Webinar: Practical AI for Busywork

August 30, 2025

by Royce Sin

Date: August 29, 2025

Hosts: Christopher Wiebe & Royce Sin

Topic: A practical framework for implementing AI to automate administrative tasks, empower employees, and scale small businesses effectively.

The Big Picture: Why Most AI Projects Fail

The webinar begins by addressing a startling statistic from an MIT report: 95% of Generative AI pilot projects fail to make it into production.

Royce Sin explains the primary reasons for this high failure rate:

Unclear Goals: Companies adopt AI for the hype ("fear of missing out") rather than to solve a specific, well-defined problem.
Lack of Trust: Employees are often hesitant to adopt AI tools because they don't understand how they work or trust their output.
Poor Data Quality: AI is only as good as the data it's given. The principle of "garbage in, garbage out" is especially true for AI.

The key takeaway is not to rush into a massive AI overhaul. Instead, the advice is: "Don't rush AI, but don't delay AI." Start small, focus on clear problems, and build from there.

A Framework for Smart Implementation: F.I.R.S.T.

Before jumping to a technological solution, it's crucial to analyze the underlying process. The webinar introduces the F.I.R.S.T. framework as a way of thinking, always putting Process first, People second, and Technology last.

Work through these steps in order and stop at the simplest effective solution.

Filter: Can you eliminate the task or process entirely? This is the easiest fix.
Improve: Can you make the existing process better or more efficient without new tools?
Reassign: Is the right person or team handling the task? Can it be shifted to someone else?
Systematize: Can this process be automated using technology (including AI)? This step comes only after the previous ones have been considered.
Track: How will you measure the results to ensure the new system is working better than the old one?

Key Quote: "If you have a bad process and you automate it, you just make the bad process go faster."

Case Study: Martha the Bookkeeper

The webinar uses the example of "Martha," a fictional bookkeeper, to demonstrate how to apply the F.I.R.S.T. framework to common business problems. Martha's goal is to spend less time on low-value "busywork" and more time on high-value strategic tasks like financial analysis and forecasting.

Here are her three biggest time-sinks and their AI-powered solutions.

Problem 1: Constant Policy Questions

Martha is constantly interrupted by staff asking simple questions about the company's expense policy because nobody wants to read the full document.

Framework Application:
- Filter: Eliminate the need for staff to ask Martha directly.
- Improve: Make the process of updating the policy document faster for Martha.
Solutions & Tools:
- Create an AI-Powered Q&A Bot with Google Notebook LM:
  - This tool is built into Google Workspace (Microsoft has a similar tool called Copilot Notebook).
  - Simply upload your policy documents (Google Docs, PDFs, etc.) into Notebook LM.
  - It instantly creates a private chatbot that staff can ask questions to (e.g., "What's the limit for meals?"). The bot answers only using the information in your documents, ensuring accuracy.
- Write & Update Policies Faster with Google Gemini:
  - Use Gemini (or another AI assistant) to write policy drafts from scratch or update existing ones.
  - You can give it commands like, "Find all mentions of our old office address and update it to the new one." This saves significant time.
- Pro Tip - Use Voice Typing: Use the built-in voice typing feature on your computer (Windows key + H on Windows) to dictate notes, logs, or documentation ideas directly into Gemini. The AI can then clean up your spoken words into a polished document.

Problem 2: The Tedious Reimbursement Process

The staff finds submitting expenses tedious, and Martha spends hours chasing receipts, doing manual data entry, and reviewing every submission.

Framework Application:
- Filter: Eliminate the reimbursement process altogether. If the company pays directly, staff don't need to be reimbursed.
- Reassign: Shift the responsibility of approvals from Martha to individual team managers.
Solution & Tool (Non-AI):
- Implement a Spend Management Platform like Float Financial:
  - These platforms (many are free for small businesses) allow you to issue virtual or physical corporate cards to employees.
  - This eliminates personal spending and the need for reimbursements.
  - You can set specific spending limits, rules, and merchant categories for each card.
  - Receipts can be captured automatically via email or photo, and data flows directly into your accounting software, eliminating manual entry.

Problem 3: Manual Invoice and Bill Entry

Martha manually opens every vendor invoice from her email, reads it, and enters the data into QuickBooks a slow, error-prone process.

