Greenspace Health

Artificial intelligence is a feature layer on a measurement platform whose value stands without it. What most customers buy is the administration and scoring of published assessments, more than 500 of them, with automatic charts, intake screening and outcome reporting, and those instruments are scored as published rather than by a model. Greenspace says it uses AI and machine learning within the platform. Its MBC 2.0 features include Session Prompts, which generate conversation prompts for the clinician from a client's symptom data, a generative function that runs on a trained model, and natural language risk scoring, described as in beta when it was announced in July 2024.

Remove these features and the measurement platform remains, which is why this is C.

Ask which features run on a learned model, which model it is, and which features are rules applied to assessment scores.

The AI features advise the clinician, and the structure is described with the risk part missing. Session Prompts are suggestions a provider requests, can regenerate when one does not fit and can rate in the product, and Greenspace says they are designed to support clinical judgment, not to replace it. The Caseload Assistant highlights clients on clinically relevant measures, reliable deterioration, suicidal ideation, engagement and therapeutic alliance, for the clinician to review, and the measures behind each highlight are the client's own assessment results.

What is not published is what happens after a suicidal ideation highlight: who is notified, how quickly, and whether anything routes beyond the clinician's dashboard.

Ask how suicide risk responses are escalated, and what Session Prompts are prevented from suggesting.

The approach is described in general terms with nothing identified. Greenspace says Session Prompts are generated from a client's Symptom Insights and Focus Areas using instructions developed in house with clinical experts, and that provider feedback improves the next suggestion. No model, model provider or version is named, and nothing describes how natural language risk scoring works or how customers learn of changes.

Ask which model generates Session Prompts, how it is versioned, and how a change is communicated.

Hosting is named and the model layer is not. Greenspace names Aptible on AWS and Google Cloud Platform as hosts, and says other authorized service providers support the platform under contractual privacy and security obligations. No subprocessor list is published, and no model provider is named for Session Prompts or natural language risk scoring.

Ask for the subprocessor list, including any language model provider, and the retention and training terms that govern data sent to it.

Named customers and vendor reported averages, with no method behind the figures. Greenspace reports a 49 percent average symptom improvement, an 81 percent average therapeutic alliance score and an 83 percent average assessment completion rate, and says aggregate data from 482 partner implementations shows that clients who complete more assessments have better outcomes. No period, denominator, comparison group or study design is published with these figures, and they describe measurement based care as practiced on the platform rather than the AI features. Named customers include Sheppard Pratt, Merakey, Catholic Charities, Maryland Behavioral Health Solutions, Homewood Health, PathStone and UConn Health.

No evaluation of Session Prompts, the Caseload Assistant or natural language risk scoring is published.

Ask for the method behind the outcome averages, and for any evaluation of the AI features against clinician judgment.

A categorical commitment on AI data use is published, without the retention schedule behind it. Greenspace says only de identified data, with all direct identifiers removed, is used for AI purposes, that customers can opt out of AI and machine learning data use at any time, and that its use of AI is governed by a Responsible Use of AI Policy covered by its annual SOC 2 Type II audit. The platform privacy policy says personal information is kept no longer than necessary, and that non identifying, aggregated information is used for platform improvement, publication and a research study at the University of Toronto.

No retention period is stated, the de identification method goes no further than removing direct identifiers, and nothing says whether a language model provider receives identifiable data when Session Prompts are generated. Safety engineering, meaning how generated prompts are checked and how a safety event is handled, is not published.

Ask for the retention period, what data reaches the model behind Session Prompts, and whether that provider retains or trains on it.

Business associate status is stated in the platform's privacy policy, and the agreement is not published. The patient privacy policy for the Greenspace app says Greenspace will only use or disclose information as permitted by the controlling business associate agreement, and gives patients rights of access and correction. Greenspace's privacy and security page states HIPAA compliance and displays PIPEDA compliance for Canada. The website privacy policy, updated July 2025, covers the public site only and points to the separate platform policy, last updated in May 2021.

No business associate agreement is published, and nothing says whose agreement governs, Greenspace's standard form or the customer's.

