Home · Solutions · Management & planning
Solution · Management & planningThe owner asks in Teams how much the clinic made on toxin after product cost. A minute later she has figures, period and source, instead of waiting a week
The clinic’s finance agent in Teams
An AI agent in Microsoft Teams has read-only access to the clinic system, the books in the accounting system such as Comarch Optima or iFirma, bank statements and the product register. It understands questions asked in plain words, answers with figures stating period and source, builds reports on request and every morning sends one message with the key numbers. It books nothing, pays nothing and sees no medical data.
Executive summary
A clinic’s numbers sit in several places: revenue per treatment and doctor in the clinic system, costs in the books, money in the bank, product cost in invoices; answering a simple question means collecting them all.
An AI agent in Teams reaches those sources read-only, answers questions asked in plain words with figures, period and source, builds reports on request, sends a morning report and watches the margin per treatment.
Decisions on prices, purchasing and schedules made on yesterday’s numbers, not last month’s; fewer hours over spreadsheets and fewer out-of-scope questions to the accounting office.
the clinic system via export or API; the accounting system via API or exports; bank statements; the batch and payout registers; conversation and reports in Microsoft Teams
Business problem
The clinic produces numbers every day, and the owner sees them once a month, in someone else’s summary
The owner of an aesthetic clinic makes decisions every week that depend on numbers: whether to raise a treatment price after a product price rise, whether to give more hours to a doctor with a full schedule, whether to order a larger batch of toxin with an early-payment discount. The numbers needed for this exist. Just not in one place.
Revenue per treatment and doctor is in the clinic system. Costs and invoices are in the books, kept in an accounting system or by an accounting office. Money is in the bank. Product cost is in wholesaler invoices. A question about toxin margin needs all four, so it either takes an evening over exports or goes by email to the accountant and comes back a week later.
The result is predictable. Decisions are made by feel or on last month’s numbers. A product price rise eats into the margin for a quarter before anyone notices. The report that was meant to be weekly happens when there is time, which is rarely.
A finance agent does not replace the accountant or make decisions. It does one thing: it lets you ask about numbers the way you ask a person, and get an answer from the data within a minute. Stating where it comes from and which period, so the owner can trust it or check it.
How it works today
Below is what the work looks like before anything is automated.
- PersonThe owner has a question about numbers on Wednesday evening
- Risk of errorData in the clinic system, the books and the bank, separately
- WaitingThe question goes by email to the accountant
- Risk of errorThe accountant builds an answer from exports when she has time
- WaitingThe answer arrives a week later; the decision waits
- Risk of errorThe next question starts the same cycle
Why the current process costs more than it appears
The bill that never shows up in a budget.
- How long does it take you today to answer a question about margin on one treatment type last month?
- When did you last raise a treatment price, and was it before or after a product price rise ate the margin?
- How many questions about numbers do you send to the accounting office each month, and how many fall outside the contract?
- Do you know this morning what came in yesterday, what is in the account and which invoices must be paid this week?
Cost of inaction
The model assumes about 30 questions about numbers and reports a month, 16 minutes to assemble an answer, a fully loaded owner hour of 45 € and an out-of-contract accounting office hour of about 50 €. You will know your own number by noting every question about numbers for a month and the time until the answer.
The third line has no amount but is often the largest. A product price rise noticed a quarter too late is a quarter of treatments sold at a lower margin.
A model organisation with realistic proportions – the numbers exist so you can run the same maths on your own data; they are not a client result.
An aesthetic clinic: 3 treatment rooms, 5 doctors and cosmetologists, about 600 treatment visits a month, books kept in Comarch Optima by an accounting office, one business account, Microsoft 365.
Every day at 7:30 the owner gets one message in Teams: yesterday’s visits and takings, today’s schedule with free slots, the account balance, invoices due this week and products whose stock will last less than a week.
On Wednesday evening she writes: “How much did we make in September on toxin and on fillers after product cost?”. A minute later the agent replies: toxin 20 400 € revenue, 5 000 € product, 15 400 € after product, which is 75%; fillers 26 500 €, 9 100 €, 17 400 €, which is 66%. Source: the clinic system and the batch register, as of 7:00.
