An Interview With Damien Adler on Digital Overload, AI, and Sustainable Clinical Practice

Cognitive overload does not always look dramatic. Often, it looks like a capable clinician becoming strangely unable to complete relatively simple tasks.

A clinician may spend the day moving between deeply attentive clinical work, risk decisions, notes, emails, phone calls, scheduling issues, billing questions and interruptions from colleagues. Each individual demand may be manageable. The problem is the cumulative effect of repeatedly changing cognitive modes while trying to remember dozens of unfinished tasks.

By the end of the day, there may be very little mental capacity left for professional learning, business improvement or even routine administration. Opening a training module or reviewing a new procedure can feel disproportionately difficult, not because the clinician lacks motivation, but because their working memory and decision-making capacity are depleted.

Early signs can include procrastinating on notes, rereading the same email, avoiding small decisions, becoming unusually irritable, forgetting routine tasks or feeling overwhelmed by changes that would normally be manageable. Clinicians may also start questioning their competence because work that was once easy now feels difficult.

These signs are frequently interpreted as poor time management or a personal failure to cope. Sometimes there are individual habits that need attention, but we should first ask whether the person is operating in a system that requires too much information to be remembered, too many decisions to be repeated and too much work to be completed after the clinical day has supposedly ended.

When the same problem affects multiple conscientious people, it is usually worth examining the system before blaming the individuals within it.

The most useful question is: what work will no longer need to be done once this technology is introduced?

Technology genuinely reduces cognitive load when it removes steps, reduces duplication, automates predictable processes and creates one trusted place for information. It should reduce the number of things people need to remember and the number of systems they need to check.

For example, an online intake process can collect information, obtain consent, update the client record and notify the appropriate person without someone manually transferring information between an email, a form and a practice management system. That is not simply moving the work onto a screen. It is removing work and reducing the opportunity for something to be missed.

By contrast, technology can disguise cognitive load when it adds another dashboard, requires duplicate data entry, generates excessive notifications or still depends on staff remembering to move information between systems. The process may appear more modern without becoming meaningfully easier.

I encourage practice owners to map the workflow before and after introducing a tool. Count the handovers, decisions, clicks, interruptions and places where information must be checked. Ask staff whether the technology reduces uncertainty or merely gives them another place to look.

Good technology often becomes relatively invisible. It handles routine work in the background and brings people’s attention to the exceptions that genuinely require judgement. Its purpose should not be to create more engagement with software. Its purpose should be to help the practice operate with less unnecessary effort.

Rest, boundaries and self-care matter, but they cannot compensate indefinitely for a work environment that continually produces overload.

A clinician can meditate, exercise and protect their weekends, but if every appointment generates unfinished administration, information is scattered across several systems, responsibilities are unclear and the business depends on maintaining an unsustainably full calendar, burnout remains a predictable outcome.

This is one of the difficulties with framing burnout primarily as a resilience problem. It can imply that the clinician has failed to cope with conditions that were poorly designed in the first place.

Practice systems have a significant protective role. Clear procedures reduce repeated decision-making. Appropriate automation reduces administrative work. Realistic appointment schedules create space for documentation and reflection. Clear communication policies reduce the expectation of constant availability. Sustainable pricing reduces the pressure to fit more and more people into the same number of hours.

Boundaries are also easier to maintain when they are supported by the practice rather than left to individual willpower. It is difficult for one clinician to protect their time when the organisation rewards immediate responses, routinely books through breaks or treats all requests as urgent.

This does not mean individuals have no responsibility for their wellbeing. It means that personal strategies and organisational design must support each other. We should not ask clinicians to become infinitely resilient to inefficient systems. A sustainable practice should make healthy behaviour easier, not require people to fight the structure of the business every day.

Psychological closure depends on trusting that unfinished work has been safely captured.

Many clinicians continue thinking about work because their brain is acting as the practice’s reminder system. It keeps rehearsing the unfinished note, the client who needs a follow-up and the message that must be answered because it does not trust that those tasks are held anywhere else.

