AI for the work behind the business
Fulcrum™. One chief of staff. A team behind the work.
Ask what changed. Hand off the follow-up. Keep the decisions that need you. We're building Fulcrum™ into an AI chief of staff for operators: one point of contact that coordinates specialist work around your data, your systems, and your rules. The automated analyst you can trust is the foundation. Approved operational work is the next step.
Built by two people who ran the numbers at a national healthcare company as it grew from 2 markets to 50+.
The chief-of-staff layer is in development with design partners. Which workflows, integrations, and access you get is confirmed for each engagement.
We built this before · in healthcare and outside it
- 40+ yrs
- Combined analytical experience, seventeen of them inside healthcare
- 2 to 50+
- Markets scaled during their tenure
- 5 yrs
- Side by side at the same company, on the same system
- Every layer
- Warehouse, pipelines, definitions, reporting, models, agents. Between them, they've built the data infrastructure that drives business decision-making.
Together they built the supply and demand system a national healthcare company ran on as it scaled, and the data science, ML, and data engineering organizations behind it. Between them they have built every layer Fulcrum™ is made of: the data underneath, the definitions, the reports, the models, and the agent on top.
One request, coordinated work
Give it the outcome, not instructions for every handoff.
A useful assistant does more than answer the last message. Fulcrum™'s chief of staff is being built to break a request into work, hand each piece to the right specialist, keep track of what's still open, and bring back a result you can check. The specialists are the mechanism. You never have to manage them.
You
“Prepare the weekly operating review. Check the unusual changes and flag what needs my decision.”
-
Analyst
Review the approved metrics and run down the exceptions.
Every number carries its source and the definition it was computed from.
Reviewed
-
Data checks
Confirm the sources are current and the totals reconcile.
Freshness confirmed. Totals tie across the systems of record.
Reviewed
-
Reviewer
Check the evidence behind each conclusion.
One number depends on a definition you haven't settled. It goes back to you, not into the report.
Question open
Back to you
Review assembled, sources attached. One definition question is waiting on you. Nothing was changed in any system.
You set the priorities. Fulcrum™ coordinates the work between them, and anything outside what you've approved comes back as a question instead of a guess.
The foundation
Ask for the number. Get three. Pick one before the board meeting.
Finance has one revenue figure, the CRM has another, contracts a third, and one analyst spends the week reconciling them in a spreadsheet. That analyst is your integration layer. Then AI shows up and makes it worse, because the machines inherit the mess.
Every system is right. They just don't agree.
Scheduling counts a booking, billing counts a claim, finance counts what the payer paid. One agreed definition per number, written down, so all of them roll up to the same figure in the board deck and the dashboard.
The definition lives in one analyst's head
Why Q2 got restated, what counts as a referral. One person knows, and it isn't written down. Every definition goes in one place, owned by you, so when someone leaves, the number stays.
AI that never says "the data can't support that"
Point a copilot at ungoverned tables and it launders the garbage into a fluent, plausible, wrong answer. Fulcrum™ answers only from agreed definitions, shows its sources, and flags the gap.
The answer arrives after the decision
A payer wants a rate, a site is off plan, and the number lands a week later, after the call got made. Fulcrum™ answers the follow-up from the same definitions the report used, while the decision is still open.
The model was never your bottleneck. The data underneath it is. Fix the in, and the out becomes a number you can actually believe.
Data maturity is a ladder, and each rung holds up the next.
Most AI programs never deliver what they promised, and the model is rarely why. Companies jump to the top rung with nothing under it, and the initiative dies in the reconciliation meeting.
Rung 1
Data you can rely on
The systems of record wired in, reconciled across each other, with quality checks in front of every number.
Rung 2
Definitions everyone shares
One written definition behind each number, agreed by the people who use it, so finance, operations, and the board mean the same thing.
Rung 3
Metrics you trust
Backward-looking reporting that survives the board meeting. What happened, why, and where the source is. This is where trust gets earned.
Rung 4
ML and AI that hold up
Forecasts, models, and agents built on the three rungs below them, so the answers are supported by a framework instead of a hope.