Framework Application:
- Systematize: This is a perfect task for automation because it can't be eliminated or easily reassigned.
Solution & Tool (Advanced AI):
- Build an Automated Workflow with make.com:
  - Using a low-code tool like make.com, you can build a "digital employee" to handle this entire process.
  - The Workflow:
    1. All vendor invoices are sent to a dedicated email address (e.g., payables@yourcompany.com).
    2. The automation monitors this inbox.
    3. When an invoice arrives, AI reads the attachment and extracts key data (vendor name, invoice number, amount, tax, etc.).
    4. It checks your accounting software (e.g., QuickBooks) to see if the vendor exists and if there is a matching purchase order.
    5. If everything matches, it creates the bill automatically.
    6. If there's a discrepancy (e.g., the invoice amount is higher than the purchase order), it automatically flags it and notifies Martha with the specific details.
- Tracking the Automation: This workflow creates a perfect audit log. You can feed this log into Notebook LM (from Problem 1) so anyone in the company can ask, "Did we receive the latest bill from BC Hydro?" and get an instant, accurate answer without bothering Martha.

Final Takeaways

Start with "Why": Be crystal clear about the problem you are trying to solve before you look for a tool.
Process F.I.R.S.T.: Always analyze and improve your process before applying technology. Remember, not every problem needs an AI solution.
Good Data is Not an Accident: Clean, well-structured data is the foundation of any successful AI or automation project. Implement systems (like Float) that create this clean data automatically.

Webinar Q&A Session

Q1: I'm overwhelmed by all the different AI tools available. Where should I start?

A: Forget about all the new, flashy tools. Start with what you already have and are likely already paying for: Google Gemini (if you use Google Workspace) or Microsoft Copilot (if you use Microsoft 365). These tools can handle 70-80% of your basic needs, like writing documentation and answering questions, without any extra cost or managing new subscriptions. Only seek out specialized tools or developers when you have a more advanced, specific process you want to automate.

Q2: How can AI help me stay on top of changing government regulations?

A: You can create a system to get customized alerts. Many government websites offer email subscriptions for updates.

Subscribe to the updates so you receive an email whenever a regulation changes.
Set up an automation (using a tool like make.com) that watches that inbox.
When a new email arrives, the automation sends the text to an AI.
You can give the AI a prompt like, "Read this regulatory update. My business is [describe your business]. Tell me if this change impacts me and explain how."

This creates a personalized watchdog that filters out the noise and only alerts you to what matters.

Q3: Can AI take my financial books and prepare all the necessary tax filings automatically?

A: It is very possible to get close to full automation, but it must be done one step at a time. Because accounting is highly structured, it's a great candidate for AI. The process would be:

First, focus on automating all the data entry and reconciliation to ensure your data is perfectly clean.
Once your data is reliable, you can build workflows to prepare standardized filings (like GST).
The final step of a human double-checking and submitting the payment will likely remain, but you can automate almost everything leading up to it.

Q4: When you upload a policy document, like a Word or PDF file, to the AI chat, does it transfer correctly?

A: It depends on the ecosystem you're using.

In Google Workspace, it's best to first convert your document into a Google Doc before adding it to NotebookLM. This is a very simple process.
In the Microsoft ecosystem, Copilot Notebook should be able to handle Word and PDF documents natively, as it's designed to work within that system. The key is to use the format that is native to the platform you're working in.

Q5: Can a user accidentally change a policy document through the AI chat? How do you maintain security?

A: That's a very important point. The AI chat tools like NotebookLM are in read-only mode. They can only read the information from the documents you provide; they have no ability to write or make changes back to the source file.

To ensure security, you control the permissions on the document itself. In Google Drive or Microsoft SharePoint, you can set the policy file's permissions so that general employees have "View Only" access. The AI will respect those permissions.

Tools Mentioned

Float Financial (Spend Management)
Get a complimentary upgrade to the Professional Plan for 12 months when you sign up with our link. https://get.floatcard.com/SolopilotA2
Make .com (Automation Platform)
The powerful low-code tool for building automated workflows, from simple tasks to complex systems.
https://www.make.com/en/register?pc=scalewithekko