howtheF
All case studies/Bright Brains

Bright Brains

Bright Brains

6-Agent AI System for Personalized Neuromodulation Protocols

Client Project
What we built
6
AI Agents
Orchestrated workflow
80%
Time Saved
Protocol creation
3
Protocol Variants
Generated per patient
At a glance

“The multi-agent system transformed how we create and validate neuromodulation protocols, giving us unprecedented efficiency while maintaining the human oversight our industry demands.”

Company

Bright Brains is a healthcare company specializing in neuromodulation therapy—using electrical stimulation to treat neurological and psychiatric conditions. They create personalized treatment protocols backed by scientific research for each patient.

Visit Bright Brains

Industry

Healthcare / Neuromodulation

Size

~50 employees

Bright Brains faced a classic healthcare scaling dilemma: their neuromodulation protocols were clinically excellent but operationally unsustainable. Each patient required manual WhatsApp intake, hours of PubMed research, and multiple rounds of doctor-patient communication before delivering a personalized treatment plan.

The bottleneck wasn't expertise—it was time. Doctors were drowning in administrative work instead of doing what they do best: making clinical decisions. Without automation, they couldn't serve more patients without hiring proportionally more staff, making growth economically unfeasible.

“The multi-agent system transformed how we create and validate neuromodulation protocols, giving us unprecedented efficiency while maintaining the human oversight our industry demands.”

Clinical Director

Bright Brains

The problem

Manual processes limiting growth

  • •

    Time-intensive patient intake: Each new patient required 30+ minutes of WhatsApp back-and-forth just to collect basic information and explain neuromodulation therapy.

  • •

    Research bottleneck: Doctors spent hours searching PubMed for relevant studies before creating each protocol, pulling them away from clinical work.

  • •

    Protocol creation delays: Multiple rounds of revision between doctors and staff meant 3-5 days from intake to protocol delivery.

“We were turning away patients because we simply couldn't create protocols fast enough. Our clinical expertise wasn't the bottleneck—our processes were.”

Use case01

WhatsApp Patient Intake & Anamnesis

We built a conversational AI interface on WhatsApp that handles the entire patient intake process. Instead of a static form, patients engage with a friendly, intelligent agent that guides them through initial screening.

Once eligibility is confirmed, the system intelligently routes the patient to one of four specialized anamnesis agents—each trained on specific neurological conditions—to conduct a deep-dive clinical interview. This ensures every relevant medical detail is captured before a doctor ever sees the file.

1
AI Agent

Initial Patient Contact

WhatsApp agent greets patient, collects basic info, and briefly explains what neuromodulation can do for them.

Core Agent Logic
1
Greets patient with warm, professional introduction.
2
Collects essential identification and contact details.
3
Provides brief, accessible explanation of neuromodulation therapy.
4
Assesses initial eligibility criteria.
5
Routes to appropriate anamnesis pathway based on profile.
Inputs
Patient messageBasic info request
Outputs
Patient profileRouting decision
2
Automation

Anamnesis Routing

Based on patient criteria, routes to one of 4 specialized anamnesis agents to profile the patient with targeted questions.

Core Agent Logic
1
Analyzes patient profile against routing criteria.
2
Determines optimal anamnesis pathway (1 of 4).
3
Triggers handoff to specialized anamnesis agent.
4
Logs routing decision for audit trail.
Inputs
Patient profileRouting criteria
Outputs
Assigned anamnesis type
3
AI Agent

Structured Anamnesis Collection

Specialized anamnesis agent asks handful of targeted questions to build comprehensive patient profile for protocol generation.

Core Agent Logic
1
Loads condition-specific question template.
2
Conducts conversational medical history collection.
3
Validates responses for completeness and consistency.
4
Identifies contraindications or red flags.
5
Compiles structured patient profile for protocol generation.
Inputs
Anamnesis templatePatient responses
Outputs
Complete anamnesis data
Use case02

PubMed Research & Protocol Generation

With a complete patient profile in hand, the Research Agent accesses PubMed's API to retrieve the latest clinical studies relevant to the specific case. It filters thousands of papers to find the 5-7 most applicable citations.

Armed with this research, the Protocol Agent then generates three distinct treatment variations (conservative, standard, and aggressive). This gives the clinical team options to choose from, rather than starting from scratch.

