The missing piece between your therapy sessions.
Therapist-directed AI homework, validated outcome measures, and a dashboard that shows what's actually changing — not an autonomous chatbot, an extension of your clinical work.
Build the formulation that drives everything
A visual 5-part model and longitudinal formulation builder. Drag to link components, export to PDF for your clinical records. Every homework exercise your client receives is shaped by this — not a generic script.
Why Companion is therapist-directed, not autonomous
Companion's design choices map directly to four areas of published research: homework adherence, measurement-based care, alliance rupture detection, and the limits of autonomous AI therapy. Not founder opinion — the literature your sceptical colleagues will already know.
A meta-analysis of 17 CBT studies (2,312 clients) found that homework compliance was significantly associated with treatment outcome at post-treatment (quantity g = 0.79, quality g = 0.78) and follow-up (quantity g = 0.51, quality g = 1.07). The strongest finding: homework quality matters more than completion quantity.
Over 40 RCTs support routine outcome measurement (ROM) with feedback as a system for improving outcomes, reducing dropout, and increasing efficiency. A UK NHS Talking Therapies study of 2,233 patients found 8.01% more patients reliably improved when therapists received feedback alerts. Intake-and-discharge measurement alone misses the signal.
A meta-analysis of adult individual psychotherapy found a moderately strong relationship between weaker therapeutic alliance and dropout (d = .55). Pre-dropout sessions show progressive disengagement — reduced participation, withdrawal, missed homework — rather than a single abrupt event. Early rupture detection is the highest-leverage retention move.
A 2026 mixed-methods RCT of Wysa within NHS Talking Therapies concluded there was “no evidence that Wysa treats depression in this study”, with engagement limited by the conversational agent. Woebot's consumer app shut down in mid-2025. The regulatory direction — FDA, MHRA, APA — is toward supervised AI as a clinical adjunct, not autonomous therapy. Companion is built for that direction of travel.
Further reading: Lebeau et al. (2013) on homework adherence and outcome variance; 2023 systematic review (PMID 37104804) on between-session collaborative design; Safran & Muran rupture taxonomy. Full reference list available on request.
CBT is a 168-hour-a-week therapy. We see clients for one.
Most therapeutic change happens between sessions. But homework completion rates sit at 50–80% in best-case studies — and the worksheets that do come back are often skimpy, late, or never reviewed properly.
Built for clinical work, not engagement metrics
Evidence-based exercises, validated outcome measures with RCI and NHS Talking Therapies benchmarks, suicide-risk monitoring built in, and pattern detection that runs continuously. Built by a practising CBT therapist, used inside the formulation you set.
Formulation-informed conversations
You set the 5-part formulation and longitudinal history. The AI uses it to shape every homework exercise — Socratic, evidence-based, never freelancing. No competitor offers this.
PHQ-9, GAD-7, CORE-10, WSAS with RCI + NHS Talking Therapies
Automatic scoring, Reliable Change Index calculations, comparison against NHS Talking Therapies recovery benchmarks. Outcome evidence commissioners and insurers can use — one-click PDF report.
Q9 suicide-risk monitoring
PHQ-9 Question 9 monitored automatically with configurable alert thresholds. Historical risk tracking. A clinical safety net — without turning the app into a crisis service.
Pattern + alliance detection
Cross-exercise pattern analysis identifies recurring core belief themes. Engagement signals flag potential alliance ruptures before they become dropouts. The clinical intelligence runs in the background.
Who Companion is for
Qualified CBT therapists who care about outcomes — and want their clinical expertise to extend beyond the room.
Differentiate on outcomes
Show prospective clients you measure what matters. Hand commissioners and insurers RCI-backed evidence. Cut admin time on homework review and write better, more focused sessions.
Built around the standards you already use
PHQ-9, GAD-7, CORE-10 and WSAS with NHS Talking Therapies benchmark comparisons. Q9 monitoring. Reports that line up with the metrics your service is measured on. UK data residency available.
Stretch your clinical reach
Make therapist time go further by giving clients structured between-session support that you direct. Demonstrate impact to funders with outcome reports they can quote in their next bid.
What's coming next
We build deliberately. Each addition has to earn its place against the same standard as everything that's already shipped: built by a practising supervisor, grounded in evidence, and useful in real clinical work.
Pre-session intelligence briefing
A 30-second read before each session. What your client did this week, where engagement dipped, any new patterns the system surfaced, and a suggested agenda — so the time you have with them is the highest-value time it can be.
Voice mode for client conversations
Some clients won't type. Older adults, dyslexia, neurodivergence, busy parents, low literacy. Voice mode opens Companion to all of them — same engine, same formulation, same outcome tracking. Just spoken.
Crisis-aware response protocol
If a client's PHQ-9 Q9 elevates mid-conversation, Companion shifts its stance — acknowledge, ground, signpost (Samaritans, 111, A&E), and immediately alert you. Built around UK crisis pathways and the safeguarding standards your regulator already expects.
Specialty modules
Perinatal mental health, OCD with ERP scaffolding, insomnia (CBT-I), health anxiety, chronic pain, eating disorders. Each module purpose-built around its evidence base, not a generic exercise library re-skinned. Sequenced by clinical need and your network's reach.
Your clients' patterns flow into your supervision — with your consent.
Anonymised client patterns from Companion can flow into the supervision side of Supervisia, so the things you'd struggle to remember by Friday afternoon are already there when you sit down with your supervisor. No competitor can do this because no competitor has all three pillars — train, supervise, practise.
Waitlist members get early access to each capability as it ships, plus a say in what we build next.
Be among the first therapists using Companion.
Companion is rolling out to a select group of UK CBT therapists first. Join the waitlist for early access, founding-member pricing, and direct input into the roadmap.
No spam. Unsubscribe with one click. We won't share your address.