This is a hypothetical SasaDaily business case.
A 6-person management consulting firm mainly serves:
Small and medium enterprises with about 20–100 employees.
When a client approaches them, the real questions usually aren't:
“What are management theories?”
But rather:
“Why can’t managers effectively lead their teams?”
“Why do employees leave?”
“Why do so many meetings happen but nothing moves forward?”
“Should tasks be redistributed or should more managers be added?”
The consultant’s real challenge is dealing simultaneously with two very different things:
Professional methods.
And:
What’s actually happening with this client now.
After Google launched Expert Intelligence on August 27, qualified e-books purchased from Google Play Books can be directly added to Gemini Notebook.
This hypothetical consulting firm decided to test a new workflow:
The books provide principles.
Client data provides the current situation.
AI identifies gaps between the two.
The consultant makes the final judgment.
Before: Preparing Client Diagnoses Meant Reviewing Lots of Material
Suppose a client struggles with:
Managers unsure how to give negative feedback.
The firm might have simultaneously:
Initial owner interview notes.
Manager interviews.
Anonymous employee feedback.
Organizational charts.
Meeting minutes.
Existing SOPs.
And several management books the consultant had read previously.
The time-consuming part isn’t necessarily:
“Not knowing management theories.”
But rather:
Remembering a book discussed a topic but not finding the exact passage.
Or:
Knowing a method is good,
but forgetting the author’s important premise.
In the end, the consultant had to:
Re-read books.
Find notes.
Search electronic files.
Then re-compare findings to client issues.
Step 1: Create a Notebook for Each Client
This firm doesn’t build:
a big “all company books + all client data” Notebook.
Instead, it creates a separate Notebook for each formal project.
For example:
“Project on Manager Feedback at Company A.”
It includes:
Client interview records.
Anonymous employee feedback.
Existing company processes.
Meeting notes.
And only the professional books truly relevant this time.
The reason is simple:
Gemini Notebook is designed for each Notebook to be a separate source collection for a specific project.
And different Notebooks remain independent.
This avoids mixing all project data together.
Step 2: First Round—Only Query Management Books, Not Client Data
This is the core of the workflow.
Suppose this round explores:
“How should managers handle employees who repeatedly make the same mistakes?”
The consultant first avoids letting Gemini Notebook see:
Employee interviews.
Owner opinions.
Their own preconceptions.
The first round selects only:
Relevant management books.
Then asks:
“Answer solely based on the currently selected books. What principles do the authors offer regarding ‘employees repeatedly making the same mistakes’? Each key conclusion should be traceable to its source. Please separately present the author’s explicit claims, examples they provided, and your summaries organized by chapter.”
This isn’t an official Google prompt but a workflow designed by SasaDaily for this hypothetical case.
Why Intentionally Hide Client Data at First?
Because if everything is opened immediately,
AI might generate a smooth but mixed answer.
For example:
“Authors recommend first understanding why the employee errs, then increasing feedback frequency, followed by redesigning job responsibilities.”
It sounds good,
but could blend together:
Author principles.
Consultant interview content.
Already-decided client actions.
And AI’s own summaries.
In the end, no one knows:
Which sentence came from whom.
Thus, the first task is clear:
Clarify:
What does the professional source actually say?
Step 3: Check Citations for Every Key Point
Gemini Notebook uses source content to create citations.
When consultants see:
“The author recommends managers identify problem causes first,”
they don’t just copy this into presentations.
They click the citation,
go back to the original text,
and check:
Did the author really say this?
Are there conditions?
Is this a general rule or just a case the author cited?
This step is critical.
Because what consultants sell isn’t:
“AI summarized everything nicely.”
But rather:
Whether professional judgments stand up.
Citations don’t replace verification;
they speed it up.
Step 4: In the Second Round, Include Client Data
After confirming book principles,
now add:
Client interviews.
Employee opinions.
Meeting minutes.
Process documents.
Then ask:
“Using both professional books and client data, separate content into four sections:
- Clear principles from professional sources.
- Client’s actual current situation.
- Where these align.
- Where they don’t align or data is insufficient.
Do not treat client data as the author’s viewpoint.
Do not directly decide company layoffs, salary adjustments, manager changes, or restructuring.”
This last point is vital.
AI can assist research,
but decisions like:
Changing managers.
Adjusting salaries.
Layoffs.
Promotions.
Changing responsibilities,
are management decisions that really impact people and shouldn’t be automatic.
Step 5: AI Produces a Draft Diagnosis, Not the Final Client Answer
This hypothetical firm divides AI output into three layers.
First layer:
What sources prove.
For example:
A management principle applies under certain conditions.
Second layer:
What client data shows.
For example:
Four out of five manager interviews mention inconsistent feedback standards.
Third layer:
What still requires consultant judgment.
For example:
Does this mean retraining managers is needed?
Redefining responsibilities?
Changing the performance system?
Or is it just a problem with two particular managers?
AI can suggest possibilities,
but don’t let:
“AI found a reasonable explanation”
automatically become:
“This is the company’s real problem.”
Step 6: Consultants Decide What to Tell the Client
Final advice delivered to clients might be:
First, establish manager feedback standards.