A chief of staff is only useful if it understands the business. A bigger team of agents doesn't fix an unreliable number. We start with an assessment of where your data sits on that ladder, then build the missing rungs in order. For analytical work that means connected data, agreed definitions, and calculations anyone can check. For operational work it means written procedures, current records, and hard limits on what may change. Then the chief of staff has something worth coordinating.
One Fulcrum™, two capabilities
Understand what happened. Follow through on what comes next.
Fulcrum™ Analyst
Verify before you answer.
An automated analyst built on your business definitions. It investigates what changed, produces the recurring reviews, answers the follow-up with sources attached, and says so when the data can't support an answer.
“What changed in collections, and which part is timing rather than performance?”
See how the analyst gets builtFulcrum™ Operator
Verify before you act.
An operational assistant for back-office work you've specifically authorized. It reads the current record, asks when two readings are plausible, makes the approved change, and checks what the system actually saved.
“Check this week's hours and prepare the corrections. Don't submit payroll.”
See Fulcrum™ OperatorStart with the one you need. The chief of staff is being built to coordinate both as your approved workflows grow, and the scope you sign decides which functions you actually get.
What happens when Fulcrum™ Analyst takes control of your data.
The report is done before you sit down
The morning report is already built. Every number checked, the sources attached. The questions you have in the staff meeting already have verifiable answers.
Ask the follow-up. Get an answer, not a project.
Something looks off, so you ask Fulcrum™ why. The answer comes back with its sources, from the same definitions the report used. No export, no spreadsheet. No multi-day analysis from your data team before you have more questions and still aren't at the root of the question.
It finds you when something's off
Exception alerts, not dashboards you have to remember to check. When metrics become concerning, you get an alert with details, before it becomes a discussion topic and a deep dive at the next board meeting.
It says so when it doesn't know
When the data can't support the answer, Fulcrum™ flags the gap instead of filling it. You know which numbers to bet on and which ones to wait for. Fulcrum™ learns as it manages, gaining critical business knowledge to provide continuous improvement over time.
All of it runs on the same agreed definitions, so the number in the report, the alert, and the board deck is the same number. Which reports come first goes in your Blueprint. Each one moves over as it earns trust.
Fulcrum™ Operator · back-office work for small operators
Text it the work. It does the work.
An AI on a phone number. You text it the way you'd text an office manager: move this shift, add these hours, who's on Saturday night. It checks the request against what's already on the calendar, asks when two readings are plausible, makes the change inside the software you already use, and texts back what it did.
Running today at Love Never Fails, a Colorado home care agency, and at an assisted living community one of our founders co-owns.
Today it's a narrow job: scheduling changes, hours entry, who's on. The chief-of-staff layer that coordinates that work across specialists and tracks what's still open is in development. Pilot scope is confirmed for each operator.
From the pilot
- 132.5
- caregiver hours entered from plain text messages in two days of production use
- 5
- errors caught before entry: duplicate shifts and totals that didn't match the clock
- 0
- entries made on a guess. Every ambiguity went back to the owner as a question
- 2
- platforms it runs today, ShiftCare and Gusto, neither through an API
Five steps, and a rule at every one.
-
1 · A text arrives
You text a dedicated number. Only the people you name can give it work.
-
2 · Read the record
It reads the current schedule before it touches it, never what it remembers.
-
3 · Ask, don't guess
Two plausible readings means one question. A guessed entry in payroll is a wrong paycheck.
-
4 · Do the work
Inside the software you already run, with its own credentials, on screen. No API required.
-
5 · Confirm
It reads back what the system computed, not what it typed, then tells you in one line.
Every rule is there because something went wrong once.
Check the math against the clock
Totals get stated from memory. Clock times are the record. A texted shift said seven and a half hours, and the start and end times said eight. It went back as a question with the difference attached, and nothing was entered until the owner chose.
Verify against what the system computes
Reading back a field you just typed proves nothing. The night the clocks change, an overnight shift is thirteen hours, not twelve. The scheduling system's own total caught it. Our arithmetic didn't.
Published is not the same as saved
In most schedulers an edited shift notifies the worker right away and a new one sits as a draft nobody can see. Before it tells you a schedule is live, it checks the state of each shift it names.
