This is a hypothetical business case by SasaDaily.
A small property management company with five staff members
opens Gmail every morning,
and what they see most often are not complicated business problems,
but issues like:
"The bathroom is leaking."
"The air conditioning isn’t cooling."
"The door lock is broken."
"The stairway light is out."
"Can someone come take a look?"
The problems themselves aren’t difficult.
What really consumes time is the chain of small tasks that follow:
Open email.
Confirm which building.
Which unit.
Check photos.
Find the folder for the building.
Create a repair record.
Move attachments into it.
Notify maintenance personnel.
Reply to the tenant:
Received your request.
Each task takes only a few minutes,
but with a dozen or more requests daily,
time slowly disappears.
The new Automation Steps announced by Workspace Studio on September 2
are starting to address this kind of work.
But this company wouldn’t set up:
“Let AI handle all tenant emails entirely by itself.”
Instead, it only automates:
organizing, moving files, and internal notifications.
Commitments involving:
time,
cost,
responsibility,
or official statements to tenants
remain under human control.
Clarifying: This article does not mean all new features are live today
Google has announced:
Move Drive file.
Copy Drive file.
Send a Chat reply.
Reply to email.
But today is:
September 3, 2026.
According to Google’s rollout schedule,
the new Drive and Chat Steps are expected to launch from
September 8
onwards.
The Gmail Reply Step will rollout starting
September 14.
This case is designed to show
how the features can be deployed after they become available,
not pretending every Workspace account has these four steps yet.
Step 1: Only allow specific repair emails into the Flow
The property company wouldn’t set:
“Any received email.”
as the sole condition,
because the inbox also contains:
bills,
landlord messages,
vendor quotes,
contracts,
spam,
internal messages.
The more reasonable Flow only processes emails sent
to designated repair mailboxes,
or matching:
specific subjects,
certain form notifications,
or specified management processes.
Only those enter this automation.
The first layer of safety is not
AI intelligence,
but
limiting its scope upfront.
Step 2: Gemini organizes first, doesn’t decide yet
After a tenant’s email arrives,
Gemini’s first job is just
organizing.
For example, extracting:
which building,
which unit,
type of issue,
whether there are photos,
tenant’s preferred contact time,
whether data is missing.
AI doesn’t immediately ask questions like:
“Who is responsible?”
“Should the landlord pay?”
“Is the tenant at fault?”
Because these involve more than just data organization.
The AI’s first round only creates a
repair intake.
That is, transforming scattered information into a workable case summary.
If data is incomplete, don’t guess
For example, if a tenant writes only:
“Air conditioner is broken.”
but provides no:
unit number,
photos,
fault details,
or whether the AC is not working at all or just cooling poorly,
the Flow shouldn’t automatically fill in
“3rd floor, unit 302, compressor failure.”
when no evidence exists.
A better approach is to mark
“insufficient information.”
and create a
“pending data”
task rather than turning guesses into official repair records.
Step 3: Automatically copy the standard repair folder template
Each repair case may require a fixed folder structure, such as:
Tenant’s email,
original attachments,
site photos,
quotes,
before and after work records,
invoices,
closing documents.
The team first prepares a
standard repair case template folder.
When a new case qualifies,
Workspace Studio’s new
Copy Drive file or folder
step can copy this template.
Each case therefore has the same data structure,
saving employees the trouble of searching, copying, renaming, and deleting unnecessary old files daily.
Step 4: Automatically move attachments to the correct location
Using
Move Drive file or folder,
if a tenant attaches,
photos of leaks,
air conditioner images,
or door lock videos,
the Flow can move the confirmed files
into the case folder.
Note:
The key here is not
“AI can move files,”
but
only move once case and target folder are clearly confirmed.
If building or unit can’t be confirmed,
the files stay
pending classification,
instead of forcing the automation to place them in the wrong folder.
Step 5: Automatically notify internal staff via Google Chat
After data organization,
the next question is:
Who should know?
In the past, managers manually copied and pasted messages like:
“Leak on 5th floor of building A, photos attached, please check.”
into Google Chat.
Workspace Studio’s new
Send a Chat reply
step can send the organized data directly to specified internal chat conversations,
for example:
plumbing maintenance,
air conditioning repair,
public areas,
emergency cases.
Different types follow different internal workflows.
