How Kimi + Gemini Notebook can turn a pile of files into a report — without turning your brain into mashed potatoes

You know that folder.

The one called:

“Q3_Final_Final_v7_USE_THIS_ONE.xlsx”

Inside it are 14 spreadsheets, three PowerPoints, six PDFs, a few emails, one meeting transcript and something called new_new_FINAL2.xlsx.

Your boss asks:

“Quick” is corporate language for:

“Please spend your weekend doing archaeology.” 😂

But something important has changed.

AI tools are increasingly capable of taking a collection of documents, spreadsheets, reports and research sources and turning them into something much closer to a working analysis rather than merely answering one question at a time.

The interesting part isn't asking AI:

The interesting part is saying:

That is a completely different workflow.

And it works surprisingly well for both your job and your investing research.

1. The new workflow: Files → Evidence → Analysis → Report

A recent demonstration by My Online Training Hub showed this concept using Genspark Super Agent.

The input was deliberately messy: customer emails, sales data, returns information and sales targets.

The output wasn't just a summary.

Genspark produced:

• an analysis spreadsheet
• a management presentation
• an action plan

The author correctly emphasised the important bit: the AI output still needs checking.

That is the lesson I want to steal — not the software.

Because you can build a similar workflow using several other AI tools.

My preferred mental model is:

COLLECT → QUESTION → CROSS-CHECK → ANALYSE → REPORT → CHALLENGE

The AI does the heavy lifting.

You remain the person holding the steering wheel.

2. At work: turn “Where is the problem?” into a report

Imagine you're responsible for regional sales operations.

You have:

• monthly sales by customer
• gross-margin data
• overdue receivables
• customer complaints
• sales targets
• inventory data
• management emails

Normally you might open six applications and start copying numbers into PowerPoint.

Instead, create a project and upload the relevant files.

Then give the AI a proper assignment.

Try something like:

Notice what happened.

You didn't ask:

You gave it a job description.

That's the difference between using AI as a fancy search box and using it as a junior analyst.

3. But PLEASE don't upload your company's crown jewels 😬

This is where the productivity conversation gets uncomfortable.

Your company may have:

• customer pricing
• employee information
• bank details
• contracts
• acquisition plans
• unpublished financial results
• customer disputes
• strategic plans
• confidential management discussions

Don't casually dump these into whichever shiny AI website appeared on TikTok yesterday.

Your employer's AI policy comes first.

The question isn't simply:

The better question is:

That's a very different question.

Enterprise offerings can have materially different security, governance and data-handling arrangements from consumer accounts. For example, Perplexity's Enterprise tiers advertise controls including encryption, access controls and a policy that enterprise data isn't used for model training.

ChatGPT's Business, Enterprise and Edu projects similarly inherit workspace security and data controls, while consumer plans have different data-use settings.

So use a simple rule:

🟢 Green

Public information, your own notes, public filings, public research.

Upload freely, subject to the tool's terms.

🟡 Yellow

Internal company material that isn't highly sensitive.

Check your company's AI policy first.

🔴 Red

Customer personal data, confidential financial information, trade secrets, unpublished results, passwords, legal documents or material that could cause damage if leaked.

Don't upload it unless your organisation explicitly authorises the workflow.

AI productivity is wonderful.

Explaining to your boss why confidential customer data appeared in a public AI conversation is considerably less wonderful.

4. Now comes the fun part: YOU become the analyst

Here's where I think retail investors can get enormous value.

Most individual investors don't have an information problem.

We have an information soup problem.

You might collect:

• annual reports
• quarterly filings
• earnings transcripts
• investor presentations
• company presentations
• analyst reports
• industry reports
• competitor filings
• valuation spreadsheets
• news articles
• management interviews
• your own notes

Then you stare at all of it and ask:

😂

Instead, build your own Stock Research Notebook.

5. Example: Build your own NVIDIA research report

Let's use NVIDIA as an illustration.

Suppose you collect:

1. NVIDIA's latest 10-K
2. latest 10-Q
3. earnings release
4. investor presentation
5. earnings-call transcript
6. competitor information
7. your valuation spreadsheet
8. selected industry research

You now have a mini research database.