She follows up: “And how did August compare, and how many packages are still open?”. The agent shows the month comparison and 38 packages with 112 treatments still to deliver, worth 9 800 € together. Finally she asks for the same report on the first Monday of every month; the agent sets it up.
A week later the agent itself reports that after a new wholesaler invoice the margin on one filler has fallen below the threshold the owner set. She decides on a price change; the agent only shows the numbers.
After a quarter, out-of-contract questions to the office are down by three quarters and reports produce themselves. The modelled value is about 5 600 € a year, plus decisions made on time. These are model figures, not the clinic’s records.
Proposed solution
The agent works in Microsoft Teams, where the owner already works. It has read-only access to the clinic system, the books in Comarch Optima or iFirma through interfaces or exports, bank statements and the registers the clinic already keeps: product batches, stock and doctor payouts. It turns a question asked in plain words into queries to those sources and builds an answer with figures, period and source.
Beyond conversation, the agent works in three ways. At a set time every day it sends the morning report. At the owner’s request it sets up recurring reports, such as a monthly review of margin per treatment. It also watches thresholds: when a treatment’s margin falls after a product price rise, when the account nears an agreed minimum or when vouchers are about to expire, it says so itself. It records every answer and every report, so it is always possible to check what data they were built on.
The limits are clear. The agent only reads and reports: it books nothing, pays nothing and changes no prices. It sees no medical data or patient names, working with treatments, amounts and visit numbers. When data is missing or incomplete, it says so plainly rather than guessing. Tax matters, such as how VAT rates differ between treatments, it shows as the office booked them and refers to the accountant rather than interpreting them.
UiPath Agents as a conversational agent in Microsoft Teams, with a tool for each source; UiPath Orchestrator for the report schedule and a trail of every answer; UiPath AI Trust Layer to control data sent to models
Plain-language conversation about numbers, answers with period and source, a morning report, recurring reports on request, alerts on margin, account balance and expiring vouchers
Comarch Optima or iFirma via API or exports; the clinic system via export or API; a CSV or MT940 bank statement or the bank’s API; clinic registers in SharePoint
How the automated process works
- PersonThe owner types a question in Teams, in plain words
- AutomationThe agent reaches the clinic system, the books, the bank and the registers
- AutomationAn answer with figures, period and source within a minute
- SystemA report on request, or every day at 7:30
- AutomationAn alert when a treatment’s margin falls below the threshold
- PersonThe owner decides on prices and purchasing; tax with the accountant
Human-in-the-loop model
Automation handles
- Answers to questions about numbers, stating period and source
- The morning report and recurring reports set up in conversation
- Alerts on margin, account balance and expiring vouchers
People decide
- Decisions on prices, purchasing, schedules and pay: the owner
- Bookkeeping, tax and interpreting regulations: the accountant
- Alert thresholds and who has access to the agent: set by the owner
Before and after
Systems and integrations
The stack is short on purpose: one engine, one execution layer, one place where a person decides.
Inputs
- visits, treatments and revenue per doctor from the clinic system
- invoices, costs and balances from Comarch Optima or iFirma
- bank statements
- clinic registers: product batches, stock, doctor payouts, packages and vouchers
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Action Center
Target systems
- answers in Teams with figures, period and source
- the morning report and recurring reports
- alerts on margin, account balance and expiring vouchers
Human touchpoints: the owner asks when she needs to; sets up reports and alert thresholds once; the accountant answers only tax questions
Technologies used
understanding questions, querying sources, answers with period and source
Athe report schedule, a trail of every answer, control of data passed to models
Aconversation, recurring reports, the answer archive
Aaccounting data read-only via API or exports; scope depends on the interface
Brevenue per treatment and doctor, statements; via export or API
BIllustrative economic model
Numbers you can check against your own data.
The model counts only time and costs that can be measured. The most valuable thing about the agent is what the model does not count: a treatment price decided in the week of a product price rise, not a quarter later. Put your own figures into the calculator beside.
Run the maths on your data
An illustrative estimate based on your inputs. It models freed capacity, not promised savings.
