Luma Health Debuts Agent Launchpad for Multistep Procedure Preparation

Luma Health launched Agent Launchpad on November 5, 2026, making its AI-agent orchestration product available for procedure preparation and follow-up. The University of Arkansas for Medical Sciences and Mile Bluff Medical Center are using it for colonoscopy preparation, with Luma reporting a 95% average patient response rate across their outreach.

Agent Launchpad is part of Luma’s Fall 2026 product release. Its initial agents call or text patients, follow the health system’s protocols, answer routine questions and track readiness. The launch templates cover colonoscopy preparation and post-procedure follow-up, as well as preparation for knee replacement, cataract surgery and bariatric surgery. The product is intended to take repetitive coordination away from care teams without removing staff from clinical decisions.

Health systems consulted by Luma reported that nurses averaged 2.3 calls simply to reach a patient for the first time, before further preparation calls. Staff also confirm preparation supplies, collect anesthesia-screening answers, check that patients have a ride home and follow up with patients who have not completed a step. That workload explains why Agent Launchpad is organized around completing a workflow rather than sending isolated reminders.

How Agent Launchpad Coordinates Preparation

A health system can start with a template or select a goal and let AI determine how to reach it. It uploads approved instructions, screening questions and FAQs, then defines what agents may handle and when staff must intervene. Luma’s orchestration documentation separates those agentic jobs, in which specialized agents handle tasks toward a selected outcome, from deterministic workflows in which the customer defines each step.

For colonoscopy preparation, the described workflow follows the provider’s preparation instructions, perioperative logistics and anesthesia-screening rules, with check-ins 14 days, seven days, three days and one day before the procedure. An agent can switch from a phone call to text at the patient’s request, retry after missed calls using patient-engagement data, and answer routine questions in the same conversation. If the appointment is rescheduled, future outreach moves to the new timeline.

Agents working with the same patient share conversation context, allowing separate pre- and post-procedure workflows to run without staff starting each one. Concurrent preparation workflows can also run when the provider’s protocols permit. Clinical questions follow the health system’s escalation rules: the nurse receives the full conversation and can resume, repeat, skip or stop steps. Every engagement leaves a readable message history and a record of the agent’s actions.

Early Results at UAMS and Mile Bluff

Luma reports that 70% of patients responded within the first day of outreach, perioperative nurses saved an average of more than 20 minutes of coordination per patient, and half of Mile Bluff patients were fully prepared without staff phone calls. The figures describe patient responsiveness and staff coordination in those colonoscopy-preparation deployments.

Paige Anderson, assistant clinical service manager at UAMS, said in Luma’s launch blog that the system “now answers routine prep questions for patients, day or night, at the times that work best for them.” Questions requiring a nurse go directly to the team, giving nurses more time for clinical work only they can do. Travel is also a practical concern at UAMS: many patients come long distances, so understanding the requirements and arriving prepared matters.

Tamara Coughlin, director of surgery at Mile Bluff, reports fewer nursing hours spent on preparation calls and patients becoming prepared earlier with less staff burden. Patients can ask questions whenever convenient, an important consideration for colonoscopies and other procedures with multiple steps that require guidance from the perioperative team.

Referral Scheduling Is Next

Luma’s winter roadmap includes additional procedures, among them endoscopy and pain management, plus a capability for AI to read a procedure referral, contact the patient to schedule it and immediately enroll that patient in preparation outreach. These are planned capabilities, not launched ones.

That roadmap would extend agent coordination to the first scheduling contact and carry it into preparation. As the sequence expands, the launch’s staff-control model becomes central: agents handle routine coordination, while each health system defines approved guidance, operating limits and escalation points so that nurses retain decisions requiring clinical judgment.