On this page
Trusted Works began by helping hospitals solve one of their most immediate workforce problems: filling open shifts. When a shift goes unfilled, the consequences are severe. Managers spend hours searching for coverage, clinicians absorb the pressure, and hospitals may be unable to staff as many beds as their communities need. The problem costs money, drives stress, and limits patient care — which is why we were able to drive so much value by tackling it.
Solving that problem also gave us a close view of how open shifts are created. After years of working downstream — responding to vacancies, call-offs, changing patient demand, and gaps the original schedule could not absorb — we came to a simple realization. Every open shift is born in a schedule.
What filling shifts taught us about schedules
Building a schedule is complicated. Knowing how many clinicians you need is the easy part; deciding which people fill those roles — qualifications, availability, required hours, preferences, the rules of that specific unit — is where the real work lives.
Unfortunately, there is no one-size-fits-all success metric for scheduling, either. Schedulers continually balance coverage, skill mix, fairness, employee preferences, labor costs, and the disruption of changing assignments people have already accepted. The weight of each factor varies by organization and unit, and it can shift from one schedule cycle or season to the next. It's a complicated and nuanced process, and anyone telling you they have scheduling figured out hasn't watched a scheduler work.
Most of the knowledge needed to navigate those tradeoffs does not live in existing scheduling systems. It lives with experienced schedulers who know which clinician can charge, who has already been moved twice, who is trying to avoid another weekend, and what accommodations were made during the last cycle. That context is distributed across spreadsheets, text messages, disconnected systems, and institutional memory. What appears to be scheduling software working is often a skilled person compensating for what the software cannot see.
And those tradeoffs don't stop when the schedule is published. They're often just shifted to a new group of people. Managers, staffing teams, and clinicians keep renegotiating the schedule through call-offs, swaps, time-off requests, and changing coverage needs. By the time those changes become open shifts, the work is more urgent, more expensive, and more disruptive.
Why we acquired ShiftOS
Our experience filling those shifts led us to a clear conclusion: we could not solve the entire downstream problem without moving upstream.
That is why Trusted acquired ShiftOS, including Holly, its AI scheduling agent.
Built by a clinician-led team, Holly uses more than 50 specialized AI agents, each responsible for a distinct part of the scheduling workflow. It coordinates schedule generation, call-offs, shift swaps, paid-time-off requests, and credential tracking. Staff can reach it through the channels where scheduling conversations already happen — text and phone — and it completes approved actions rather than recommending what someone else should do.
What interested us most, however, was not a collection of AI features. It was the architecture and know-how underneath them.
Combining intelligence with control
Building an AI-native platform from the beginning forces a team to confront questions that rarely arise when AI is added to an existing product. How should a system respond when a clinician texts at 5 a.m. and expects a shift swap resolved before work begins? How can it retain the operational context that previously lived only in a scheduler's head? How can it determine whether an agent made the right decision, catch the failures, and improve over time?
Answering those questions takes more than a conversational interface. It takes an evaluation system that continuously measures what the agents did, whether the result was correct, and where performance needs to improve. That architecture — and the operational learning embedded in it — is not something you get by adding a chatbot to existing software.
ShiftOS built this architecture with this exact workflow in mind — and proved it in production across organizations from 25 to more than 2,000 employees. What Holly learns from each deployment is precisely the soft context this work runs on: individual preferences, unspoken rules, the interpersonal dynamics of a unit, the accommodations that shape who can take which shift. And the people using it never have to learn new software — staff text or call Holly the way they already negotiate their schedules with each other today.
At the same time, a hospital schedule is not a place for unchecked experimentation. Credential requirements, union rules, hour limits, required rest, and approval chains are not preferences for a model to interpret. They are hard constraints that must be enforced consistently and verified afterward. No hospital should receive a probabilistic answer to whether a clinician was properly credentialed for a shift.
Trusted Works already provides that foundation. Its configurable rules, approvals, credentialing, auditability, and integrations have been built through years of supporting complex healthcare workforce operations. Deterministic, verifiable, and boring on purpose.
The opportunity ahead is to combine our purpose-built infrastructure with the AI-native architecture and workflows already proven by Holly. Hard constraints stay enforced by rules that cannot bend. Contextual decisions — balancing employee preferences, fairness, coverage, and cost — leverage technology built to hold context and weigh shifting priorities. We believe the result will be better, more approachable technology: hard rules enforced without exception, soft preferences weighed with context, and answers that come back in the channel where the question was asked.
The people using the system should never need to know where one kind of decision ends and the other begins. A clinician requests a shift swap in the channel where the conversation is already happening. Behind the scenes, the system checks credentials, hour limits, and labor rules while weighing availability, preferences, and the effect on the rest of the schedule. If the request clears the organization's requirements and approval policies, the system completes the action and answers in the same conversation — in minutes, not days.
Where we go from here
Trusted Works already integrates with the systems health systems run, and Holly's intelligence is designed to work across them. Our immediate focus is accelerating schedule generation within Trusted Works and improving how clinicians and managers interact with their schedules.
Healthcare will always produce unexpected changes, so open shifts will never disappear entirely. But fewer of them should become emergencies. A workforce platform that understands demand, schedules, credentials, preferences, and organizational rules can respond earlier — and coordinate the work when circumstances change.