Ohio’s consent-based smart-documentation design places AI where it can reduce burden without confusing assistance with authority.
A crisis specialist should not have to choose between being present with a person and documenting the interaction.
Yet every additional field, duplicate entry and after-call task competes for attention. During the contact, the specialist is listening for words, pauses, changes in affect, immediate risks, protective factors and the practical next step. Afterward, the record must support continuity, reporting and accountability.
We are using modern language technology to reduce that burden.
We are giving that automation clear boundaries from the start.
As part of Ohio’s State Centralized 988 Platform, RingMD is developing a consent-based smart-documentation workflow designed to transcribe an interaction and organize relevant information into a structured draft. The crisis specialist reviews, corrects and approves the material before it becomes the final record.
Our design principle is simple:
The system can draft. The specialist decides.
In our workflow, a generated draft becomes accountable documentation only after the crisis specialist has reviewed and approved it. That approval carries professional judgment, not just a click.
Documentation Is Part of Crisis Care
Documentation is sometimes described as an administrative task that follows care.
In a crisis system, it is part of care.
A good record helps the next professional understand what happened, what the person communicated, which risks were considered, which resources were offered and what follow-up may be needed. It supports handoffs among 988 centers, mobile crisis, MRSS, emergency services and community providers. Aggregated appropriately, it also helps leaders understand service patterns and resource needs.
But documentation can become counterproductive when the workflow demands the same information repeatedly or forces the specialist to divide attention during the conversation.
Ohio’s nineteen 988 centers have developed local practices within a broader national network. RingMD’s discovery work with those centers and other stakeholders identified fragmentation across telephony, documentation, referral and reporting environments as an operating challenge.
We are developing smart documentation to solve that workflow problem: preserve the value of the record while reducing unnecessary friction around its creation.
A Transcript Is the Starting Point
An accurate transcript gives the specialist a useful foundation.
Clinical meaning comes from the person who understands the conversation.
A transcript captures the words, while the crisis specialist applies context, identifies what matters and organizes the information for continuity of care.
Language systems can summarize, classify and propose structure. They can also omit, flatten or misinterpret. A fluent draft may make an error harder to notice because it sounds complete.
Our design keeps the transcript and generated structure in an assistive role. The specialist can compare the draft with the interaction, change the language, add missing context and reject material that is inaccurate or inappropriate.
The product gives the crisis professional a useful starting point without asking the machine to imitate their expertise.
Consent Comes Before Capture
Voice data can contain some of the most sensitive information a person will ever share.
In a 988 interaction, that may include suicidal thoughts, substance use, trauma, health conditions, relationships, location and the identities of other people. The fact that transcription can make documentation easier does not make capture automatic or ethically neutral.
Ohio’s smart-documentation approach is being designed as consent-based. That means the workflow must communicate what is happening and give the specialist a clear path when consent is not provided or when transcription is not appropriate.
Consent must also be more than a sentence hidden inside general terms. The exact implementation should reflect Ohio’s approved policies, applicable law, modality and operating context. The system should make the state of capture visible to the specialist and prevent accidental use outside the authorized workflow.
The HHS HIPAA Security Rule requires safeguards for electronic protected health information maintained by regulated entities and their business associates. HHS guidance also recognizes transcription vendors as business associates when they create, receive, maintain or transmit protected health information on behalf of a covered entity.
RingMD is designing privacy protections across the full information lifecycle—from capture and processing through review, storage, access and deletion.
The Draft Must Remain Visibly a Draft
Automation bias begins when people give extra weight to a machine-produced answer simply because the system produced it.
That is why we are designing the interface to keep generated content unmistakably in a supporting role.
If generated text appears in the final record with no clear status, a busy specialist may reasonably assume it has already been validated. If the workflow requires active review, highlights uncertainty and makes editing easy, the interface reinforces professional responsibility.
Our interface is being designed to distinguish clearly between:
- source transcript versus generated summary;
- required fields versus suggested content;
- reviewed content versus unreviewed content;
- human edits versus machine-generated text where auditability requires it; and
- saved draft versus approved final record.