This kind of notification is ideal to automate
because it’s
internal collaboration
and not an official commitment to tenants.
But AI does not directly decide “someone will arrive this afternoon”
This is the most crucial boundary in property management.
AI might see
a leak
and based on past experience suggest
“priority handling recommended.”
That’s fine.
But it cannot reply itself
“The technician will arrive at 3 PM today.”
because AI does not know
where the technician is now,
whether the previous job is delayed,
whether building access is possible,
if the tenant will truly be home,
or if parts are available.
Once that message goes out,
it becomes a
company commitment.
So from here on,
a human must get involved.
Step 6: Separate tenant replies into two types
Not all email replies need to be written manually.
The company can divide them into two categories.
First:
Low-impact confirmations.
For example:
“We have received your repair request and are scheduling it.”
This does not include:
price,
time,
responsibility,
compensation,
or clear outcomes.
This kind of standard acknowledgement,
after thorough testing,
can be considered for automated replies.
Second:
Official commitments.
For example:
“The technician will arrive at 2 PM tomorrow.”
“This repair will be covered by the landlord.”
“A fee of 1,500 is required.”
“Damage is tenant’s responsibility.”
“We will replace the entire unit.”
All these require
prior approval.
Where does Google’s Approval come into play?
Workspace Studio’s Approval mechanism allows steps controlled by admin security policies to
pause,
wait for human review,
and then be approved or rejected
before executing.
This company expects managers to place
parts involving sharing information externally,
modifying critical shared data,
or causing external consequences
within the Approval process.
But the team doesn't assume
having Approval makes the entire Flow safe.
Because Approval only protects the steps it’s applied to
Consider the process:
Gemini determines the unit number.
↓
Move Drive file.
↓
Notify maintenance.
↓
Final Email reply which requires Approval.
If Gemini mistakenly reads unit 501 as 510,
files may be moved incorrectly,
and maintenance could be notified about the wrong case.
Although the final email goes through Approval,
prior mistakes can’t be automatically corrected.
Therefore, besides Approval,
the company also adds
check-if conditions
on key data.
For example, confirming three conditions first:
Only if
building name exists,
unit number exists,
and the repair type is properly classified,
will the case be created.
If any are missing,
the process diverts to
manual data supplementation
instead of forcing the next step.
This is where Workspace Studio’s Conditional Flow proves truly valuable,
not just a simple
AI → Action → Action → Action
but a
conditional path that proceeds only if criteria are met.
Step 7: Never test on real tenants the first time
Google specifically warns that Workspace Studio’s
Test run
will actually execute actions — it’s not just a preview.
So if the flow involves:
sending Chat messages,
sending emails,
or modifying files,
those actions may really happen during testing.
This company first tests using:
test tenant emails,
test buildings,
test Drive folders,
and internal test Chats.
Replies to tenants initially only go internally,
to verify:
cases were categorized correctly,
files moved properly,
Chats sent to the right people,
and variables assigned accurately,
before gradually connecting to real operations.
Launch won’t be full scale at once either
The company plans three phases.
Phase One | Organize only
First, have AI:
read emails,
organize cases,
and create drafts without changing official data.
Run 20 cases
with manual verification one by one.
Phase Two | Add internal actions
Once organization stabilizes,
start:
copying folders,
moving files,
internal chats,
but all tenant emails still replied to manually.
Phase Three | Enable only very low-risk automated replies
Such as:
“Received, we are processing your request.”
This kind of fixed acknowledgement.
All formal arrangements involving:
time,
cost,
responsibility,
still require
human approval.
This is a reasonable agent rollout.
How much time can this really save?
For example, assuming
SasaDaily hypothetical numbers.
The company receives roughly
40
repair emails per week.
Originally, each took about
7 minutes
to complete reading, organizing, folder creation, attachment filing, and colleague notification.
40 × 7 minutes =
280 minutes, or about
4 hours 40 minutes per week.
After automation?
SasaDaily hypothetical numbers again.
If the Flow handles initial organization,
creates case folders,
files attachments,
and sends internal notifications,
staff only need to
spend an average of
3 minutes
checking results and handling exceptions.
40 × 3 =
120 minutes, or
2 hours per week.
Theoretical difference:
about
2 hours 40 minutes saved weekly.
Estimating 4 weeks a month:
approximately
10 hours 40 minutes saved monthly.