Authority expands only after the workflow earns it.
Access, required approvals, and what counts as done are defined before it runs. Preparing payroll data is not permission to submit payroll. Drafting a schedule is not publishing it. A configured task runs inside its limits. A new kind of action needs a new scope decision.
Running now, in pilot
Scheduling changes, hours entry, who's on
By text, on ShiftCare and Gusto. It reads the current record, asks when two readings are plausible, makes the change, and verifies what the system saved. Each operator's exact scope is confirmed before the first text.
In development
A coordinator across the tasks
One request split across specialists, tracked to completion, reported back as one status with the open questions separated from the routine follow-up. Same rules, same limits, one point of contact.
Design partners: we're taking three.
Fulcrum™ Operator is in pilot, not on a price list. We're looking for three operators to build the next ninety days with: home care agencies, assisted living communities, or small multi-site practices where the owner or administrator is doing the scheduling and payroll entry by hand. It runs in healthcare today; the job is the same in any business with shifts and timesheets. You bring a platform you're already on, one person who texts it real work, and twenty minutes a week telling us what it got wrong. ShiftCare and Gusto work today. Tell us what you run.
Become a design partnerHow it works
Answers start in week one. We produce them at first. Over time, Fulcrum™ takes them over.
We start as your embedded analyst. You ask, we answer, and we write down the definition behind every answer as we go. Underneath, we build Fulcrum™ on those definitions: the automated analyst that takes over the recurring work. Reports move over one at a time, each with its definition agreed, its sources attached, and a check in front of it. Then we go lighter.
Fulcrum™, your embedded analyst
Questions answered. Definitions written as we go.
Fulcrum™, the automated analyst
Morning report, alerts, follow-up questions. Sources on every one.
Blueprint
We answer your questions and write the definitions down.
Build
Working software every two weeks. Each report moves over once you trust it.
Operate
Fulcrum™ runs the recurring work. We keep it up and bring in what's new.
What allows you to trust your data?
Semantic layer
One written definition behind every number.
For each number: what it counts, what it includes and excludes, how it's calculated, how the data connects, and what it's for. Agreed with you, written down in one place, and updated as the business grows. Your team and Fulcrum™ work from the same definitions, so the board deck, the dashboard, and the answer to a follow-up question all come from the same source. Curated and documented answers you can trust.
Guardrails
Hard rules, written in code, not in a prompt.
Rules decide what Fulcrum™ may see, what it must cite, and what it may never do. Every answer shows its sources, so anyone on your team can trace a number back to where it came from. When the data can't support an answer, Fulcrum™ says so and flags the gap. Quality checks run before a number reaches you. Patient data gets its own rules: PHI stays inside infrastructure covered by a business associate agreement, access is scoped the way you'd scope a new hire's, and nothing patient-level moves without a BAA in place. We review the answers and keep the checks current.
Inside your systems
A digital worker, not a new system.
Fulcrum™ works inside the systems you already run: EMR, billing, CRM, scheduling, payroll. It logs in the way a new analyst would, with its own credentials and only the access you grant. Nothing gets replaced, migrated, or rebuilt. Your core systems stay exactly as they are. Fulcrum™ reads from them, reconciles across them, and reports on them.
Control
Delegation without giving up authority.
Every workflow runs the same sequence: understand the request, check its scope, assign the work, verify the result, report back. Anything outside the approved scope comes back to you as a decision. Permissions, spending limits, and approval requirements live in the application's controls, not in an agent's judgment about whether a rule matters. A blocked task says why. A proposed change is never reported as done.
Models
You shouldn't have to pick the model for every task.
The model router lives inside Fulcrum™'s chief-of-staff application. It picks among approved models by the job, measured results, and operating limits: a small model or a plain function for routine work, a stronger one for a hard investigation. Amazon Bedrock is the planned inference foundation; specific models, regions, and capabilities are approved per deployment. The product is the maintained workflow, not access to a model brand.
Yours from day one
Nothing lives in our heads.
Definitions, documentation, and the system belong to you. The Blueprint is written so another team can take it forward. Ongoing support stays your choice. Trust that depends on us isn't trust.