Redesign one-on-one meetings.
Conduct two additional employee interviews.
Temporarily hold off on organizational changes.
Or:
Insufficient evidence now; avoid major personnel decisions.
These must be decided by the consultant.
Because AI doesn’t know:
Who is about to resign.
How long the company’s cash can last.
How much the owner is willing to change.
Team politics.
Labor laws.
Whether an anonymous interview comment is a special case.
So the workflow isn’t:
Books + interviews → AI → Client answer.
But rather:
Professional source → Verification.
Client data → Comparison.
AI → Gap analysis.
Consultant → Judgment.
How Is This Different From Just Asking ChatGPT?
Suppose a consultant asks a generic AI:
“What should I do about employees repeatedly making mistakes?”
AI might suggest:
Improving feedback.
Increasing training.
Clarifying responsibilities.
Reviewing processes.
The answer might be reasonable.
But consultants also need to answer:
“What is your suggestion based on?”
The value of Expert Intelligence is that when consultants have qualified professional books,
they can make those books Notebook sources,
and return to:
Original author content.
Citations.
Their own client data.
This is far more reliable than:
“I think a certain book said something like this.”
Can You Compare Multiple Management Books Together?
Yes.
At launch, Google’s Expert Intelligence supports over 100,000 qualified books from major publishers.
Gemini Notebook can hold multiple sources.
So consultants can build a project Notebook with:
Two leadership books.
One organizational design book.
Client interviews.
Internal documents.
Then compare whether authors’ views on the same issue conflict.
But the more sources,
the less you can rely solely on summaries.
AI might mix parts from three books to create a:
Fourth method no author fully stated.
So the professional approach is still to:
Check citations back to original texts.
If the Team Buys One Book, Can Six People Use It Together?
This is a crucial business limitation.
Expert Intelligence isn’t:
“If one person in the company buys a book, the entire company automatically gets AI access.”
Google has designed rights control for Play Books.
If a Notebook is shared with others,
co-collaborators might also need their own legitimate copies to access or use book content.
Without proper rights,
the book source will be restricted.
So before establishing a formal workflow, companies must confirm:
Which members really need access to book sources?
Does everyone need it?
Or does one research consultant complete source study and then share legally compliant internal research results with the team?
AI’s convenience doesn’t override content licensing.
How Much Time Could This Case Save?
Below are all:
SasaDaily’s hypothetical figures.
Not official Google data.
Not tested results from any real consulting firm.
Suppose the firm needs to prepare weekly:
10 client research and meeting reports.
Previously each took about:
40 minutes.
This includes:
Consulting books.
Finding notes.
Confirming author views.
Organizing client interviews.
Making initial comparisons.
Totaling:
400 minutes per week.
Using Gemini Notebook,
AI first finds potentially relevant sources,
Consultants check citations and align with client data.
Assume the average time drops to:
22 minutes.
For 10 reports:
220 minutes.
Saves approximately:
180 minutes—or about 3 hours weekly.
Assuming internal consultant cost of NTD 1,800 per hour,
this equals roughly:
NTD 5,400 per week.
Or about:
NTD 21,600 per month.
Again, these are SasaDaily’s hypothetical numbers.
Actual time savings depend on:
Book qualification.
Number of sources.
Document length.
Consultant familiarity.
Case complexity.
Amount of verification needed.
The Real ROI Might Not Be Just Time Saved
For professional service firms,
the bigger value may not be:
“Flipping fewer pages.”
But rather reducing another cost:
Consultants mistaking memory for source.
After years, it’s easy to say:
“I’ve read that before.”
“That author probably said something.”
“Management usually is like this.”
But when facing clients,
“Probably”
versus
“I can now verify the source directly”
are completely different levels of professionalism.
This workflow can improve:
Traceability.
Research speed.
Internal quality control.
And consultants’ awareness of:
When they’re quoting sources versus making professional judgments.
What Tasks Should Not Be Entrusted Directly to Gemini Notebook?
Even if all sources are accurate,
AI cannot directly decide:
Who to fire.
Who to promote.
How much to adjust salaries.
Which manager has issues.
Whether to restructure.
Whether discrimination is involved.
Legal labor responsibilities.
Such decisions involve:
Human rights.
Law.
Ethics.
Company culture.
And circumstances AI cannot see.
The mature approach isn’t:
Turn AI into the consultant.
But rather:
Let AI help consultants find evidence faster.
What Can Small Companies Learn Most from This Case?
It doesn’t have to be a management consulting firm.
If your work is:
Corporate training.
Marketing consulting.
Design consulting.
Financial planning.
Research services.
Content strategy.
Education.
Or any task requiring simultaneous handling of:
Professional knowledge + real client data
You can separate the two.
First round, ask:
What do professional sources really say?
Second round, ask:
How does this relate to my client?
Finally, third step is:
What should we actually do?
AI can speed up the first two.
The third step
still belongs to the responsible person.
Because true value in professional services
is never:
Knowing the most theories.
But knowing:
Which theory to apply—or not apply—to this client.
Today, grow a little with AI.
Learn an AI tip every day.
Save a little time every day.
Improve a little skill every day.
SasaDaily, growing with you.