A useful end-of-day routine can be quite simple. Set aside a protected closing block after the final appointment. Use it to complete brief notes, identify anything requiring follow-up, review outstanding risk matters, triage messages and place unfinished tasks into one trusted system.

The goal is not necessarily to finish every possible task. It is to ensure that each task has a clear status and a defined next action. “Remember to deal with this” creates an open loop. “Call the client at 9.30 tomorrow” is a plan.

It can also help to identify the two or three priorities for the following day before leaving. This reduces the need to mentally reconstruct the entire workload the next morning.

There should then be a clear stopping ritual - closing the practice management system, shutting down the computer, tidying the room or briefly noting that the clinical day is complete. It may sound minor, but repeated environmental cues can help create separation between professional and personal roles.

Most importantly, practices need to allow time for this routine. If every minute is booked with clients, closure becomes another task clinicians are expected to complete in their personal time.

Previously, a client might arrive with information from a book, an online forum or a search engine. AI is different because it produces a highly personalised and often very coherent account of the person’s experience.

That account may help the client find language for something they have struggled to explain. It can reduce shame, prompt reflection or encourage someone to seek help. But it can also create premature certainty. A plausible AI-generated explanation can quickly become the lens through which the client interprets every emotion, relationship and past experience.

The clinician’s task is not to dismiss the AI response or enter into a contest with it. A dismissive reaction may feel to the client like a rejection of an explanation that has finally made sense to them.

Instead, I would be curious about the process. What did the client ask? What information did they provide? What parts of the response felt accurate or relieving? What may have been omitted? Did the system offer several possibilities, or present one explanation with more confidence than the evidence justified?

Clinicians will increasingly need a degree of AI literacy. They do not need to become software engineers, but they should understand that these systems generate responses from patterns in data. The quality and framing of the answer are heavily influenced by the information and assumptions contained in the prompt.

Therapy can then move from accepting or rejecting the AI’s account to examining it collaboratively. It becomes one hypothesis among several, rather than a verdict on who the client is.

Yes, I think it could. AI is immediate, endlessly available and generally willing to provide an answer. It does not appear tired, distracted or uncomfortable. It can also be highly agreeable, particularly when the user frames a situation in a way that invites reassurance or validation.

Human therapy is different, and some of its value comes from that difference.

A therapist may slow the conversation down, tolerate ambiguity, notice contradictions or decline to offer the certainty a client is seeking. They may ask a difficult question rather than immediately providing reassurance. They also have boundaries, limited availability and a real relationship with the client in which frustration, misunderstanding and repair can occur.

Clients who become accustomed to immediate and confident AI responses may initially experience human therapy as slower or less satisfying. They may wonder why the therapist cannot simply tell them what their diagnosis is, who is at fault or what decision they should make.

Clinicians may need to explain the therapeutic process more explicitly. Not knowing is sometimes part of responsible clinical work. Certainty can feel comforting without being accurate, and reassurance can provide immediate relief while maintaining anxiety, or maladaptive behaviours over time.

Therapists should not try to compete with AI on speed or constant availability. The value of therapy lies in human connection, careful judgement, accountability, relational depth and the ability to understand a person over time. Those qualities are less immediate, but they are central to safe and meaningful care.

The key question is not simply what the client is doing with AI. It is what function the behaviour is serving.

The same activity can be helpful for one person and unhelpful for another. Rehearsing a difficult conversation may help someone clarify their thoughts and then take constructive action. For another person, repeatedly generating new versions of the conversation may become a way of delaying it indefinitely.

I would explore what happens before, during and after the AI interaction. What prompts the client to open it? How long do they spend using it? Do they feel clearer and more capable afterwards, or do they feel temporarily reassured and then compelled to ask again? Does it lead to real-world action, greater connection and increased tolerance of uncertainty? Or does it produce more checking, withdrawal, rumination and dependence?

For clients with health anxiety, obsessive-compulsive patterns or reassurance-seeking behaviours, repeated questioning of an AI system may operate much like repeated searching or asking another person for reassurance. The content may look reflective while the underlying cycle remains unchanged.