Converted to money
SasaDaily hypothetical numbers.
If the company's internal labor cost is
NT$550 per hour,
10.67 hours × 550 =
about
NT$5,869 per month.
Note:
This is not Google’s claimed ROI
nor any real company’s measured result.
It’s a demonstration of
turning “feeling faster” automation
into
actual saved labor time.
And it can’t be directly counted as “additional NT$5,869 earned”
Because the company may not actually reduce headcount.
What likely happens is
managers have more time
to track repair progress,
communicate with tenants,
review site photos,
and handle truly difficult cases.
So more accurate to say
this Flow
frees up about 10.7 working hours.
How much business value results from that time
requires further tracking.
The KPIs to record go beyond time
The company would also track:
number of weekly repair requests,
auto-classification success rate,
number of attachments wrongly filed,
cases with missing data,
number of summaries needing manual edits,
approval rejections,
number of tenant follow-ups,
average first reply time,
and how long cases actually take to close.
Because AI can quickly organize emails,
but if repairs aren’t completed faster,
customer experience may not improve.
Automation can easily create a false impression
Before:
cases created 10 minutes after request.
Now:
cases created in 30 seconds.
Dashboards show
great efficiency gains.
But if technicians only arrive 2 days later,
and tenants get results 2 days later,
the company merely
sped up administrative tasks,
not the entire repair service.
The true business KPIs should measure whether cases
are completed faster,
have fewer omissions,
less data errors,
and less need for tenants to repeat information.
Which repair requests are best to automate first?
Pick those that are:
high frequency,
have relatively fixed formats,
and low consequences,
such as:
public lighting,
routine air conditioning inspections,
door lock repairs,
non-urgent leak photo organization,
regular equipment maintenance.
The cases that shouldn’t be automated from day one include:
gas leaks,
major water leaks,
fire,
power hazards,
personal safety issues,
formal claims,
and disputes.
These require
immediate escalation to human staff,
not letting the agent attempt to handle the entire case alone.
AI’s most valuable role might be knowing “Not to handle this itself”
For example, if the system detects
“Smell of gas,”
the best automation isn’t
searching repair procedures,
creating a neat summary,
and sending a reply email,
but rather
immediately stopping the normal flow,
marking the case as urgent,
alerting a human,
and avoiding automatic commitments.
This is a mature process.
Another crucial issue: tenant data
Property companies may hold
names,
phone numbers,
addresses,
leases,
house photos,
repair records,
and even in-home images.
“Just because Workspace Studio can technically handle it”
does not mean
“all the data should flow through this automation.”
Companies must still decide based on
data classification,
access controls,
retention policies,
tenant contracts,
and regulations
which data can enter,
and which can only be viewed by certain people.
This is the same as any enterprise AI setup.
No tool permissions replace proper data governance.
How does this case differ from the August 12 food wholesaler example?
August 12’s case focused on
how Workspace Studio integrates
Gmail,
Drive,
Gemini,
Sheets,
and Chat
into a first AI automation workflow.
Today’s case adds a new question:
once flows can actually move data, reply in Chat, and even reply to email, how does a company distinguish between “organizing” and “making commitments”?
In other words:
going from
information automation
to
action automation.
This is a completely different step.
Finally, the entire process in one diagram
Tenant repair email
↓
Check if from designated repair source
↓
Gemini organizes:
building/unit/issue/attachments/missing info
↓
Check if data is sufficient
↓
Copy repair case template folder
↓
Move attachments into correct folder
↓
Send Chat reply to notify internal staff
↓
Low risk:
acknowledgment sent
↓
Consider automated reply
Involving:
time,
cost,
responsibility,
compensation,
formal repair scheduling
↓
Approval
↓
Manager reviews original data
↓
Official reply to tenant
This company really automates
not
the entire property management,
but the repeated, rule-based administrative tasks
performed dozens of times daily.
Humans stay to handle
judgment,
exceptions,
responsibility,
commitments,
and real tenant communication.
This is what makes Workspace Studio’s new actions most valuable to small businesses:
not AI doing the whole process from first email to last, but removing repetitive middle steps while keeping humans involved where consequences matter.
Today, progress a little with AI.
Learn one AI skill a day.
Save a little time daily.
Improve a little every day.
SasaDaily grows with you.