For example, NVIDIA's latest reported second quarter fiscal 2027 revenue was $96.2 billion, up 106% year over year, while Data Center revenue reached $89.0 billion, up 117%. Gross margin was 75.0%.

But here's the important bit:

Don't ask AI whether NVIDIA is a good investment.

That's lazy.

Ask better questions.

Prompt #1 — The bull case

Prompt #2 — The bear case

Then deliberately reverse the question:

Now you've got something much more interesting.

Bull vs Bear.

6. Then ask the AI to attack BOTH sides

This is my favourite step.

Tell it:

This is where AI becomes much more useful.

Not because it magically knows the future.

Because it can help you interrogate your own thinking.

7. Then make your own research report

Ask Kimi or Gemini Notebook to structure the final report as:

Now you've created something much more valuable than an AI summary.

You've created your own investment research framework.

8. Kimi vs Gemini Notebook vs the others

Here's the practical cheat sheet.

The exact limits change frequently, so don't treat “free” as meaning unlimited.

For example, Kimi's current membership documentation shows paid tiers beginning at $49/month, with Agent Swarm appearing on higher tiers and dedicated Kimi Claw on the $199/month tier and above.

Perplexity currently offers a free Standard tier with limited file uploads and research functionality, while Pro provides expanded research, file analysis and access to additional models.

Google's current AI plans provide expanded access to Gemini Notebook; Google AI Plus, Pro and Ultra tiers.

And Genspark continues to operate a credit-based model, with free users receiving daily credit refreshes and paid tiers providing substantially greater access.

Translation: don't choose an AI because someone says “it's free.” Choose it because it fits the job.

9. My practical AI stack

If I were building a retail-investor workflow today, I'd keep it simple:

🔎 Perplexity

Find the information.

Use it for current news, competitors, industry developments and source discovery.

📚 Gemini Notebook

Organise the evidence.

Load your filings, transcripts, reports and notes into one research environment. Google's latest version can reason across sources and generate formats such as charts, spreadsheets and slide decks.

🤖 Kimi

Work the evidence.

Use it when you want heavier multi-file processing, document generation, spreadsheet work or agent-style workflows.

🧠 ChatGPT

Challenge and refine the thesis.

Use a project to maintain your research framework, ask follow-up questions and turn the analysis into something readable. Projects support uploaded reference files and custom instructions.

You don't need all four.

One good tool used properly beats four subscriptions collecting digital dust.

10. The secret weapon isn't the AI

Here's the uncomfortable part.

AI can make a mediocre analyst faster.

It can also make a good analyst dramatically more productive.

The difference is the quality of the questions.

Garbage in → garbage out.

But there's an even more dangerous problem:

Convincing garbage in → convincing garbage out.

The SEC, NASAA and FINRA explicitly warn investors not to rely solely on AI-generated information because AI can produce inaccurate, incomplete, outdated or fabricated information.

So don't let AI become your investment guru.

Make it your:

research assistant + calculator + organiser + devil's advocate + report writer.

Then verify the important stuff yourself.

For U.S. stocks, the SEC's EDGAR database remains an excellent primary-source starting point for company filings.

11. The 30-minute Retail Investor AI Workflow

Here's the practical version.

Step 1 — Gather

Collect the latest:

10-K + 10-Q + earnings release + transcript + investor presentation

10-K = annual report.

10-Q = quarterly report.

Step 2 — Upload

Put them into your research notebook/project.

Step 3 — Understand

Ask:

Step 4 — Measure

Ask for:

• revenue growth
• gross margin
• operating margin
• free cash flow
• return on invested capital
• debt
• valuation
• dilution
• segment growth

Step 5 — Attack

Ask:

Step 6 — Compare

Add two or three competitors.

Step 7 — Value

Give the AI your valuation assumptions.

Ask it to calculate:

Bear / Base / Bull

Step 8 — Monitor

Create a quarterly dashboard.

Step 9 — Report

Generate your final investment memo.

Step 10 — Verify

Go back to the original filings.

AI drafts. You decide.

12. The biggest productivity upgrade is not speed

It is reusability.

Don't create one NVIDIA report.

Create an NVIDIA Research System.

Then next quarter:

Now you're not repeatedly researching NVIDIA.

You're maintaining a living research process.

Do this for:

NVDA → MSFT → AMD → AVGO → CRWD → PLTR → ETFs → whatever you actually own.

