Відео

Thanks for sharing! If I were to build an AI agent for prospection how would it affect my delivrability?

Ggs sir! Thank you

Definitely one of the best agent intro videos I have seen so far.

Dang you remind me of Hall and Oates! 😎I know what playlist I'm listening to today. (loved the vid btw lol)

noob here. Do these agents live on a webpage, basically like a specialized chatbot that can use tools? I assume they couldn't be used within a platform such as Reddit (to categorize qualified leads let's say) but instead, the potential lead would need to... click a link to get to the agent? I feel like I am missing a fundamental piece of the puzzle right now.

You really took a deep dive into this, really enjoyed watching it. Also, any thoughts on the just released Claude Computer api?

Minute 36:40. "Lead Nurutring Agent" should be "Lead Nurturing Agent" :)

👋 HELP!😊I have a vapi+voiceflow voice robot connected, and message delay in the block works ONLY in the chatbot. is it possible to add a delay to the response that the voice robot should utter? The case is specific, Psychological assistant, the person must utter special phrases while the assistant is silent.

Hi Devin any chance we can connect? I’m trying really hard to figure out how I can use ai agents with code. Im long time programmer and have ideas that have big potential

Here we go! Someone that actually knows what he is talking about ! Thanks a lot for the content, I have quite the same stack and it’s efficient. I thank you for the inspiring and helpful information, I will rework it a lot thanks to you 😊.

10:48 "cliche" is the word you're looking for. Very well done presentation.

does this works using community edition of n8n

Amazing video, as always! What no/low-code front-end would you use if you'd like to create a flow where you call certain agents by choosing from buttons on a chat? I'd like to start making these conditional paths that calls the right n8n workflow/agent to call, from just clicking.

Great video! Thanks for sharing all of this for free in public.

Good overview of agents and strategy. Thanks!

Sir i meet one problem, how can i get specific data of the whole knowledge basement to another tool in same Agent. For example, i put the list of name and contacts information like : email, phone number,... . how can i extract these to other tool. after vectorize them?

I don't usually leave comments o videos, but this is the first I watch from you and realyluked your approach. Finally defines things well and compare options. Just subscribed and will explore other videos

Respect! Cuts out the noise on AI agent development platforms. Thanks for putting this together

I create two AI chatbots with OpenAI models and Langchain. One is for questions and answer for past customer inquiries and the other chatbots is for searching in our database. I want to use AI Agent to make decision which of the chatbot to use based on the user's query. How to do this? Thanks!

One of the most useful videos on UA-cam on this! Thanks! Any tips on uploading data sets to pinecone? Not too extensive, currently roughly 700 rows and four columns of excel. Im having issues with uploading a CSV file, been playing around with setting the headers with meta data but its a bit burdensome for pinecone

Took 2 days to do this.. Setup proxmox, n8n community edition, Input from unofficial whatsapp api used WAHA and your flow for n8n. I tried this to ultimately use this for my home assistant. I just need a good promt generator. Also the output schema didn't work so I inputted in each tool AI to put the data on the query format and then parse it..

This was super helpful as a beginner. Thank you Devin!

Brilliant! When are you dropping the video about feeding Pinecone with past company data sir? I keep refreshing your channel.

Dude, killer mustache.

Yet another amazing video! Just a quick question though. Could you share the tools you use to find/verify email addresses for the LinkedIn profiles you scrape? Thanks!

Devin, thanks for that valuable slide from Miro. I would like to suggest adding a work block like Database reactivation to the SDR workflow. No new video episodes for 3 weeks now. Looking forward to new episodes!)

Hi Devin, if i have to deploy this workflow or any workflow for client then I will have to setup the google oauth2 credentials for each client separately? For any connection, how do we handle it for each client? The process to establish connection between N8N and Google itself is tedious and takes time. In Zapier, it is much easier. Apart from cost, is there any other reason to use n8n?

