Therapy is about one hour a week. The other 167 hours, the client is on their own. A lot of new AI tools are rushing to fill that gap, mostly by trying to act like a therapist you can reach any time. We wanted to know what therapists actually think about that. So we built a real tool for the gap, sat down with 15 of them, and asked a blunt question: where does this cross the line?
The short answer surprised us. The therapists we most respected did not want an AI that fills the gap. They wanted one that protects it.
How we did this, and why to take it with a pinch of salt
Between late June and mid-July 2026, we talked to 15 therapists from the TinT community, about 45 minutes each. During the call we showed them the real, working product and watched them react. They ranged from a year into practice to more than seven, across very different styles of therapy, in India and the US.
This is not a clinical trial. It is 15 people from one community, and the person asking the questions was the founder showing his own product. People are kinder when the founder is in the room. We know that. So we paid the most attention to the therapists who disagreed with us, because that is where the honest signal was.
One thing we learned about the method itself: a real, working tool got far more honest answers than slides would have. People only said "this part feels wrong" when a real feature actually crossed a line.
The main thing: therapists do not agree
Everyone said the 167 hours matter. But they wanted completely different things to happen in that time. They fell into four camps, and they genuinely disagree with each other.
Four camps, one gap
Tap a camp to see what that therapist wants.
These are not preferences you can average out. A Plan therapist and an Empty therapist disagree about what actually heals. You cannot please both by meeting in the middle. You have to let each one set the tool their own way. A tool that hard-codes one of these camps will quietly alienate the other three.
Where they all agreed: help me, do not replace me
Every one of them drew the same line. AI is welcome to write up notes, summarise, spot patterns, hand out resources. It is not welcome to become a therapist in the client's pocket, to make clients dependent, or to decide what matters clinically.
As one put it: "When I'm working through something hard, I don't just want answers. If I wanted answers, I'd Google."
The tools they trusted had one thing in common: the therapist stayed in charge. The therapist pulls information when they prep, instead of getting pinged all day. The client chooses, each time, whether to share an entry. The AI drafts, the therapist decides. Nothing goes out on its own.
The surprise: it is not really about admin
We assumed the pain was paperwork. The sharpest voices dismantled that. Many therapists are not full, so they do not need a big admin system. Note-taking is not even taught in Indian training. Clients ignore booking apps and just message on WhatsApp.
The real pains turned out to be simpler: getting clients, and, for some clients, the gap itself. And for a surprising number, writing the note is the thinking, not a chore to automate away.
Same tool, two countries
The two US therapists showed an almost opposite picture. No WhatsApp, and privacy and insurance rules that change everything. But in the US, insurance actually pays for time spent supporting clients between sessions.
So the same feature is a growth tool in India and a billable one in the US. Channel, privacy, and payment are not details you add at the end. They change what the product is.
The part we have not solved
Almost every therapist eventually asked the same thing: what does it do on the worst night? A tool a client can use alone at 2am has to know when to step back and get a real human involved.
We have a clear rule for that hand-off. We have not finished building it. It is one of the two things that matter most, and we are being honest that it is not done.
What this changed for us
We stopped thinking in features and started thinking in stances. A feature encodes one clinical belief and imposes it on everyone. The therapist's camp, not the feature, is the real unit. Build for the style, not the flow.
This work was done with TinT, whose community made the whole thing possible, and whose founder Harshali Paralikar helped shape it throughout.
If you are a therapist and any of this resonates, or you think we have it wrong, we would genuinely like to hear from you. You can see what we are building, and try it for your own practice, at InnerOS Practice.