1
AI Agent

Scientific Research

Research agent searches PubMed for 5-7 relevant scientific papers based on anamnesis criteria and saves findings to database.

Core Agent Logic
1
Extracts key medical parameters from patient anamnesis.
2
Constructs optimized PubMed search queries.
3
Retrieves and ranks 5-7 most relevant scientific papers.
4
Extracts protocol-relevant findings from abstracts.
5
Saves research summary to patient database record.
Inputs
Anamnesis criteriaResearch parameters
Outputs
Research summaryPaper citations
2
AI Agent

Protocol Variant Generation

Protocol agent creates 3 variations of the treatment protocol based on research findings and patient profile.

Core Agent Logic
1
Analyzes research findings for treatment options.
2
Generates conservative protocol variant (safest approach).
3
Generates standard protocol variant (typical treatment).
4
Generates aggressive protocol variant (maximum efficacy).
5
Structures each variant with dosage, frequency, and duration.
Inputs
Research dataPatient profile
Outputs
3 protocol variants
3
Automation

Slack Review Request

Protocol variants sent to company Slack for doctor approval with full research context.

Core Agent Logic
1
Formats protocol variants for Slack readability.
2
Attaches research citations and key findings.
3
Sends to designated physician review channel.
4
Awaits approval, rejection, or modification request.
Inputs
Protocol variantsResearch summary
Outputs
Slack notification to doctors
Use case03

Human-in-the-Loop Approval & Delivery

The generated protocols are sent directly to a private Slack channel for doctor review. Physicians can approve a protocol with a single click or reply with feedback like "Lower frequency to 2x/week" to trigger an instant revision.

Once approved, an automation script generates a branded, professional PDF treatment plan and delivers it to the patient via WhatsApp, completing the cycle in hours instead of days.

1
Human Review

Doctor Review & Iteration

Doctors review protocols in Slack. If rejected, they reply with reasons and research/protocol agents iterate until approval.

Core Agent Logic
1
Physician reviews all 3 protocol variants in Slack.
2
Evaluates against clinical judgment and patient specifics.
3
Approves preferred variant OR provides rejection rationale.
4
Rejection triggers automated iteration with feedback loop.
5
Final approval locks protocol for patient delivery.
Inputs
Protocol variantsResearch context
Outputs
Approved protocol OR rejection feedback
2
Automation

Branded PDF Generation

Upon approval, Google Apps Script generates branded PDF treatment plan for client delivery.

Core Agent Logic
1
Receives approval trigger from Slack workflow.
2
Populates branded template with protocol details.
3
Generates professional PDF with company branding.
4
Stores PDF in patient record for compliance.
5
Triggers delivery workflow notification.
Inputs
Approved protocolBranding templates
Outputs
Branded PDF protocol
3
AI Agent

WhatsApp Delivery & Handoff

WhatsApp agent sends client-facing protocol PDF and connects patient with human team for next steps.

Core Agent Logic
1
Retrieves patient contact from database.
2
Sends personalized message with protocol delivery.
3
Attaches branded PDF document.
4
Explains next steps and scheduling process.
5
Initiates warm handoff to human clinical team.
Inputs
Branded PDFPatient contact
Outputs
Delivered protocolHuman handoff
Impact

Transformed Protocol Creation at Scale

80%

Faster Protocol Creation

From days to hours

5x

More Clients Served

With same team size

100%

Doctor Oversight

Every protocol reviewed

“We can now serve significantly more clients without compromising on the quality and safety our industry demands.”

Clinical Director

Bright Brains

Tech Stack

LangGraphWhatsApp APIPubMed IntegrationSlack WorkflowsGoogle Apps ScriptMulti-Agent Orchestration
The Fourth Path to Automation

Stop Evaluating. Start Building.

You've seen the results. We bring the infrastructure to launch working agentic workflows in 90 days—that your team actually owns.

Assess → Build → Train Methodology

30-minute call. No pitch deck. Honest feasibility check.

Related

Explore more customers

Discover how other businesses leverage AI automation to transform their operations and achieve measurable results.

WWF

The chatbot made complex carbon credit concepts accessible to communities who had never engaged with such programs before.

P

Program Director

WWF Environmental Programs

1000+

Users Educated

HowTheF (Internal)

This is our methodology made transparent - what you see is what you get when you work with us, but productized.

O

Oliver Gattermayr

Founder, HowTheF

<60s

Generation Time