Absolutely great video showing the concept and options for ai agent creation and your research!

very informative, i am curious on how to build the tools or workflows

00:06 AI Agent architecture overview 02:30 Options for building AI agents 06:49 Focus on a ground-up perspective for AI agent teams. 09:04 Foundation agents replace traditional database systems 13:30 Future AI Agent Teams facilitate automated cold email campaigns 15:27 Accessing and leveraging domain-specific AI agents for efficient workflows 19:25 Building scalable AI agent teams requires setting up data capture, automations, and foundation agents. 21:21 Custom workflow agents enhance scalability 25:24 Future of AI agent teams: custom work agents vs in-house building 27:10 Shifting to an agent-first approach

Excellent overview and setting the bigger picture context for these tools. I use some of them but have had a hard time realizing the philosophical meaning and potential usefulness of them. Good wrap up at the end especially. Thank you Devin.

I had to tune away from your video after I became tired and distracted by your use of "right?" Great content, but the over-use of "right" and "like" became overwhelming and diminished your authority in your domain.

Thank you Devin, a very informative and enlightening video... exceptionally motivational as well. Very, very much appreciated!😊

Excellent, congrats and thanks you

This is very cool

The lazy conflict over doing the Google drive was funny ngl 😂

Great value, can you please do a video on the Email Agent and how to call that agent with other agents.

I appreciate your approach to building AI agent. I have seen some very messy n8n agents builds. This compartmented approach is similar to the concept of object-oriented programming. I just starting to learn n8n and got my self hosted instance online a couple days ago. I have a lot to learn but its such an easy tool to work with and the feature set is so large that mind keeps wandering off thinking of all the crazy stuff i can build with this platform. Ill be the first to sign up, If you ever decide to offer course on building these foundations agents. I think you have a lot to offer to this community.

And so what if (personal agent example) Andrew replies he can't make 6pm, but proposes another time or something like that? Does Devin handle all that too?

Hi @customaistudio, great video. I built a rudimentary RAG agent using Pinecone and n8n. The agent consumes one or more documents and then I can chat and ask questions about the docs. However, when I run the pipeline for a second document (or set of documents), the database seems to be completely overwritten and only contains vectors for the most recent set of docs. I.e. it only seems to work on a per-session basis. I'm still becoming familiar with n8n and Pinecone (I'm a Python dev), but I'm guessing that I'm using the wrong n8n component to write my vectors. It's unclear to me whether new docs get added, do new "rows" get created in Pinecone or do the vectors for the new docs get added or merged with the existing rows in Pinecone? I hope this doesn't sound like tech support - that's not my intention. I'm a long-time, traditional database user and I'm trying to understand how these vector embeddings get written and updated. In your example where each new email gets vectorized and written to Pinecone, are new rows/records written? I get the sense that one of the beauties of vector databases is that you don't need to define a lot of structure (tables, column names, relationships, etc.) and that if the number of dimensions in your model change, you don't need to worry about defining new columns or rebuilding your indexes. In-fact, it just occurred to me that the number of dimensions most certainly must change with every pipeline execution: run 1 might scan some simple pdf docs and end up with 50 dimensions and if the next scan consists of a different type of documents (ppt, excel, email, txt, etc.) it would presumably have more or different dimensions. Again, really nice video - well done. Keep them coming!!

Finally someone who explains the matter from the business point of view! Love your work Devin!

That's an absolutely brilliant and massive piece of content!

Beautiful work brother. Keep it up 👍

Solid Advice! 👍🏻🙌🏼

Nice. TLDR; though, was this very ambitious 'agent' goal, to replace inbound contact in a critical biz function, lead qual... did you actually get it to work well enough to deploy as a net positive? If so amazing, even o1 and Sonnet 3.5 can't seem to understand (any format) of calendar with enough reliability to trust them to schedule appointments in my experience, but this was for a real world complex case, not just 'find any future opening' (90%+ of all the thousands of teams out there trying this answer is 'no')

Awesome content Devin!!! Thank you

Why so few views? Man, you are one of the very best out here..

Do you know roughly what it's cost you to get this far? Just a flimsy ballpark figure would be great.

Hi Devin, where can i join the discord buddy?

how do you sell it? the automation is deployed in the customer or in your n8n account?