Business benefits
- An answer to a question about numbers within a minute, with period and source
- Margin per treatment, doctor and room always at hand
- One message in the morning instead of checking three systems and the bank
- An alert when a product price rise eats the margin, at the first invoice
- Recurring reports set up in conversation, without spreadsheets
The management view
- The owner runs the clinic on yesterday’s numbers, not last month’s
- Conversations with the accountant move from summaries to tax and strategy
- The same agent grows with the clinic: a new source is a new tool, not a new project
Board-level KPIs
time from question to answer · reports produced automatically · out-of-contract questions to the office · days from product price rise to a price decision · answers flagged as incomplete data
Security and governance
The automation has exactly the permissions it needs. Not one more.
- The agent has read-only access; it books, pays and changes nothing
- The agent sees no medical data or patient names; it works with treatments, amounts and visit numbers
- Only people named by the owner talk to the agent; the data scope depends on the role
- Every answer records its sources and period; AI Trust Layer controls what data reaches the models
- Data stays in your Microsoft 365 tenant; the robots run in the EU region of UiPath Automation Cloud
Why now
Product prices change more often than the treatment price list; without the margin at hand, the clinic pays the difference.
A conversational agent in Teams is proven technology today, not an experiment; it only needs well-connected sources.
A quick-win deployment takes weeks and starts with a morning report that works from day one.
Relevant executive roles
Numbers when she needs them, in conversation, with no waiting and no spreadsheets
Fewer questions about summaries, more conversations about tax and planning
The morning report and alerts in the same app she works in all day
Common questions and objections
The agent answers only from source data and always states period and source. When data is missing or the question is ambiguous, it says so and asks. It does not make up figures, and every answer can be checked in the record.
No. If the office keeps the books in Comarch Optima or iFirma, the agent reads the data through the interface or exports, without changing anything in the office’s work. We agree this with the office at the start.
No. The agent works with treatments, amounts, doctors and visit numbers. Patient names and health data are out of its reach by design.
When this is not the right solution
- A clinic without electronic access to its accounting data: access first, then the agent
- Expecting the agent to book entries or make payments: that is a different solution with different permissions
- Expecting tax advice from the agent: that is the accountant’s role
A question for the next management meeting
Which three questions about numbers would the owner ask today if the answer came within a minute?
Implementation approach
Scope without ambiguity, before anything is signed.
We deliver
- A finance agent in Microsoft Teams with read-only access to your sources
- Connections to the clinic system, Comarch Optima or iFirma, the bank and the clinic registers
- A morning report built around your morning routine
- Alerts on margin, account balance and expiring vouchers with thresholds set by the owner
- A guide to asking questions and two weeks of assistance after go-live
We need from you
- Read access to the clinic system and to Comarch Optima or iFirma, agreed with the office
- Bank statement exports or access to the bank’s API
- A list of people who may talk to the agent and their data scopes
Stages
Discovery
Sources, the questions asked today, the owner’s morning routine; half a day at the clinic
Sources
Read access, agreements with the office, a dictionary of treatments and categories
Build
The agent in Teams, source tools, the morning report, alerts
Trial
Two weeks of the owner’s questions; we check every answer against the data
Go-live
The agent for named people; a review of questions and thresholds after a month
A quick win. Connecting the sources and agreeing the treatment dictionary take the most work; the agent itself works from the first day of the trial.
A question about margin on Wednesday evening. An answer with figures and source a minute later. A decision the same week.
Write down the three questions about numbers you wait longest for today. Send them to us with the names of your clinic and accounting systems; we will show how the agent would answer them in your clinic.
Send us your three questions about numbersThe neighbouring process usually has the same problem
Yesterday’s takings, today’s staffing, overdue payments and the balance. One message at 7:00 instead of four logins and three calls.
See the solution Finance & accountingDaily takings reconciled every morningThe till says one thing, the card terminal another, the booking system a third. The difference gets explained on Sunday evening.
See the solution Finance & accountingDoctor payouts per treatment, without the spreadsheetA percentage per treatment minus the product, different rates for every doctor. Every month the same spreadsheet and the same discussion.
See the solutionIndustries where we deploy this most oftenAesthetic medicineSmall business & services