Business process automation consulting
When your systems stop keeping up with your growth.
We go into your business and find where manual work, disconnected systems and unreliable reporting cost you time and money.
Then we connect the information, automate the repeated work and help you see what is working, tested against the work your team already trusts. Best fit: owner-led businesses with several locations or business units.
Our flagship work is with a five-center behavioral health network. Results in other industries are not yet published.
BeforeDays
AfterEach morning
Growth adds the same work again.
It usually starts small: one export, one spreadsheet, one person who knows how it fits together. Each new location or team adds another copy of that work, until the business runs on people carrying information by hand.
- Numbers carried by hand
- Work finished in one system is retyped or exported into another before anyone can bill, pay, report or decide.
- Reports nobody fully trusts
- Leadership numbers arrive late, disagree between teams, or quietly include stale inputs.
- One person holding it together
- A critical weekly process lives in someone's head. When they are out, it waits.
More on this: seven workflows that break at the third location.
What we build, and where it sits in the work.
Four kinds of work, scoped to what the audit finds. Each one covers part of the same flow: information comes in, gets checked, a person decides, the team acts, and the result is measured.
Sometimes the fix is AI. Sometimes it is a better process, a missing field, or two systems finally talking to each other. At our flagship client, the most important fix was a column nobody had added: a record number on the attendance log. When to automate a process, and when to hire.
A repeatable method. The audit decides what is worth building.
- AuditInterview the people doing the work, map how it really runs, and find where time and money slip.
- PrioritizeRank each finding by impact, evidence, effort and dependencies.
- Build and verifyImplement a focused fix and test it against the work your team already trusts.
- Run and improveMonitor the agreed processes, handle failures, and keep the rules documented.
An audit finds plenty. Prioritizing decides what is worth building.
Every finding is placed by its impact, the effort to fix it, how strong the evidence is, and what it depends on. The first build has the strongest case and nothing it is waiting on. Some findings are left alone on purpose.
- 1A record number missing from a logCheap to fix, and the fix that makes matching possible
- 2Numbers retyped every weekThe most hours lost; depends on 1
- 3A leadership report rebuilt by handWorth it, once the data underneath is right
- 4Two teams keeping the same listQuick, but confirm it costs anything first
- 5A one-off custom dashboardAsked for once; probably not worth building
What an audit delivers.
Usually two to three weeks on your real files, not a demo. Quoted flat, and credited toward any build that follows.
| Finding | Evidence | Business effect | Recommended action | Owner |
|---|---|---|---|---|
| Weekly report assembled by hand | Workflow observation and a time log | Repeated preparation every week | Connect the sources and validate the output | Operations lead |
| Customer records do not match | A sample reconciliation | Exceptions need investigating by hand | Agree identifiers and matching rules | Systems owner |
| Dashboard contains stale inputs | Source timestamps | Decisions may use outdated figures | Add freshness checks and failure reporting | Finance lead |
What you receive
- The findingsEach with its evidence and business effect
- A build scopePrioritized, with what it depends on
- The assumptionsWhat the scope rests on, written down
- An operating procedureYour team can run it without us
- Named ownersOne for every recommended action
If nothing is worth building, the audit says so. The full automation audit checklist.
What it looked like at a five-center network.
One operations lead carried attendance from four EMR instances into the billing census by hand, more than a day a week, and leadership reports waited on that work. Here is what happened, in order.
- AuditWith the operations lead and every center's files
- Root causeAttendance logs had no record number
- Census engineDrafted and checked five ways
- Parallel weekRun beside the manual census, signed off
- EMR by APIEvery center read each morning, read-only
- Daily boardsCensus, executive and clinical views
- NextThe fourth site and a new center
Try the weekly check, on made-up data
When attendance and the billing census disagree, the check flags it, a person decides, and the report updates. Reviewed corrections become documented rules where they apply.