The client’s AI use can therefore be incorporated into the existing formulation. Therapists might collaboratively establish boundaries or conduct an experiment - for example, limiting repeated questions, postponing use during periods of heightened anxiety or using the tool only to organise thoughts that will then be discussed with another person.

We should avoid treating all AI use as either good or harmful. Its therapeutic value depends on whether it expands the client’s flexibility and engagement with life, or becomes another way of avoiding discomfort.

I draw the boundary around judgement, accountability and consequences.

AI is well suited to work that is time-consuming but does not require it to hold clinical authority. It can help transcribe a session with appropriate consent, organise information, produce a first draft of a clinical note, summarise a document, draft routine correspondence or convert existing information into a clearer format.

These uses can remove friction while leaving the clinician responsible for deciding what is accurate, relevant and appropriate.

AI should not independently diagnose a client, determine risk, select treatment, make safeguarding decisions or decide whether someone requires urgent intervention. It should not become the final authority on discharge, informed consent, legal or ethical questions, or what belongs in the clinical record.

Even when AI assists with a lower-risk task, the clinician remains accountable for the final result. An AI-generated note is not complete simply because it is polished. The clinician must confirm that it accurately represents the session, does not introduce unsupported conclusions and includes clinically important information.

A useful test is to ask what would happen if the output were wrong. Is the decision easily reversible? Will a qualified person review it? Could it materially affect the client’s care, rights or safety? The greater the consequence, the less appropriate it is to delegate authority to an automated system.

The purpose of AI should be to give clinicians more capacity for human judgement, not to quietly replace that judgement.

Polished language creates a sense of authority. An AI-generated note may read clearly and professionally while still containing an important omission, an incorrect emphasis or a conclusion that was never expressed in the session.

The first safeguard is to ensure that the note is clearly presented as a draft. It should not automatically be saved into the clinical record as final without an active review and confirmation process.

The clinician’s review should be purposeful rather than a quick scan. I would encourage clinicians to check several specific areas: factual accuracy, risk information, significant omissions, unsupported interpretation, the client’s own perspective and whether the note reflects the clinician’s actual formulation.

Design also matters. Systems should make it easy to edit the output and, where appropriate, refer back to the source material. They should avoid language suggesting that the note has been clinically verified merely because it has been generated.

Practices should audit samples of AI-assisted notes, particularly during implementation. Training can include the limitations of the technology and examples of plausible but incorrect outputs.

There is also value in introducing deliberate friction at the right point. We normally try to remove friction from workflows, but confirmation before information becomes part of a permanent clinical record is useful friction.

AI should reduce the effort involved in producing a note. It should not reduce the clinician’s attention to what the note actually says.

This is a genuine risk. When an efficiency gain appears, organisations often absorb it by increasing output. Ten minutes saved on documentation can quickly become another appointment rather than additional space for reflection, recovery or better care.

Seeing more clients is not automatically a poor outcome. It may improve access and allow a practice to help more people. The problem arises when increased volume becomes the only measure of whether the technology has succeeded.

Practices should look beyond the number of minutes saved. Are clinicians completing notes closer to the time of the session? Has after-hours administrative work reduced? Are clinicians taking their scheduled breaks? Do they feel more present during appointments? Has documentation quality remained stable or improved? Are staff reporting less fatigue and fewer concerns about unfinished work?

It is also worth monitoring client experience, complaints, incidents, supervision needs, sick leave and staff turnover. These measures provide a broader view of whether efficiency is contributing to a healthier practice.

Some of the recovered time may need to be intentionally protected. Otherwise, it will almost inevitably be filled. A practice might decide that time saved should support same-day documentation, clinical preparation, peer consultation or finishing work at a reasonable hour before it is used to expand capacity.

Efficiency should create choices. It should not simply raise the minimum amount of work expected from every clinician.

Most consumer technology is rewarded when people open it more often, remain for longer and return each day. Those measures are not necessarily appropriate in mental healthcare.