Ask for the business associate agreement, whether it binds Greenspace's subcontractors, and when the platform privacy policy will be updated to cover the MBC 2.0 features.

A recognized certification is named without the report. Greenspace says it has completed a SOC 2 Type II review by an independent AICPA auditing firm, renewed annually, and describes annual independent penetration testing, monthly automated vulnerability scanning, and AES 256 encryption at rest. No report date is given and no trust portal is published. The ISO 27001 and HITRUST certifications shown on its security page belong to its hosts, Google Cloud Platform and Aptible on AWS, and do not count here.

Ask for the current SOC 2 Type II report with its period, scope and any bridge letter.

No device claim is made, and the AI features are scoped as support for the clinician. Greenspace positions the platform as measurement based care software, scores published assessments by their own rules, and says Session Prompts support clinical judgment rather than replace it. No regulatory position is published for the Caseload Assistant, which prioritizes clients partly on suicidal ideation and deterioration, or for natural language risk scoring. The highlights rest on the client's own assessment answers, which a clinician can open.

Ask how Greenspace classifies its risk scoring and caseload prioritization for regulatory purposes in the United States and Canada.

Responsible AI is committed to in policy, with no evaluation published. Greenspace says a Responsible Use of AI Policy governs its use of AI and that the policy is covered by its annual SOC 2 Type II audit; the policy itself is not published. No results are published for Session Prompts or natural language risk scoring across age, language, diagnosis or cultural groups, which matters for prompts and risk scores applied across child, adult and addiction services.

Ask for the policy, any evaluation of risk scoring by group, and how clinicians' feedback on prompts is reviewed.

Review mechanisms exist and nothing stands behind the AI output. Session Prompts can be regenerated and rated by the provider, and Caseload Assistant highlights rest on assessment answers the clinician can check. The platform privacy policy gives patients a right to access and correct their personal information by contacting Greenspace directly, which covers what Greenspace holds rather than a correction of a risk score or prompt. No error rate, warranty or remedy attaches to any AI output.

Ask how a wrong risk score or highlight is corrected, and what Greenspace commits to if a feature misses a client at risk.

Named systems and standards with named deployments, with two way exchange claimed in general. Greenspace lists more than 85 record systems, including Epic, Oracle Cerner, Meditech, athenahealth, NextGen Healthcare, Netsmart myAvatar, Welligent, Qualifacts, Streamline Healthcare, Kipu and AdvancedMD, over HL7, FHIR, SMART on FHIR, a REST API and SFTP, with single sign on. It names deployments including Merakey on Netsmart myAvatar and Advanced Behavioral Health on Dr. Cloud, and says many integrations exchange demographics, program information, assessment results and outcome insights in both directions, without saying which. Greenspace also has a partner listing in the Qualifacts marketplace.

Ask which integrations write results back into the record at your record system, and for a reference customer running it.

A hosted service with the US data location committed, and isolation not described. Greenspace says all protected health information for US customers is stored in the United States, on Aptible running on AWS and on Google Cloud Platform, and runs a separate Canadian application. Where Canadian customers' data is stored is not stated on the US pages, no on premises or private option is offered, and tenant isolation is not described.

Ask where Canadian data rests, which cloud hosts which data, and how customer data is separated.

No price is published and how the product is sold can be established. Greenspace sells three platforms, for measurement based care, client intake and population health, to organizations rather than to individuals, offers a demo route and a clinic sign up form on its app, and lists more than 85 record system integrations. No price, tier, unit of charge or implementation fee is published.

Ask for the price per clinician or per organization, whether the MBC 2.0 features cost extra, and what integration work is charged.

Coverage is stated clearly, with nothing validating it by setting. Greenspace names outpatient mental health and addiction clinics, psychiatric hospitals, health systems, community organizations, schools and campuses, workplace programs and insurers, in Canada and the United States, and its customers include child and family services such as The Children's Guild alongside adult services. The assessments it administers carry their own published validation, but the platform's outcome averages are not broken out by setting or age group.

Ask which settings and age groups the outcome figures come from, and whether the AI features are offered in every setting.