The convention in therapy is the 50-minute hour. It comes from psychoanalysis, where it gave the analyst time between patients to write notes. It survived in modern practice because it's structurally convenient — clinicians can see roughly one client per hour, and the cadence works for billing, scheduling, and the rhythm of a working day.
It's the dominant unit of mental health care. It's also not the unit users tell us they actually want when we offer them other options.
Why we constrained session length in the first place
The choice to bound sessions wasn't originally about engagement. It started as a clinical safety decision.
Most AI-assisted mental health products being built today sit on top of general-purpose large language models. Those models are powerful, but they're open-ended by design. There's no natural endpoint to a conversation, no built-in arc, no moment where the system asks what the user wants to do with what they've shared. Someone using a general-purpose chatbot for emotional support can talk for an hour, two hours, three — drifting through whatever surfaces, validating things that shouldn't be validated, escalating into territory the system isn't equipped to handle, and ending without anything therapeutic actually having happened.
The experience is therapy-shaped. It isn't therapy.
When we started designing Thoughtful, we wanted to build something different. The first design decision wasn't about session length — it was about session structure. Every session needed to have a beginning, a middle, and an end. Every session needed to land on something specific: a reframe, a script for a hard conversation, an action the user could take. The system needed to know when to stop.
Bounded sessions were the only way to deliver that. You can't build a reliable therapeutic arc inside an open-ended conversation. Therapeutic intent requires a container.
The question of how long the container should be came after. We started with the structural commitment and worked toward the optimal duration. Ten minutes is where we landed — but the bound came first.
The bandwidth barrier
What we discovered as we tested shorter sessions in pilots is that the constraint we'd built for clinical reasons turned out to solve a second problem we hadn't initially designed for.
The conversation about why people don't use mental health benefits has historically been about stigma. What we hear from members and what the engagement patterns suggest is that the dominant obstacle for today's workforce is different. It's not how they feel about asking for help. It's how much time, attention, and emotional bandwidth it takes to actually access it. In a working culture where the permission to be present is structurally low, even the minutes required to find a provider, book a session, and prepare for one can be enough to keep someone from doing it.
We've started calling this the bandwidth barrier. Stigma still exists, but for most workers today, it isn't the load-bearing one. The harder barrier is room — and the support that gets used is the support that fits inside the room people actually have.
The clinical decision to bound sessions and the workforce reality of constrained bandwidth turned out to align. Ten minutes is short enough to deliver a complete therapeutic arc and short enough to fit into the gaps that already exist in someone's life. The shape we needed clinically and the shape members could actually use turned out to be the same shape.
That alignment isn't an accident, but it also wasn't engineered for engagement. It emerged because we started with the harder constraint.
What we found
Thoughtful sessions come in three lengths: a five-minute structured session for a quick reset, a ten-minute open session to talk about something new or continue a topic, and a thirty-minute deep dive for a fuller emotional triage.
Ten-minute sessions are the most popular and the most effective. Users see behaviour change with three ten-minute sessions per week. The deep dive matters — it has a specific role in the architecture — but it's not where most of the work happens. Most of the work happens in bite-size support, used more frequently.
This wasn't a finding we engineered for. We had constrained session length for clinical reasons. Within that constraint, ten minutes emerged as the sweet spot — shorter than the regulation-focused five-minute session, longer than what's possible in a single check-in, and frequent enough to compound across a week. Therapy's convention is fifty minutes, weekly. The most engaged Thoughtful users are doing roughly the inverse — bite-size support, several times a week. The total time investment is comparable. The shape is different.
Why this shape works for the population we serve
If you're an employee dealing with a stressor — a hard conversation at work, a sleepless night, money pressure, an argument at home — the unit of help you actually need is rarely a fifty-minute session next Tuesday. It's ten minutes now.
Ten minutes is long enough to talk something through, get a reframe, and end with an action you can take. It's short enough to fit into a lunch break, a walk, the gap before bed, or the ten minutes between meetings when everything feels like too much.
Most importantly, it's short enough that the user doesn't have to feel like they're 'doing therapy' to start one. That's not a small thing. The threshold for opening a mental health app is much lower than the threshold for scheduling a session. The unit of engagement has to match.
The before-work group
Three weeks into one of our enterprise pilots, a usage pattern we hadn't predicted emerged. A meaningful share of sessions were happening before 9am — users opening the app first thing in the morning, often before they'd opened anything else.
This is exactly the use case that a ten-minute session is built for and that a fifty-minute model can never serve. It's not therapy. It's regulation. It's an emotional warm-up before a day that's about to ask things of them.
That pattern wouldn't show up in a product where the minimum unit of help is an hour. It only shows up because the support is short enough to belong in that moment.
When longer is right
The thirty-minute deep dive has a specific role. It's the unit for when something needs unpacking — a situation that's been building for weeks, a feeling that doesn't resolve into a single topic, a decision the user wants to work through carefully.
We built the deep dive because some sessions need that runway. The phase architecture inside a deep dive is different from a ten-minute session: more time in the open phase to set context, a longer insights-and-reframe phase before moving into planning. The structure matches the duration.
But it's the secondary unit, not the primary one. The primary unit is bite-size support, used often. The deep dive is what you do when bite-size support isn't enough.
What this changes about how we design
Bite-size support only works if memory is real. The whole model depends on a user not having to re-explain themselves between Monday and Wednesday — not re-introducing the situation at work, the conversation they're dreading, the pattern they noticed last week. If memory isn't there, every session starts from zero, and ten minutes isn't enough time to start from zero and arrive somewhere useful.
What "real memory" means in practice is that each session leaves behind durable artefacts the next one can pick up: a structured summary of what the user was working on, a Recent Topic the assistant can open with, a longitudinal view of the themes that keep returning. Memory isn't transcript playback — it's the shape of the user's situation, held across sessions, so the next session can start in the middle.
Everything else about the experience is downstream of memory doing its job. The home screen can stay frictionless — tap a card, the assistant speaks first, no setup — because the context the assistant needs is already loaded. And bite-size support can fit into the moments between other things — before work, on a break, at midnight — because none of those moments have to be spent re-establishing where the user left off.
The broader point
Therapy's fifty-minute hour was designed around the convenience of the clinician and the rhythm of the office. It worked when therapy was the dominant model of mental health care. It's the wrong unit for a workforce facing a bandwidth barrier — where the permission to be present is structurally low and the minutes between meetings are most of what people have.
Most people who need support don't need it for an hour next Tuesday. They need it for ten minutes, three or four times this week, in the gaps that already exist in their lives. That's what we built for — within a clinical constraint that came first. Get the unit right, get the structure right, and the rest gets easier.