The specialist’s approval finalizes the note and confirms that a qualified professional has reviewed and stands behind it.
This principle also reflects the direction of federal health-technology policy. The Office of the National Coordinator’s HTI-1 final rule established transparency requirements for predictive algorithms in certified health IT. Ohio’s crisis workflow follows its own approved policies, but the practical principle is the same: when technology influences healthcare work, users should be able to see what it did and where its limits begin.
Structure Should Follow the Encounter
Traditional forms often make the user adapt the conversation to the database.
Ohio’s adaptive-documentation concept is intended to move in the other direction. Fields and prompts can respond to what has already occurred, reducing redundant entry and focusing attention on information that remains necessary.
That can make the workflow more efficient, but only if structure remains clinically and operationally appropriate.
The system must preserve uncertainty: silence remains silence; ambiguous statements remain unresolved; risk fields require confirmed information; and data collection stays within the approved workflow.
We focus the automation on practical work:
- pre-populate information already established;
- organize content under the correct headings;
- identify incomplete required elements;
- reduce duplicate typing; and
- present a draft for professional review.
Those tasks keep automation assistive and leave clinical judgment with the specialist.
Human-in-the-Loop Must Describe a Workflow
For RingMD, ‘human in the loop’ is not a slogan.
It is a defined workflow with named people, clear decisions and recorded accountability.
Our design answers concrete questions: Who reviews the output? What information can they see? Are they trained to detect errors? Can they change or reject the draft? Does the system record approval? What happens when the model is unavailable? How are recurring errors reported and corrected? Who is accountable for changes to the workflow?
For Ohio’s smart documentation, the human is the crisis specialist responsible for the final record. The loop includes consent, draft generation, comparison, correction, completion and approval. Supervisors and authorized administrators require their own governance views. Technical and clinical teams need a controlled process for evaluating defects and updating the system.
The National Institute of Standards and Technology’s AI Risk Management Framework emphasizes governing, mapping, measuring and managing AI risk across the lifecycle. We apply that discipline by carrying feedback from human review into ongoing testing and improvement, including patterns of omissions, false structure or workflow confusion.
That process protects each record and helps RingMD strengthen the system over time.
Measure Burden Without Trading Away Quality
For RingMD, success is measured by the time and attention we can return to specialists while keeping documentation quality and human judgment intact.
The result we are working toward is simple: specialists can devote more attention to people while producing records that remain accurate, complete, timely and useful.
That requires measurement on both sides of the equation.
Operational measures may include time spent documenting, after-contact work, duplicate entry, completion rates and specialist experience. Quality measures may include correction rates, omitted information, inappropriate additions, supervisor review findings and performance across different accents, speech patterns, languages and contact types.
Saving three minutes only helps if documentation quality holds. A polished draft that forces the specialist to reconstruct the conversation during review merely moves the burden.
Evaluation should also examine whether performance differs among people or settings. Crisis conversations do not follow standardized speech. Technology that works well for one communication style may work poorly for another.
That is how we approach governed AI: we measure the automation against the work it is supposed to improve.
Development Is the Time to Set the Boundary
Ohio’s State Centralized 988 Platform remains under development, with the smart-documentation workflow now being configured, tested and validated ahead of statewide use.
We are using this development stage to make the boundary between assistance and authority explicit in the design itself.
RingMD and Ohio’s stakeholders can test how consent is obtained, how drafts are displayed, which fields can be suggested, how review is recorded and where automation should abstain. User-acceptance testing can reveal whether the workflow genuinely reduces burden or merely changes its form.
Once a design is operating at scale, ambiguous responsibility becomes much harder to correct. Setting the principle early keeps the technology aligned with the mission.
As RingMD’s Chief Operating Officer and a leader in our government-health work, Varun Arora has made that boundary a core principle across our AI programs. Under his leadership, we use automation to help clinicians, crisis specialists and program teams navigate, document and work more effectively while keeping consequential judgment with the people accountable for it.
Ohio’s documentation design turns that philosophy into a concrete workflow.
This is the kind of AI RingMD is building: technology that gives the human author back attention, time and structure. The system can listen, organize and draft; the specialist interprets, corrects and decides.