Most AI documentation tools ship with a short dropdown. SOAP, sometimes BIRP, sometimes DAP, and that is the menu. If your agency uses something else, you reformat by hand after every note. If your county built its own template ten years ago, you reformat by hand. If your supervisor wants a six-section progress note with two sections that do not exist in any standard case note format, you reformat by hand.
That is most of human services. Workers do not get to pick from a tidy list. They inherit whatever their organization, funder, or state agency decided was correct, and the format is rarely on a vendor's roadmap.
Thirteen native case note formats, out of the box
TryCaSIE ships thirteen case note formats natively. SOAP, BIRP, GIRP, SIRP, DAP, DAOP, BPS (BioPsychoSocial), PCAP, FIRP, FIRPP, and a plain standard narrative cover the eleven generic formats used across social work, case management, behavioral health, and clinical settings. On top of that, two specialized variants are already in the app: a Virginia-compliant BPS for clinicians billing Virginia Medicaid, and a ClinicalTracker-aligned BSP for organizations using that EHR. More state-specific and organization-specific formats are on the way.
If you work in a setting where one of those thirteen is the right answer, pick it and move on. The model is tuned for each one.
Custom case note formats for the work that doesn't fit
When none of the standard formats fit, build your own. A custom case note format in TryCaSIE has up to eight sections. Each section has a name and a short description that tells the AI what content belongs there. Save it, name it, use it for the next note. You can keep up to ten custom formats per account.
A residential program supervisor can build the five-section shift summary their state licensing board reviews. A community mental health director can build the program-specific progress note that combines BPS elements with internal outcome tracking. A re-entry program coordinator can build a four-section weekly check-in that maps directly to the fields in a grant reporting system.
Why the wrong format costs workers real time
A case note that uses the wrong format is a note that fails review, fails audit, or sits in a draft folder while the worker rewrites it on a Saturday. A worker who has to manually re-section every AI output ends up doing the work twice and saving no time at all. The whole point of an AI documentation tool is to deliver a finished note that pastes cleanly into the system of record. If the format is wrong, none of the rest matters.
Format also has to survive review by people who were not in the room. A funder reviewing a progress note expects specific section headers in a specific order. An EHR with a structured intake form expects those same headers so a paste lands in the right fields. A supervisor signing off on residential staff documentation expects the language pattern they trained the team on. The format is not decoration. It is the part the reviewer reads first.
Three workers, three custom formats in real use
A residential program in Virginia uses a five-section format their state licensing board reviews: presenting concerns, interventions used, resident response, safety check, and shift handoff. The supervisor built that format in TryCaSIE once, set it as her team's default, and now the staff record their shift summaries by voice and paste a finalized note into the resident chart at the end of each shift.
A community mental health agency uses an eight-section progress note that combines BPS elements with their own outcome tracking fields. They built it in TryCaSIE, named it after the program, and the case managers no longer keep a Word template open in a second window for reference.
A re-entry program documents weekly check-ins with a custom four-section format that maps directly to the fields in their grant reporting system. When the program coordinator finalizes a note and pastes it into the reporting tool, every section lands where it belongs.
How a custom format guides the AI
A custom case note format is not just a header skeleton. Each section has a description, and the description is what the model uses to decide what content goes where. A section called "Functional Status" with a description like "current daily living capacity, mobility, and self-care observations" produces a different paragraph than a section called "Functional Status" with no description at all. Worth spending a minute on each description when you create the format. The model is matching content to your intent.
The format definition you used at the time of generation is snapshotted into the note's metadata. If you later edit the format or delete it, every old note still displays correctly. The history is preserved.
Custom formats are not custom AI prompts
A custom format is not a custom AI prompt. The transformation prompt, the system that instructs the model on tone, voice, audience, and grounding, is something TryCaSIE writes for you and refines based on the editing level you choose. What you provide is the structure of the note. What we provide is the language model behavior that fills it in.
If you want both, the app supports custom transformation prompts as well, separate from custom formats. You can save up to ten of those, set one as your default, and combine it with any format. That combination is what most senior workers eventually settle into. Your structure, your voice instructions, our model behavior.
The tool bends to the documentation
A documentation tool that only knows three case note formats is a tool for a small slice of the field. Human services runs on agency-specific, county-specific, state-specific, and program-specific paperwork, and most of it was designed long before anyone was thinking about AI. TryCaSIE was built around that reality from the start.
If your agency has a format that is not on our standard list of thirteen, build it once and it is there. If your state changes its requirements next year, edit the description and your team is current. The tool bends to the documentation, not the other way around. See TryCaSIE plans.