Eventually your biggest asset isn't the AI subscription.

It's your research library + repeatable framework + accumulated judgment.

That is how retail investors start behaving less like people reading financial headlines over breakfast...

…and more like a tiny research department.

Without needing to hire 12 analysts.

Or buy everyone lunch.

🧰 Your “Don't Be an AI Potato” Checklist

Before trusting an AI-generated report:

☐ Did I use primary sources?

☐ Did I upload the latest filings?

☐ Did I distinguish facts from assumptions?

☐ Did I ask for the bear case?

☐ Did I ask what could falsify the thesis?

☐ Did I compare competitors?

☐ Did I check the numbers against the original filing?

☐ Did I check whether the information is current?

☐ Did I consider valuation rather than just business quality?

☐ Did I identify what the market already expects?

☐ Did I avoid uploading confidential employer information?

☐ Did I make the final investment decision myself?

If you tick all 12:

Congratulations.

You've upgraded from AI user to AI-assisted analyst. 🤖📈

💡 Wealth Builder Wisdom

The modern investor doesn't necessarily need more information.

We need a better way to organise, interrogate and act on information without drowning in it.

That's where newsletters like Wealth Builder can help. Instead of jumping between hundreds of headlines, opinions and “hot stock” calls, a good investing newsletter gives you frameworks, ideas and research angles you can investigate yourself.

AI then becomes your research multiplier: it can organise filings, compare companies, challenge assumptions and turn scattered information into a repeatable process. The goal isn't to outsource your judgment. It's to free more of your limited time for the decisions that actually matter — valuation, risk, position sizing and patience.

If you're trying to build passive income and long-term financial freedom without becoming a full-time market detective, discover other like-minded investing newsletters here: 👉 Explore Wealth Builder and other investing newsletters​

Final Thought

The biggest mistake is thinking:

No.

The better mindset is:

That's a subtle difference.

But it could become a very profitable one.

Because when everyone gets access to the same AI...

the edge moves to the person who asks better questions, supplies better evidence and checks the answer more ruthlessly.

Research. Challenge. Decide.

#AIInvesting #RetailInvesting #KimiAI #GeminiNotebook #NotebookLM #InvestmentResearch #Productivity #ArtificialIntelligence #WealthBuilding #PassiveIncome

Notes & sources

1. Mynda Treacy, My Online Training Hub, “From Business Files to Management Reports with Genspark AI,” June 18, 2026. This is the structural inspiration for the workflow: messy business files → analysis → spreadsheet → presentation → action plan. My Online Training Hub
2. Kimi Help Center, current membership documentation. Used to correct the original assumption that advanced Kimi Agent/Swarm/Claw functionality is universally free. Kimi
3. Google, Gemini Notebook / NotebookLM, July 16, 2026. Google renamed NotebookLM to Gemini Notebook while retaining its research-oriented purpose. blog.google
4. Google, “Do better research with NotebookLM,” June 8, 2026. Supports the description of source-grounded research, advanced reasoning, cloud-computer analysis and generated outputs. blog.google
5. Perplexity Help Center, September 2, 2026. Used for current free/Pro/Enterprise capabilities and file/research limits. Perplexity AI
6. OpenAI Help Center, “Projects in ChatGPT,” updated 2026. Used for current project/file capabilities and data-control distinctions. OpenAI Help Center
7. Genspark Help Center, current credits and enterprise documentation. Used for the current free-credit/paid-plan description. Genspark
8. U.S. Securities and Exchange Commission (SEC), NVIDIA FY2027 Q2 Form 10-Q, July 26, 2026. Used for NVIDIA revenue, Data Center revenue and margin examples. SEC
9. U.S. SEC, NVIDIA FY2027 Q2 earnings release, August 26, 2026. Used for NVIDIA's reported quarterly results and Jensen Huang quotation. SEC
10. SEC, Investor Alert: “Artificial Intelligence (AI) and Investment Fraud,” SEC/NASAA/FINRA, January 25, 2024. Used for the warning against relying solely on AI-generated investment information. Investor.gov
11. SEC EDGAR, the SEC's Electronic Data Gathering, Analysis, and Retrieval system, used as the example of a primary-source repository for public-company filings