It Took Your Treatment Coordinator 45 Minutes to Build the Alvarez Case Packet. Now You Assign the Outcome Instead.

You already tried this. Sometime in 2024 your office manager pasted a treatment plan into a chat box and asked it to write up the case for Mrs. Alvarez, and it produced four paragraphs of warm, empty language about the importance of oral health, with a made-up insurance estimate in the middle. You closed the tab and went back to doing it by hand. That was the correct call at the time.

Here is what is different now. In January 2026 Anthropic shipped Claude Cowork, and on 9 July 2026 OpenAI shipped ChatGPT Work on GPT-5.6. Both do the same fundamentally new thing: you give a goal rather than a prompt, they connect to your files and your applications, they break the job into steps themselves, they work for an hour or three on their own, and they hand back a finished thing — a completed spreadsheet, a priced quote packet, a filled set of forms. Both were built for people who have never written a line of code, which in a dental practice means the treatment coordinator and the office manager, not a consultant you hire.

The job in question: the packet behind a $7,400 treatment plan

Take a real one. The doctor diagnoses generalised moderate periodontitis with two failing crowns and a molar that needs an implant. Four quadrants of scaling and root planing, two crowns with buildups, one extraction and implant referral, periodontal maintenance every three months afterwards. Call it $7,400.

What has to exist before that patient will say yes:

• A phased plan showing what happens in what order and roughly when.
• An insurance estimate per phase that reflects this plan — remaining annual maximum, deductible, frequency limits, the alternate-benefit downgrade on the molar crown, the waiting period on the implant.
• A sequencing decision worth real money: whether to split the case across two benefit years so the patient gets two annual maximums instead of one.
• A predetermination request for the periodontal work with a narrative and the periodontal charting and radiographs attached.
• A financial page with the patient portion per phase and the third-party financing options, with the actual monthly figures rather than "we have financing available."
• A signed financial arrangement form.

Your treatment coordinator does all of that by hand. It takes her 35 to 50 minutes per case if nothing goes wrong. She has maybe four of those windows in a week, and the case was diagnosed on Tuesday at 10:40 while she was on the phone. So the patient goes home with a printed treatment plan and a promise that "we'll call you with the numbers," and the practice's case acceptance quietly depends on whether that call happens by Thursday.

What changed: from a task list to a stated goal

Handing over work as a goal means describing the finished result and the rules it must satisfy, then letting the system decide the steps — instead of writing out the steps and checking each one yourself.

That sentence sounds like a small distinction and it is not. The old way required someone to know all fourteen steps and to be present for each. The new way requires someone to know what "done" looks like. Those are completely different jobs, and the second one is the one your treatment coordinator is already good at — she knows exactly what a good packet looks like, because she has built three hundred of them.

flowchart TD
  A["Goal: Alvarez packet ready before she leaves the chair"] --> B["Read the treatment plan and chart notes"]
  A --> C["Pull remaining benefit, deductible and frequency limits"]
  A --> D["Price each phase against the office fee schedule"]
  A --> E["Draft the predetermination narrative from the perio charting"]
  B --> F["Assemble the phased packet and financial page"]
  C --> F
  D --> F
  E --> F
  F --> G{"Treatment coordinator reviews — numbers and wording"}
  G -->|Fix the phasing| A
  G -->|Approved| H["Printed at the front desk, doctor signs the narrative"]

The goal, written the way you would say it out loud

This is what the instruction actually looks like in a dental office. Not technical, just precise:

"Build the case packet for the Alvarez plan the doctor entered this morning. Phase it so the periodontal therapy goes first. Use her plan's remaining maximum and the frequency limits from the verification sheet. Show me both options — all of it this year, or split across December and January so she gets two maximums — with the patient portion for each. Draft the predetermination narrative from the periodontal charting for the doctor to sign. Put the financing options on the last page with the real monthly figures at 24 and 60 months. Do not invent any coverage detail; if something is not on the verification sheet, flag it instead of guessing."

That last sentence is the one that separates a useful packet from the 2024 mush. State it every time.

