques. Those things are, in principle, more easily replicable by a machine. What isn’t replicable is the biological encounter between two nervous systems — one of them regulated and present, offering something the other can use to return to itself.

That encounter may be at the core of what the profession actually has to offer.

The AI disruption isn’t a threat to what human providers actually provide. It’s a clarifying pressure on the profession’s understanding of itself. What’s being disrupted is the technical scaffolding: the models, the techniques, the theoretical apparatus that have come to define the field. What’s at risk is the way the work is modeled, not the work itself or the people who do it.

For the provider who has been offering organizing contact all along, the disruption changes very little about the work itself. But it does change what needs to be said about the work.

For the provider whose work has primarily relied on the scaffolding, the disruption is more serious. AI may take over the parts of the work that involve applying models. What it doesn’t replace is the provider’s capacity to co-regulate. The client may have been co-regulated by their provider’s nervous system all along, even when both of them thought that all they were doing was cognitive behavioral therapy.

The profession has been handed an opportunity it didn’t ask for. Whether it takes that opportunity won’t be decided by the profession as a whole, but by one provider at a time, in one consulting room at a time, with whoever walks in next.

How we ended up here isn’t a mystery. Training programs are necessarily organized around transmittable content. You can teach a theoretical framework. You can teach a set of techniques. You can assign readings, run role-plays, evaluate competency in applying a model.

What you can’t easily teach — can’t break into modules or certify in a curriculum — is the capacity to remain present to another person’s distress without being destabilized by it.

That capacity can be developed — supervision at its best does help develop it — but it resists measurement. And what can’t be measured tends not to count.

Credentialing follows the same logic. A license certifies that the holder has completed approved coursework, accumulated supervised hours, and passed both written and oral examinations. I hold these credentials; I have for 40 years. My license says nothing about whether I can offer organizing contact. It can’t say this, because organizing contact isn’t what examinations can measure. The credential is a proxy: a reasonable attempt to approximate legitimacy through something objectively verifiable.

Reimbursement takes the same shape and goes even further. Insurance systems require a diagnosis, a treatment plan, and documentation of progress toward specified goals. They pay for identified disorders treated by recognized methods. The encounter between two nervous systems — one regulated and present, offering something the other can use — isn’t billable.

Diagnosis. Treatment planning. Documented progress toward treatment goals. The primary mechanism of change doesn’t appear anywhere in that list — and there’s no slot in the reimbursement system where it could. Not because it isn’t real, but because it’s not so easily specified.

The result is a profession organized around a description of its work that’s accurate, as far as it goes, but stops just short of what matters most. The models are real, the techniques have effects, the diagnostic categories capture something. But the mechanism that operates underneath all of it remains unnamed, and therefore unexamined.

Take EMDR as an example: it’s one of the most procedure-driven methods in the field, with well-defined outcomes and a substantial evidence base. Even there, though, the mechanism of change isn’t clear. Multiple explanations exist in the published literature, but none is authoritative. Whatever is producing the effect, some of it is surely coming from what’s happening in the interpersonal field — from the encounter between two nervous systems, not just from the procedure itself.

Without a clear account of its mechanism, the profession has no reliable way to select for the people who are good at it, train for it, or recognize when it’s absent.

Some people are simply better suited to this kind of work than others. Not better trained, necessarily — though training matters, and the capacity can be strengthened and honed. But it probably can’t be installed where it isn’t present. If that’s true, we may have to accept that some limits can’t be trained around.

For those who enter the profession with a natural capacity for organizing contact, it may develop further on its own. For others, it develops — or doesn’t — largely through chance circumstances: the quality of their supervision, the complexity of their caseload, the availability of a mentor who can model what the training program can’t teach.

The people who enter the profession and cannot offer organizing contact may never be identified as such. They move through training, accumulate credentials, build practices. The apparatus certifies them. But the clients who encounter them may notice something is missing. They just won’t know what, or why.

This is what the structural forces produced: a profession whose description of itself stops short, because the structure made any other description nearly impossible to sustain.

The encounters that organize share recognizable features: someone with sufficient capacity is present, the person in distress is allowed to be in the state they’re actually in, and nothing in the interaction imports additional dysregulation into what’s already there. These features don’t require a particular theoretical framework. They don’t require a particular credential. They do require a capacity, and the capacity, as we’ve seen, is distributed differently than the credential system assumes.

Any successor system would somehow need provider assessments closer to the actual work. It would need to be able to say, at a minimum: this person has the capacity for organizing contact, or they don’t. What that would look like in practice — how you’d identify and measure it without reducing it to a credential — isn’t clear. And even if you could solve that, institutional pressures would reshape it into something that can be taught, tested, and billed. In the end, it might not be the same thing anymore.

If that’s right, then the issue isn’t just how the work is defined or measured, but who actually needs to do it. Organizing contact doesn’t require a graduate degree. It requires the capacity, which isn’t evenly distributed. This is not only a reframing of what the work is, but also a potential disruption of who does it. A credentialed provider might be exactly the right person to assess, coordinate, and manage the complexity of a case, and at the same time exactly the wrong person to provide the encounter itself. The current system, which bundles both into a single credentialed role, has no way to see that distinction and little incentive to look for it.

The capacity for organizing contact isn’t institutional. It exists in people who never entered the profession, never sought a credential, never heard of Wampold. It’s always been more widely distributed than the system that was supposed to deploy it. Those people haven’t been found because no one was looking for them, maybe in part because the profession didn’t know what it was looking for.

That’s where the opening is. It’s our move.

John A. Martin is the author of Staying Close (2025) and Staying with Strain (2026), Enotis Press, and has been a psychologist in private practice in San Francisco for nearly 50 years.