1 · Two records, same week (minutes)
| Record | Source | Tue | Wed | Thu |
|---|---|---|---|---|
| DEMO-1042 | Attendance | 180 | 180 | 180 |
| Census | none | 180 | 180 | |
| DEMO-1078 | Attendance | 180 | 60 | 180 |
| Census | 180 | 180 | 180 |
2 · Flagged · 3 · A person decides
DEMO-1042 · Tue Attended 180 min. No census entry.
DEMO-1078 · Wed Attended 60 min. Census says 180.
Operational workflows are running. Financial outcomes have not yet been established. Full case study and rollout status
What leadership sees each morning.
The four views built for the network, recreated with invented numbers. Each answers one question. How to automate a weekly report · Dashboard design rules.
Running All four run today at a five-center network. Daily census and clinical outcomes rebuild every morning from Kipu. The numbers below are invented.
Shown with invented numbers. Nothing here is a real client, and no patient data appears anywhere on this site.
Where the numbers come from
At a five-center network, the implemented Kipu census connection removes the manual export step for that feed. We only read. Nothing is written back to your EMR.
Kipu's API is switched on per account. Where it is not, the same engine works from exports.
Different businesses. Familiar operational problems.
What we found in behavioral health is a common shape in growing operations. The details differ, and the audit checks which ones apply to you.
Results outside behavioral health are not yet published. Bring yours to a call and we will describe it in your own terms.
Keeping it working, and who is accountable.
Businesses change, sources change, and systems fail. Support is agreed at handoff, in writing, around these questions.
- Source unavailableA scheduled pull fails or arrives empty.
- Report marked staleThe report shows when its data was last read.
- Owner notifiedA named person on each side hears about it.
- Last good output keptNobody works from a half-built report.
- Recovery checkedThe next run is verified before it is trusted.
- What is monitored
- The processes named at handoff, each with an owner. Not everything, and never a promise of round-the-clock watching.
- When a source changes
- Keeping an agreed process working is support. A new workflow, source or report is scoped as new work.
- Who approves a rule change
- Your named owner. Rules are documented, so the logic does not live only in our heads.
- What you keep
- The documented rules, the operating procedure, and your own accounts and data. Where code runs and who holds it is agreed in writing for each engagement.
Running today at the network
- A scheduled morning run reads every center's census and rebuilds the daily boards.
- The manual export still works as a fallback on any day the API does not.
- Corrections are reviewed weekly and logged as documented rules where they apply.
Engineering-led, and accountable
Sunny Binjola spent ten years building production software. He leads the investigation and the technical build, and is on every client call from the first review to launch. Founder-led should not mean founder-dependent: the rules are documented, your team gets an operating procedure, and support responsibilities are named at handoff. More about Sunny
What owners ask first.
What kinds of businesses is this for?
Growing businesses whose people, systems and reporting have stopped keeping up, especially owner-led companies with several locations or business units. Our flagship work is in behavioral health; results in other industries are not yet published.
Is the answer always AI?
No. Sometimes the fix is AI. Often it is an integration, a matching rule, a missing field or a better process. The audit decides, and it says so if nothing is worth building.
Do we need to replace our existing systems?
No. The work runs around the software you already use, through APIs where they exist and exports where they do not. We read rather than write wherever we can, and approval requirements are agreed for each workflow.
What happens after launch?
Support is agreed in writing at handoff: which processes are monitored, who is notified when something fails, who approves rule changes, and what counts as new work. There is no open-ended or round-the-clock promise.
Where does sensitive data go?
Where sensitive data is processed, who can access it and what leaves your systems are agreed in writing before any of it is touched. In healthcare, a BAA is signed first, plus a qualified service organization agreement where 42 CFR Part 2 applies. Nothing goes to an ad platform, and no outside AI service works on records without a written agreement with you.
What if Sunny is unavailable?
The rules are documented, the audit leaves an operating procedure your team can run, and support responsibilities are named for each engagement. Where the code runs and who holds it is agreed in writing.
What does your team keep chasing, checking or rebuilding?
Bring one problem to a free 20-minute call. We will tell you what we would look at first, or that it is not worth building. No data needed.