AI case orchestration
AI case orchestration for dental labs
Cympha's AI reads what happens on each case, a new file, a message or a change of state, and proposes the next step with its reason. The lab decides which kinds of step the AI may take on its own. Everything else waits for a person to accept it.
What case orchestration means
A dental case passes through many hands: the practice, the intaker, the designer, the doctor approving the design, the people making and shipping it, and now AI coworkers. Orchestration is deciding what happens next and who does it. In most labs that is done by people watching a list. In Cympha the case itself proposes its next step, and the lab sets how much of that it trusts.
How the next step is decided
- 1Something happens on the case: files are uploaded, the state changes, or someone writes a message.
- 2The AI reads the case, its files, prescription, messages and current state, and proposes what should happen next. Each suggestion comes with the reason for it and how confident the AI is.
- 3The lab's rules decide what happens to the suggestion. For the kinds of step the lab has allowed, it is applied at once when the confidence is high enough. Every other suggestion waits on the case for a person to accept or dismiss.
- 4Whether a person accepts it or the rules apply it, the step goes through the same case rules as a click in the app. A step the case cannot take from its current state is refused.
- 5The suggestion, the event that prompted it, and who or what applied it are recorded in the case log.
What the AI can propose
Change the case's state
Move a case on, for example from intake review to design once the files and prescription are complete, or back to the practice when something is missing.
Request an approval
Ask for the approval the case type calls for once the design is ready.
Hand the case to someone else
Pass the case from one person on the lab's team to another without changing its state.
Add or remove labels
Tag cases with the lab's own labels, so lists and filters show them to the right people.
Who does the work
Orchestration decides the next step. The work itself goes to whoever the case is routed to: people and AI coworkers draw from the same queue.
People
When a case changes state it is routed to the right role on the lab's team, keeping the same person where possible and otherwise the one with the fewest open cases.
AI coworkers
AlignerBot drafts aligner treatment plans in rounds and hands each plan back for a person to approve. CBCT segmentation separates anatomy from a scan, and the result comes back for a person to check.
Lab workstations
Through the Cympha Agent, jobs run on the lab's own computers and their results come back onto the case.
The practice
When the next step is the doctor's, an approval or missing information, the case goes back to the practice and they are notified.
Controls the lab keeps
Off until switched on
AI suggestions are generated only for labs that turn them on.
Automatic only where you allow it
Each kind of step has its own setting. With no setting, nothing is applied automatically.
Your confidence threshold
The lab sets how confident the AI must be before a step of that kind is applied without a person.
A daily limit
Each kind of automatic step has a daily cap per lab, so a run of mistakes stays small.
Clinical approval stays with people
Designs are approved by the doctor or the lab, never by the AI.
Every step on the record
Accepted, dismissed or applied automatically, each suggestion stays in the case log with its reason.
Orchestration on one record
This only works because the whole case is in one place. The AI reads the same record the people do, and its suggestions change that record through the same rules. This is dental case lifecycle management (DCLM) with AI in it. The paper on Hybrid Dental Case Execution describes the approach in full.
Questions about AI case orchestration
- Does the AI change cases on its own?
- Only for the kinds of step a lab has allowed, above the lab's confidence threshold and within its daily limit. Everything else is a suggestion a person accepts or dismisses.
- Is it switched on by default?
- No. A lab turns AI suggestions on in its settings, and chooses separately which kinds of step may be applied automatically.
- Can we see why the AI suggested something?
- Yes. Each suggestion shows its reason and the event that prompted it, and both stay in the case log.
- Can a suggestion break the case's workflow?
- No. A suggestion goes through the same rules as a person's click, so a step the case cannot take from its current state is refused.
- Who approves the design?
- A person: the doctor, or the lab where the case type calls for the lab's own approval.
See a case orchestrated
The quickest way to judge it is to watch a case move from intake to approval with AI suggestions switched on.