Where we work
Built in healthcare. Not bound to it.
Seventeen of our forty-plus combined years were in healthcare, and that's where Fulcrum™ runs today. The rest were somewhere else, and the method didn't change: agree on the definitions, wire in the systems of record, put the analyst on top. If your business runs on a handful of numbers that three systems disagree about, it fits.
Where it runs today
Healthcare
Home care, assisted living, infusion, multi-site practices, and the companies that serve them. Census, referrals, payer contracts, staffing. PHI planned for before anything connects.
How we handle patient dataSame discipline
Regulated and professional services
Any business whose data comes with rules attached: financial, legal, staffing, benefits. The covered plane and the access controls are built once and pointed at your rules instead of HIPAA.
Same problem
Owner-run, multi-site operators
The owner or a controller reconciles finance, CRM, payroll, and scheduling by hand every month. Fulcrum™ doesn't care what the sites sell. It cares that the numbers agree.
Healthcare data
Every other industry gets to experiment with AI. Healthcare has to get it right the first time.
A retailer can point a chatbot at its data and see what happens. You can't. Patient data sits in scheduling, billing, the EMR, and the spreadsheet somebody exported last Tuesday. Any AI you switch on inherits all of it, and every vendor in the chain becomes a business associate. That is the extra burden healthcare carries: the same AI opportunity as everyone else, plus a compliance problem nobody else has to solve before they start.
We plan for PHI before anything connects to it, so the compliance work is done before the first model runs, not discovered after. Fulcrum™ is built to support your HIPAA obligations. It doesn't make them disappear, and nobody's software does.
BAAs before anything runs
A business associate agreement with you, and with every model and infrastructure vendor that could see patient data. We know which AI platforms will sign one and which won't, and we only build on the ones that do.
A covered plane for PHI
Patient-level data stays inside infrastructure covered by a BAA. Models and tools that aren't covered never see it. Where the data may go is a hard rule written in code, not a policy in a binder.
Minimum necessary, de-identified where the work allows
Each task gets the least data it needs. Most analytics never needs a name, so where the work can run on de-identified data it does, using documented methods rather than stripping names and calling it done. The few tasks that truly need patient-level detail are named and scoped up front.
Scoped like a new hire, audited like one
Fulcrum™ gets its own credentials and only the access you grant. Every answer shows its sources, so what the AI read and what it said are reviewable after the fact. Access is limited by rule, and the trail is there when an auditor asks.
For Operator, the same rules reach the phone. It introduces itself as an AI. Only the people you name can give it work. Health or personal details about a client or caregiver stay out of the texts, and it's built to flag them when they show up. When something falls outside its job, a person at OneLever gets the text, not a queue.
We run the same setup in our own operations. Bring your compliance officer to the first call. They should be the hardest person in the room to satisfy, and this is built for them.
The engagement
Fixed scope first. Retainer second. Dependency never.
Start with the numbers you steer by: census, referrals, and payer contracts in healthcare; bookings, utilization, and margin by account anywhere else. The Blueprint is the assessment: where your data sits on the maturity ladder, the agreed definitions, a design for Fulcrum™, and a build plan with estimates. You get real analysis on your real numbers while we write it, and you decide at the end of each phase whether to keep going.
Phase 1
Blueprint
Four to six weeks. We embed as your analyst on your real numbers, answering questions from week one and writing down what each number means as we go. You finish with agreed definitions, a design for Fulcrum™, and a sequenced plan with estimates.
Fixed fee. Fixed scope. Useful on its own, whether or not you build.
Phase 2
Build
We build the roadmap in order: data foundation first, systems of record wired in, then Fulcrum™ on top. Working software every two weeks. Reports move to Fulcrum™ one at a time, as each earns trust, with sources attached and data gaps flagged.
Monthly retainer, scoped to the roadmap from your Blueprint. Less than one senior hire.
Phase 3
Operate
Fulcrum™ runs the recurring work. We keep it up, wire in new data sources as they arrive, and add approved capabilities as the business changes, with the quality checks and definitions kept current. Automation cuts the recurring work; it doesn't end the maintenance. Or your team runs it and we step back.
