AI Assistant & Connectors
Live todayAsk Claude or ChatGPT about your ABA center — without PHI risk
Connect in one click, sign in with TargetFlo, and ask about staff, credentials, scheduling, authorizations, sessions, referrals, and tasks. Answers come from aggregates, never a client record.
What are TargetFlo's AI connectors?
TargetFlo's AI connectors use the open Model Context Protocol to give Claude and ChatGPT read-only access to your organization's operational aggregates — staff, credentials, training, scheduling and authorization utilization, sessions, referrals, tasks, and portal adoption. It is an AI operations assistant for healthcare that uses non-PHI data only. Separately, clinical form AI is live today: Gemini suggests text on ADOS-2 reports, child assessments, and Plans of Care; licensed clinicians review, edit, and sign every report. See AI integrations and Google Cloud security.
Live today
Clinical AI is assistive. Operations AI is read-only.
Gemini compose on ADOS-2 reports, child assessments, and Plans of Care is live: fill empty fields, refine a section, or draft from context. Nothing saves until a clinician accepts. The Claude and ChatGPT connectors below are a separate, non-PHI operations product.
ADOS-2 reports
Draft observations and impressions from scoring context. The examiner owns every sentence.
Learn morePlan of Care / BIP
Suggest medical-necessity narrative and goal wording from assessment fields already on the record.
Learn moreOCR, compose, MCQs
The same Gemini integration reads faxes, drafts broadcasts, and generates course questions.
Learn moreExample prompts
Questions your team can ask Claude or ChatGPT
Every one of these is answerable from counts and rates. None requires knowing who a client is.
- 01
“Give me an overview of our organization this week.”
- 02
“Whose credentials expire in the next 60 days?”
- 03
“What percentage of RBTs have completed the HIPAA refresher course?”
- 04
“How booked is the schedule against authorized hours this week?”
- 05
“Which authorizations are above 90% used?”
- 06
“How many sessions were submitted late or are still waiting for a signed note?”
- 07
“How many referrals moved through each intake stage this month?”
- 08
“How many tasks are overdue, by team?”
- 09
“How many families have signed in to the parent portal?”
- 10
“Summarize billed vs collected for the last three months.”
How booked is the schedule against authorized hours this week?
82% of authorized hours are scheduled this week. 3 authorizations are above 90% used, and 2 clients are under 60% booked.
Whose credentials expire in the next 60 days?
4 credentials: 3 RBT certifications and 1 CPR card. The earliest expires in 12 days.
What percentage of RBTs completed the HIPAA refresher?
87% organization-wide (61 of 70). Durham is at 74%, the lowest — 5 RBTs outstanding.
Data the assistant can use
Six operational areas, all aggregates
Organization overview
Headline numbers across the org: clients, staff, sessions, open tasks, and what needs attention.
Staff, credentials & training
Who is on the team, whose credentials expire soon, and course completion by course and role.
Scheduling & authorizations
Booked vs authorized hours, utilization per CPT line, and authorizations nearing their limit — as counts and percentages.
Sessions & tasks
Sessions submitted, late, or waiting for a signed note, and task counts by status and team.
Referrals & portal
Referral funnel counts by stage and parent portal adoption.
Revenue (Admins & HR)
Billed vs collected summary, only for the roles that already see revenue in TargetFlo.
Boundaries
What the assistant can and cannot see
The line is the PHI schema. Everything on the left is operational. Everything on the right is isolated and unreachable by the connectors.
| Can see | Cannot see |
|---|---|
| Organization overview and KPIs | Client names, dates of birth, addresses |
| Staff list, credential expirations, training completion | Diagnosis codes or clinical assessment results |
| Schedule and authorization utilization (counts, hours, percentages) | Session data, notes, or treatment details for a person |
| Session metrics: submitted, late, waiting for a note | Fax contents, documents, or images |
| Referral funnel counts by stage | Member IDs, payer responses, or coverage details |
| Task counts by status and team; portal adoption | Task text that mentions a client |
| Revenue summary (Admins and HR only) | Anything stored in the PHI schema |
Safety
AI access is scoped to operational aggregates. PHI never leaves your controlled environment.
The assistant is useful precisely because it is limited. It answers the questions a leadership meeting asks, and it is structurally unable to answer the questions a breach would.
Server-side aggregation
Counts and rates are computed inside TargetFlo. Row-level data is never serialized to the model.
Role-based sign-in
Each connection belongs to one person and one organization, limited to their role's topics. Disconnect at any time.
No PHI schema access
The connector's database role can read only its own views — no client, guardian, insurance, or document tables.
Audit log
Every tool call — who, what, when, which parameters — sits in the same log as PHI access events.
Architecture note
How a question becomes an answer
Four steps, all inside TargetFlo's boundary.
- 01
Connect in one click
From AI connectors in TargetFlo, choose Claude or ChatGPT and add the connector. No keys or code to copy.
- 02
Sign in with TargetFlo
Claude or ChatGPT sends you to TargetFlo's sign-in. You pick the organization and allow access — your role decides the topics.
- 03
Ask in plain language
The assistant calls a fixed set of read-only tools. Each one reads views built for this purpose, never the clinical tables.
- 04
Every call is checked and logged
Role, organization, and the connector switch are checked on every call. Tool name, timing, and result size go to the audit log — never the answer.
Tools
Live todayA fixed set of read-only tools
The assistant can only call these tools, and only the ones in your role's topics. Inputs are choices and numbers — never free-form queries.
- Admins turn connectors on or off for the whole organization
- Owners and admins see and disconnect connected apps
- Usage by day, tool, and outcome for admins
- Works with Claude and ChatGPT through the open Model Context Protocol
- get_org_overview
- list_staff
- get_expiring_credentials
- get_training_completion
- get_schedule_utilization
- get_session_metrics
- get_authorization_utilization
- get_referral_funnel
- get_task_summary
- get_revenue_summary
FAQ
AI connector FAQs
Related features
See Claude answer questions about your center
Book a demo. We connect Claude to a demo organization and ask the questions your Monday meeting asks.