An hour later the packet exists. Two options priced. The narrative drafted from the actual pocket depths in the chart. Three flagged unknowns at the top: whether the implant waiting period is satisfied, whether the buildup is separately billable, and whether her employer's plan year renews in January or July. Those are exactly the questions you want a human calling the payer about.

What the owner has to change about how work gets assigned

This is the part most practices skip and then wonder why nothing improved. Three changes, all of them managerial rather than technical.

Stop assigning steps. Start assigning outcomes with acceptance criteria. "Build the Alvarez packet, two sequencing options, nothing invented, on my desk before her hygiene appointment Thursday" is an assignment. "Verify her benefits, then price it, then call CareCredit" is a task list, and a task list is now the slow way.

Write down what "good" looks like, once. Your practice already has a house style for a case packet — the order of the pages, whether you show the full fee and the write-off or only the patient portion, how you phrase periodontal findings to a patient who is frightened. Nobody has ever written it down; it lives in the treatment coordinator's head. Write it down. It becomes the standard the finished work is checked against, and it is also the best onboarding document you will ever have when she goes on maternity leave.

Move review earlier and make it a real job. The coordinator's morning now starts with reviewing four finished packets instead of building one. That is a different rhythm and it deserves a different slot in the day — 20 minutes before the 8 a.m. patient, with the doctor available for narrative sign-offs, rather than squeezed between phone calls at 2 p.m.

Case acceptance, run through the arithmetic

Assumptions, illustrative: 14 comprehensive treatment plans presented a month, average plan value $4,900. Same-day presentation with numbers in hand accepts at a materially better rate than "we'll call you Thursday" — use whatever gap your own practice sees; here we will assume 52% versus 41%, a difference of 11 points. Margin on incremental production above fixed cost, 55%.

Prove it the boring way rather than believing the table: track two numbers for one quarter — cases presented with the full packet in hand, and cases presented with a promise to follow up — and the acceptance rate of each. If the gap is not there in your practice, you have learned something more valuable than the $50,000.

Where this still needs a person, permanently

The packet is paper. The conversation is not. Nothing produced overnight will notice that Mrs. Alvarez went quiet when the implant came up, or that the real objection is not the $2,100 but whether she can take three days off work. That is why treatment coordinators exist and why practices that try to replace them with a printout see acceptance fall, not rise.

Two hard limits on top of that. First, coverage detail must never be guessed at — it comes off a verification sheet or it comes with a flag on it, because a wrong estimate handed to a patient in writing is a promise you will end up eating. Second, the clinical narrative supporting periodontal therapy is a statement made in the doctor's name to a payer, and it gets read and signed by the doctor every time. Drafting is fine; sending unread is not.

Be deliberate about where patient information sits, too. These products connect to your files, and your files hold protected health information. Use the business tier, get a business associate agreement signed, and scope the connection to the folders it actually needs.

Frequently asked questions

Which one should a five-operatory practice pick?

Either will do this job. Claude Cowork has had six more months in the field; ChatGPT Work arrived in July. Pick on which one connects cleanly to where your files actually live, and run a two-week trial on one recurring job before committing the whole office.

Can it pull directly from Dentrix or Eaglesoft?

Treat that as a later step. Start by dropping the exported treatment plan, the verification sheet and your fee schedule into a folder it can read. Practices that wait for a perfect connection to their practice management software tend to wait a year and get nothing; practices that start with exports get value in a fortnight and figure out the connection later.

Does this replace the treatment coordinator?

No, and if that is the plan, do not start. It removes the 40 minutes of assembly per case, which is the part of her job she likes least and which stops her being on the floor. The practices getting the most out of this are the ones where the coordinator now presents six cases a week instead of three.

What is the one recurring job to hand over first?

The packet — because it repeats, it has a clear finished form, its quality is easy to judge at a glance, and it sits directly on top of revenue. Do not start with something judgment-heavy like fee negotiations. Start with the thing you would recognise as done or not done in ten seconds.

One thing a finished packet cannot do is chase the patient who said she wanted to think about it and then did not call back. CallSphere builds AI voice and chat agents that answer the practice line and web chat around the clock, book and reschedule appointments and capture enquiries — which is usually the difference between a well-built case packet and a case that actually gets started.