Reduced monthly operating fee as your business needs evolve.
What you hold at the end of the Blueprint
- A map of your data. Where it lives, how it connects, and what is missing across contracts, the referral network, CRM, financials, and operational KPIs.
- The first agreed definitions. The priority numbers, how each is calculated, and what each includes and excludes. One definition each, agreed by the people who use them.
- A design for Fulcrum™, your automated analyst. The first reports and questions it will answer, the data connections, the tool choices, the access controls, and the quality checks every answer has to pass before it reaches you.
- A sequenced roadmap with estimates. What to build first, the estimate for each step, and what it takes to run and maintain it.
What if we stop after the Blueprint?
You keep everything: the definitions, the documentation, the design, and the plan. The Blueprint is written so another team, or your own, can take it forward without us. The same holds at every phase. Definitions, documentation, and the system are yours from day one, and ongoing support evolves to support your objectives and needs.
What about patient data?
We expect it, and we treat it as the hardest constraint in the design, not a checkbox. Before Fulcrum™ touches any patient-level data or PHI, three things are in place: a business associate agreement with you, any required vendor BAAs, and the infrastructure and access controls to hold it. Inside Fulcrum™, PHI handling is a hard rule enforced in code: what it may read, where it may send it, and what it may never do. Clinical platforms and EMR interfaces are scoped separately.
Who stands behind the number
Built by operators, not a slide deck.
One of us builds the number. The other proves you can bet on it.
Between them, forty years of building data systems and living with what the numbers said.
Spencer Kline
Builds the machine that produces the number
Erie, Colorado
linkedin.com/in/spencerklineFifteen years in data and operations. Titles along the way: SVP of Analytics and Data, VP of Growth and Operations, founder-operator.
Spencer has built the whole stack for a healthcare operator: the warehouse, the BI layer, and the AI agents on top, for a 39-site infusion business. With Dan, he built the supply and demand system a national healthcare company ran as it scaled. He founded a 25-bed assisted living and runs it on an AI operating system he built, so he reads the morning number as the executive, not only as the person who built it.
On Fulcrum™ he owns the build: the data foundation, the systems of record wired into it, the agent layer on top, and the guardrails that decide what Fulcrum™ may answer, what it must cite, and when it flags a gap instead of guessing.
- 15 yrs
- In data and operations. Nine of them inside healthcare, producing the number and answering for it.
- 3 care settings
- In-home acute care, infusion, senior living. Built the data behind each and made operating decisions on it.
Dan Edstrom
Defines the number and proves it is right
Denver, Colorado
linkedin.com/in/danedstromTwenty-seven years in data. Titles along the way: CTO, SVP of Technology, VP of Data Science and Data Engineering.
With Spencer, he built the supply and demand system a national healthcare company ran as it scaled, and the analytics, data science, and data engineering organizations behind it. As CTO he ran product, engineering, data, design, IT, and InfoSec. Before healthcare, seventeen years delivering predictive models to university clients, and a first job in 1999 building quality control systems. Checking the number has been his job from the start.
On Fulcrum™ he owns what the number means: the definition of what each one counts and excludes, the KPIs built on those definitions, the statistics behind them, and the quality checks that catch a bad number before it reaches you.
- 27 yrs
- In data, from analyst to CTO
- 6 functions
- Product, engineering, data, design, IT, InfoSec. Ran all six as CTO of a healthcare company.
Together
- 40+ yrs
- Of analytical experience between them.
- 4 care settings
- In-home acute care, infusion, senior living, primary care. Between them, the data for each, built and answered for.
- 50+ markets
- Ran on the system they built together, every day
Bring one outcome you're tired of coordinating.
A recurring report. A scheduling task. An investigation that keeps bouncing between people. Bring that, or the short list of numbers you steer by. In thirty minutes we'll map what it depends on, what can be automated, and where your approval should stay. If it's a fit, the Blueprint starts there: fixed fee, standalone, yours from day one. If it isn't, you're out half an hour.
Or write to admin@onelever.ai. Same conversation, no calendar.