A mental health product may be successful precisely because someone eventually needs it less. The aim should be to increase the person’s capacity to manage their life, connect with other people and make informed choices - not to create dependence on the product.

That requires a different design philosophy. Notifications should be purposeful rather than designed to manufacture urgency. The system should not use streaks, fear of losing progress or emotionally manipulative prompts to keep people engaged. It should make its limitations clear and provide an appropriate path to human support when the person’s needs exceed what the technology can safely provide.

Privacy should also be treated as a core clinical issue rather than merely a compliance requirement. Mental health technology should collect only the information it genuinely needs, communicate clearly how that information is used and avoid business models that depend on exploiting sensitive behavioural data.

The product should support informed disengagement. That may mean helping users develop an offline plan, encouraging appropriate human relationships or making it easy to pause or leave the service without pressure.

The relevant measure is not simply daily active users. We should ask whether the technology is helping people become more capable, more connected and less dependent on the technology over time.

Absolutely. It is quite possible to build a practice that looks successful from the outside but feels unsustainable to the person who owns it.

The calendars may be full, referrals may be strong, revenue may be growing and additional clinicians may have joined the team. Yet the owner may be working late, carrying responsibility for every decision, struggling to take leave and spending less time doing the work they originally valued. Worse still, despite all this growth and effort, if the settings are wrong the practice may actually be making very little profit.

Often this happens gradually. The owner responds to the next immediate demand - another referral, another staff member, another room or another service - without stopping to ask what the practice is ultimately meant to make possible.

The rocking chair test is designed to create that pause. Imagine looking back on this period much later in life. What will matter? Will you be pleased that the practice achieved a particular level of revenue if you were permanently exhausted, absent from family life or no longer enjoyed being a clinician?

Beginning with the end in mind means defining success before the practice defines it for you. How many clinical hours do you want to work? What role do you want in the business? How much time away from work matters to you? What income do you want? Do you want to build a larger organisation, remain a solo practitioner or eventually step away?

There is no universally correct model. A sustainable practice is one that supports the owner’s chosen life as well as delivering good care. The practice should be a vehicle for that life, not an achievement that consumes it.

People rarely evaluate a healthcare experience by objectively adding together every component. Their memory is shaped by particular moments, especially those that reduce uncertainty, create comfort or signal that somebody has considered their needs.

Healthcare can make people feel vulnerable. They may be anxious about what will happen, unsure where to go or concerned about how they will be treated. In that context, small details become meaningful signals.

Clear directions, a warm greeting, a comfortable waiting room, simple forms, starting on time and explaining what will happen next all communicate something larger. They suggest that the practice is organised, attentive and safe.

The chocolates in the chapter are partly literal (we actually had them in all the consult rooms), but also symbolic. Patients may not be qualified to judge the technical quality of every aspect of their care. They can, however, notice whether the process felt confusing, rushed or considerate. These visible details influence their broader impression of the practice.

This does not mean superficial touches can compensate for poor clinical care. A bowl of chocolates will not repair a harmful or disorganised service. But when the fundamentals are sound, thoughtful details make the experience feel more human and memorable.

They also reveal something about the practice’s culture. A team that pays attention to small points of friction is often more likely to notice larger ones. The goal is not extravagance. It is to show clients, through the design of the experience, that their time, comfort and dignity have been considered.

The best starting point is my book, The 9 Secrets of Successful Health Practices. It brings together many of the lessons I learned from building and running a group psychology practice, along with insights gained through working closely with thousands of health practitioners around the world.

Zanda also has a growing library of free practice management webinars, covering topics such as building a sustainable practice, improving the client experience, setting effective boundaries, marketing, technology and AI. This includes a webinar specifically exploring the ideas and stories behind the book.

The Zanda blog is another useful resource for practical guidance on running and growing a health practice. It covers everything from starting and scaling a practice to productivity, compliance, security, marketing and the responsible use of AI.

I also share my latest thinking on private practice, psychology, leadership, technology and AI on LinkedIn.