Pennies from Redmond

In this issue: Thinking about an autonomous agent for War and Peace; a concise statement of The Low-Code Builder's Paradox, which explains why paradise on earth does not exist; and my transparent attempt to generate management twaddle from classic literature.

War and Peace: Chatbot Edition

About five weeks ago I started re-reading War and Peace.

I read the novel for the first time twenty years ago for a Harvard Extension School class titled Russian Imperial Masterworks. But this time, instead of speed-reading to keep up with an ambitious courseload, I’m taking it nice and slow – the way Tolstoy intended.

Slow is better. One chapter, with breakfast, ten minutes per day. The news can wait.

For the first few chapters, I kept track of all the Annas, Counts and Countesses, Princes and Princesses on a piece of paper. Tolstoy makes a puzzle out of the rapid character introductions. Maybe that’s a taste of the whirlwind of society, being thrown into a party where you’re meeting people one after another, and it’s the essence of diplomacy to remember who’s who.

I’m in Book Two now and the cannons have started firing, and in the era of generative AI, I can ask questions. For example, what’s a limber? What’s a picket rope? I am quite certain that when I came across those words twenty years ago, I skimmed past without even noticing, let alone researching methods of placing an artillery battery.

It also occurred to me that with Copilot Studio, I can build an agent (“LevBot”) to manage a virtual book group, through which anyone can sign up to receive a daily chapter with suggested discussion topics.

In the space of an afternoon, I trained LevBot to:

• Retrieve passages from the novel. (but only after giving assurances that it’s in the public domain)
• Answer questions about the text. (LevBot sometimes mixes up names – needs improvement)
• Retrieves specific information from a spreadsheet with book club members. (e.g. “What’s the latest chapter I’ve read?”)
• Count the number of chapters for each of 17 books and 2 epilogues, and in total. (It nailed the total of 365 but missed the count for one of the books)
• Generate sweaty LinkedIn content based on episodes in the text. (See below: “Key Aspects of Bagration’s Management Style”)
• Send daily emails via Outlook.

It was fun to discover the coincidence of 365 chapters (361 in the original, so perhaps not intentional) in the Maude translation at Project Gutenberg. This led me to A Year of War and Peace on Reddit – an excellent resource if you want to read the novel in daily portions without waiting for LevBot.

LevBot has potential, but I’m not ready to launch unless and until I get it working exactly how I envision it. Plus, I need a way to pay for it (see below).

The Low-Code Builder’s Paradox

Copilot Studio has come a long way since I first started using it way back in December 2023, when it was still called Power Virtual Agents.

Now, with the new-and-improved Copilot Studio, you can build a highly functional agent almost as quickly as you can come up with an idea.

I offer that statement with some caveats.

Even for deploying the simplest use cases, such as asking questions about a library of unstructured documents, you’ll still need to understand the moving pieces: Power Platform, Dataverse, and SharePoint.

Next, to combine unstructured and structured data, you’ll need to configure knowledge sources appropriately so that your questions have a better chance of getting the right answers.

If you build a workflow with Actions and Triggers, it helps to understand Power Automate.

But those are all surmountable technical and operational hurdles, becoming easier by the month.

The bigger challenge is making the business case.

Until last month, the Copilot Studio was only available through a $200/month subscription, for 25,000 messages per month.

Now, there’s a pay-as-you-go option for $0.01 per message.

From the pricing page:

“Messages” are a measure of the time and effort required for your agent to retrieve information and respond to your prompts and any actions that the agent takes, based on custom triggers and skills. The number of messages decremented for each response or action is dependent on the complexity of the task completed by the agent.

Suppose that LevBot merely retrieves the latest chapter of War and Peace ($0.01); adds generative content ($0.01 to $0.03, depending on complexity); and emails the result to a single list of email addresses ($0.01). That’s a constant price of a nickel per day, less than $20 per year.

What if we want subscribers to receive chapters at their own pace, generating content based on their own interests? Here, we’re looking at linear pricing, let’s say $0.05 per user per day; for 100 users, that’s $5 per day or $1,825 per year.

At the breakeven point of 132 users, we hit the monthly 20,000-message mark, and so it makes sense to switch to the $200 per month plan for the extra 5,000 messages, resulting in an annual cost of $2,400 per year.

And what if we want to allow subscribers to chat directly with LevBot? Prepare for a multiple of the cost.

And what if we want to replicate the concept with FyodorBot and PushkinBot and GogolBot? Another multiple.

And what if we ask LevBot to connect subscribers with similar interests, and moderate weekly group discussions with LevBot clones? Now we’re really running the meter.

It’s gentle fun when we’re thinking small.

When we think big, the kopeks add up.

This illustrates The Low-Code Builder’s Paradox.

The easier it is to build, the more it costs. The more it costs, the smaller you think. The smaller you think, the less you build.

Conclusion:

If you’re thinking big, do it the hard way. Or low-code prototype, pro-code in production.?

Key Aspects of Bagration's Management Style

Is this worth a red cent?

Prince Pyotr Bagration's management style during the first encounter with the French in?War and Peace?is characterized by his calm demeanor, strategic foresight, and ability to inspire confidence in his troops. Here are some key aspects of his leadership as depicted in the novel:

1. Calm Under Pressure: Bagration remains composed even in the face of intense pressure. His calmness helps to maintain order and morale among his troops, which is crucial during the chaos of battle.
2. Strategic Foresight: Bagration demonstrates a keen understanding of the battlefield and the movements of the enemy. He makes decisions that are not only reactive but also proactive, anticipating the actions of the French forces.
3. Leading by Example: Bagration is often seen at the front lines, directly engaging with his soldiers. This hands-on approach boosts the morale of his men and earns their respect. His presence on the battlefield serves as a powerful motivator.
4. Delegation and Trust: He effectively delegates tasks to his subordinates, showing trust in their abilities. This delegation allows for more efficient management of the troops and ensures that different aspects of the battle are handled by competent officers.
5. Inspirational Leadership: Bagration's leadership style is inspirational. He communicates clearly and confidently, which helps to instill a sense of purpose and determination in his soldiers.?

Dear reader,

Would you be interested in reading War and Peace in 2025 through a daily email, including an AI-generated glossary of key terms, historical background, and links for further reading?

Does that sound fun to you?

What if we throw in some wacky AI-generated art and other community-generated content?

Would you sign up for $10 per month? $5 per month? Or a one-time flat fee?

Would you sign up if it were free with purchase of the book, or free because it's supported by ads or donations or shadowy financiers with a Tolstoyan agenda?

Would someone have to pay you to read it?

Let me know.