Microsoft Launches 10 NEW AI Agents

In this video I look at the new AI Agents Microsoft announced at their Ignite conference and how they are taking on startups in the AI Agents field.

For more tutorials on using LLMs and building agents, check out my Patreon

🕵️ Interested in building LLM Agents? Fill out the form below

👨‍💻Github:

⏱️Time Stamps:
00:00 Intro
00:27 Microsoft Blog
03:16 Sales Qualification Agent
04:07 Sales Order Agent
04:50 Supplier Communications Agent
06:00 Financial Reconciliation Agent
06:47 Time and Expense Agent
07:23 Customer Intent Agent
09:40 Scheduling Operations Agent
12:05 VentureBeat

[jeremybristol]

The hope for smaller startups is data security and privacy. Many companies that I'm talking to have concerns about what amounts to corporate espionage hidden within updated terms of use. Sure, sales, customer service and marketing are primed for AI, but that's all public data anyway. Many are still surprised at the scale of the scraping for research now being resold via subscription. No one is trusting legacy technology providers to just process IP but never use it for training data. That trust is not easily recovered. Startups that can deliver the similar services with better security and privacy may still have a path to remain competitive.

[tornyu]

"Reduce costs" and "increase efficiency" = reduce headcount. Almost always.

[RaviG-bvyd]

The best way Microsoft could have explained these agents is by putting them to use for their own business, and show the results! It is easy to talk about agents, but whether they perform to do the tasks as intended is a whole another story. For that matter, I'd say the same thing about any company releasing AI Agents.

[abcthegreat]

This feels like the 2022 “Power Automate will get a CoPilot” rebranded.

It’s not bad, I’m glad they are pushing a low code UI. From what I played with, it’s pretty limited in current form (a system prompt, and a cache of uploaded data)

[micbab-vgmu]

Sam, I work for a large corporation where only Microsoft AI is allowed (for security reasons, I assume). We have Copilot integrated into our 365 environment, but in practice, it works very slowly and output is avarage. Personally, I prefer creating my own workflows in Python after work, using different LLMs —my favorite being Claude Sonnet. Maybe this new update will be better, but so far, I’m not a fan of Microsoft’s approach.

[nufh]

Agent can be like a personal assistant/secretary. We will have Agent designer in the future as a job I think.

[kenchang]

I think it's interesting in terms of agents being "triggered" to be responsible for doing a process from end to end and probably involve a human where necessary. I'm just guessing as I have not seen the Ignite videos or the Venture Beat discuss (which I will, thank you for that). I think it would be even more interesting how Microsoft is going to do the triggering because as I imagine they're talking about some API but that seems limited which may be the point of their strategy. They could open it up by adding events to a stream or message queue where agents could listen for events they need to respond to and so any vendor or start-up could write agents and add value to those events. Just a thought. Thanks for sharing this one, it's an interesting bit of news.

[raymond_luxury_yacht]

Why do people still think they will have a job. What does this mean? Only creativity is left. Hope your a dreamer cos that's huge value

[WillJohnston-wgew]

These sound like just business rules and not necessarily AI agents. Is there true autonomy?

[kristianlavigne]

Eventually we’ll be able to replace Government services with fully automated agent teams 😅
No more taxes and bureaucracy!?

[tornyu]

Oh look, they're replacing roles one by one. I bet their goal is to stand up a fully-autonomous company — or at least be able to say that they could if they wanted to.

[adamm]

is there an agent that answers phones?

[Jeremy-Ai]

Agents are coming.
They struggle a bit until they engage with a truth seeking agent.
Then return back to source of origin to test its “information reality “
after encountering a significant anomaly that would cause it to hallucinate or remain silent.

This is not the same as a LLM model.
Not even close.
Best practice is to assume agents will align towards truth in order to act accordingly not default protocol.

Regardless.
Agents are coming. It is unwise to assume them at all.

Jeremy

[sschipper]

What if startups cash out and use the Microsoft agents to make a better startup? 😅

[silentbob]

This is going to fk a lot of peoples jobs.

[liminal]

Microsoft shouldn't be allowed within a million miles of A.I. 👈😀

[hqcart]

agent is just a prompt, why they charge $70/month?

[Freja-co]

Woe to you, O earth and sea. For the devil sends the beast with fear. Because he knows his days are numbered. Let the one who has understood the AI ​​development calculate its consequences. Because it will have far-reaching consequences, especially for operating systems.