AI agents explained
What a operations reporting agent does
Step through an example
Pick an example business
Example: Property management firm · Charlotte
Overnight data pullExample - Rent ledger: 212 units
- Maintenance tickets: 37 open
- Leases ending in the next 60 days: 14
- Pulled: 10/19/2026, 6:00 AM
Step 1 of 5
Every morning it reads the latest figures from your property management system. It only reads; it doesn't change anything.
Agent does this
Your report rulesExample - Flag: rent unpaid 5 days after the due date
- Flag: urgent ticket open more than 24 hours
- Flag: lease ending with no renewal offer sent
- Send to: owner and portfolio manager, 7:30 AM
Step 2 of 5
Checks the figures against the things you've said you care about, so the summary leads with what needs attention.
Agent does this
Something doesn't add upExample - Unit 4B: October rent shows unpaid, 18 days past due
- Bank feed: $1,450.00 received 10/12/2026, reference "Unit 48"
- Unit 48 doesn't exist in this building
Step 3 of 5
Spots a payment that probably belongs to Unit 4B but came in under the wrong reference. Moving money between tenant ledgers isn't its job, so it flags the likely match for your bookkeeper.
Asks a person
Morning summary (email)Example - Mon 10/19/2026 · 212 units
- Rent: 6 unpaid past 5 days (1 probably misfiled, see 4B)
- Maintenance: 2 urgent tickets open over 24 hours
- Leases: 3 ending soon with no renewal offer
- Every item links to its record
Step 4 of 5
Writes the summary in plain English, leads with what needs attention and links every item to the record it came from.
Agent does this
Your decisionsExample - 4B payment: bookkeeper to confirm and reassign
- Urgent tickets: call the HVAC contractor about Unit 12A
- Renewals: send offers to 3 residents today
Step 5 of 5
The owner and portfolio manager decide what to act on and who does it. The agent reports; it doesn't call contractors or move money.
A person decides
Example: Parcel courier firm · Glasgow
Overnight data pullExample - Deliveries yesterday: 486
- Failed first attempts: 31
- Drivers out: 22 · Vans off the road: 2
- Pulled: Tue 20/10/2026, 06:00
Step 1 of 5
Reads yesterday's results from your delivery software and van tracking each morning, before the shift starts.
Agent does this
Your report rulesExample - Flag: failed attempts well above normal for an area
- Flag: a customer with 3 or more late deliveries this week
- Flag: a van due for a service within 7 days
- Send to: operations manager on Teams, 06:30
Step 2 of 5
Compares yesterday with a normal day for each area and each customer, instead of reporting every number.
Agent does this
Unusual patternExample - Failed attempts, G51 postcodes: 11 yesterday · usually 2 or 3
- Reason logged by drivers: "no access"
- All 11 at the same business park
Step 3 of 5
Spots a cluster of failed deliveries at one site with the same reason. It can't know why, so it flags it for a person to look into rather than guessing.
Asks a person
Morning summary (Teams)Example - Tue 20/10/2026 · 486 deliveries
- Failed first attempts: 31, including 11 at one G51 business park
- Late: Clyde Print Ltd, 3rd late delivery this week
- Van 14: service due 23/10/2026
Step 4 of 5
Posts a short summary in your team chat before the morning briefing, with the unusual items first.
Agent does this
Your decisionsExample - G51: call the business park's site office about access
- Clyde Print Ltd: account manager to call today
- Van 14: book the service for Friday
Step 5 of 5
Your operations manager decides what to act on at the briefing. Drivers, routes and customer calls stay with your team.
A person decides
Example: Physiotherapy group · Adelaide
Overnight data pullExample - Clinics: 3 · Physiotherapists: 14
- Appointments yesterday: 168
- Didn't attend: 12 · Late cancellations: 9
- Pulled: Wed 21/10/2026, 6:00 am
Step 1 of 5
Reads yesterday's figures from your practice management system across all three clinics. It sees appointments and billing, not clinical notes.
Agent does this
Your report rulesExample - Flag: missed appointments well above normal at any clinic
- Flag: a physio's diary under half full next week
- Flag: invoices unpaid after 30 days
- Send to: practice manager, 7:00 am
Step 2 of 5
Compares each clinic with its own normal week, so one quiet day doesn't set off alarms.
Agent does this
Unusual patternExample - Norwood clinic: 8 missed appointments yesterday · usually 2 or 3
- 7 of the 8 were booked online last week
- Reminder texts for those bookings: not sent
Step 3 of 5
Notices that most of the missed appointments were recent online bookings that never got a reminder. That looks like a settings problem, so it flags it for the practice manager rather than changing anything.
Asks a person
Morning summary (email)Example - Wed 21/10/2026 · 3 clinics
- Norwood: 8 missed, reminder texts not sent for online bookings
- Next week: 2 physios' diaries under half full
- Unpaid after 30 days: 17 invoices, A$2,980.00, mostly ReturnToWorkSA and CTP claims (GST-free)
Step 4 of 5
Sends a short summary with the Norwood problem first, and links each item to where it came from.
Agent does this
Your decisionsExample - Norwood: check the online booking reminder setting today
- Front desk: call the 8 patients to rebook
- Admin: follow up the 17 unpaid invoices
Step 5 of 5
The practice manager decides what to fix and who does it. Anything to do with patient care stays with your physios.
A person decides
What it connects to
- Job, project, property or practice management systems
- Booking and delivery software
- Accounting software
- Email and team chat, like Microsoft Teams or Slack
- The spreadsheets you already keep
Where a person stays in charge
- You can choose exactly which figures it can read, and keep it read-only.
- You can decide what counts as unusual for your business, and change it as you go.
- You can choose who gets the summary, when and where.
- You can keep every action that follows a flag as a decision only a person makes.
What we'd need from you to build one
- The handful of numbers you check, or wish you could check, every morning.
- Read-only access to the tools those numbers live in.
- What counts as unusual for you, even roughly: what a normal week looks like and what a bad one looks like.
- Who should get the summary, and how they like to read it.
Other jobs an agent can do
Document intake agent
Reads the documents that pile up, pulls out the details and enters them into your systems, flagging anything that doesn't add up.
Follow-up and chasing agent
Sends polite reminders about unpaid invoices, missing documents and quiet leads on your schedule, and stops as soon as someone replies.
See how we build it and what it costs · All AI agents explained
Common questions
Isn't this just a dashboard?
A dashboard shows you numbers when you go and look. The agent reads them for you, writes down what changed and what looks off, and sends it to you. You can have both.
What if the data in our systems is wrong?
Then the summary will be too, which is why each item links back to its source. You can also have it point out things that don't match, like a payment under the wrong reference.
Can it fix the problems it finds?
Only if you choose to set that up. You can start with it only reporting, then add small actions later, with a person approving each one.
Who sees the reports?
Whoever you choose. You decide who gets the summary and what each person sees.