WEBVTT 00:00:00.000 --> 00:00:01.919 Okay, so I want you to just imagine something 00:00:01.919 --> 00:00:03.859 for a second. Right. Imagine your automation 00:00:03.859 --> 00:00:07.280 workflows weren't just these silent scripts running 00:00:07.280 --> 00:00:08.900 in the background. Right, not just invisible 00:00:08.900 --> 00:00:11.839 code doing its thing. Exactly. What if they were 00:00:11.839 --> 00:00:14.179 conversational teammates? Teammates you could 00:00:14.179 --> 00:00:16.730 actually ask questions to in real time. Like 00:00:16.730 --> 00:00:18.910 a coworker who, you know, already knows all your 00:00:18.910 --> 00:00:21.629 data and can get things done instantly. This 00:00:21.629 --> 00:00:23.670 is really the fundamental shift that arrived 00:00:23.670 --> 00:00:27.449 in late 2025 with Ann Ann's chat hub. It moves 00:00:27.449 --> 00:00:29.710 automation from that old set and forget model 00:00:29.710 --> 00:00:34.170 to live steering. Okay, let's unpack this. We're 00:00:34.170 --> 00:00:37.729 diving into Max Ann's breakdown. of the NN and 00:00:37.729 --> 00:00:40.009 chat hub. Yeah. And it's not just about a chat 00:00:40.009 --> 00:00:43.289 interface. It's really about how this tool becomes 00:00:43.289 --> 00:00:46.270 a unified command center for delegating tasks 00:00:46.270 --> 00:00:48.670 and managing a bunch of different AI models. 00:00:48.789 --> 00:00:50.969 All in one place. Our mission today is to understand 00:00:50.969 --> 00:00:53.750 the three core features, see why this really 00:00:53.750 --> 00:00:56.469 changes complex automation, and then get into 00:00:56.469 --> 00:00:58.670 the practical steps for setting up agents that 00:00:58.670 --> 00:01:01.189 can analyze data and support customers. So what 00:01:01.189 --> 00:01:03.109 does this all mean for you? It means your system's 00:01:03.109 --> 00:01:04.890 just got a voice and we're going to learn how 00:01:04.890 --> 00:01:07.920 to talk back. And when the chat hop was released 00:01:07.920 --> 00:01:12.939 on December 15th, 2025, it was it was way more 00:01:12.939 --> 00:01:15.140 than just a chat window. It really became this 00:01:15.140 --> 00:01:18.040 centralized AI command center. Right. For talking 00:01:18.040 --> 00:01:21.019 to your entire infrastructure, not just one AI. 00:01:21.260 --> 00:01:23.280 Exactly. So the simplest way to think about it 00:01:23.280 --> 00:01:27.689 is this. It's like having GPT -5, Claude, and 00:01:27.689 --> 00:01:30.750 your own custom AI agents all in one spot. And 00:01:30.750 --> 00:01:33.609 crucially, they're already connected to all your 00:01:33.609 --> 00:01:36.489 automation tools, your data. That's the key part. 00:01:36.890 --> 00:01:38.670 That connectivity is the technical difference, 00:01:38.829 --> 00:01:40.590 isn't it? It is. It's what separates it from 00:01:40.590 --> 00:01:42.969 just, you know, using ChatGPT directly. A direct 00:01:42.969 --> 00:01:45.909 chat model is great for thinking, for brainstorming. 00:01:45.969 --> 00:01:47.790 But it's sealed off from the world. Totally. 00:01:47.909 --> 00:01:50.450 It can write an email, but it can't send the 00:01:50.450 --> 00:01:52.709 email from your account. It can't query your 00:01:52.709 --> 00:01:55.049 live database. It definitely can't run a multi 00:01:55.049 --> 00:01:57.590 -step workflow. Before this, automation always 00:01:57.590 --> 00:01:59.590 felt like a black box. You build something, hit 00:01:59.590 --> 00:02:02.500 run, and just hope for the best. And if it failed 00:02:02.500 --> 00:02:04.900 three steps in, you wouldn't know until later 00:02:04.900 --> 00:02:07.700 when you're digging through logs. It was so passive. 00:02:07.939 --> 00:02:10.620 And that's the critical insight here, the paradigm 00:02:10.620 --> 00:02:13.199 shift. It's a move to what they call collaborative 00:02:13.199 --> 00:02:15.759 automation. You're not just sending a request 00:02:15.759 --> 00:02:18.340 into the void anymore. You're treating the workflow 00:02:18.340 --> 00:02:20.919 like an active teammate. You're asking it in 00:02:20.919 --> 00:02:23.780 plain English. Hey, what's happening right now? 00:02:23.860 --> 00:02:26.580 Why did that last step fail? What data do I need 00:02:26.580 --> 00:02:29.960 next? It's that live steering capability. Think 00:02:29.960 --> 00:02:32.120 about it. If your daily sales report workflow 00:02:32.120 --> 00:02:35.500 fails on step four. The CRM credential node, 00:02:35.659 --> 00:02:38.099 always. Right. Instead of digging through logs, 00:02:38.219 --> 00:02:40.319 you just ask the hub, why did step four fail? 00:02:40.719 --> 00:02:43.139 And it replies instantly, credential expired. 00:02:43.789 --> 00:02:46.090 Pause the process. Please update the key. It's 00:02:46.090 --> 00:02:48.069 the difference between sending an email and getting 00:02:48.069 --> 00:02:49.810 an instant reply from someone sitting right next 00:02:49.810 --> 00:02:51.610 to you. It just removes all the guesswork. So 00:02:51.610 --> 00:02:53.949 connecting this to the bigger picture then, why 00:02:53.949 --> 00:02:56.150 is collaborative automation so much better than 00:02:56.150 --> 00:02:58.909 the old set and forget way, especially with complex 00:02:58.909 --> 00:03:01.849 data? It allows for live steering and debugging 00:03:01.849 --> 00:03:04.889 of systems, turning passive scripts into queried 00:03:04.889 --> 00:03:07.210 resources. Okay, so if the concept of talking 00:03:07.210 --> 00:03:10.069 to your systems is clear, what did they actually 00:03:10.069 --> 00:03:13.909 build to make this happen? Let's get into the 00:03:13.909 --> 00:03:17.009 three core features. The first one is the multi 00:03:17.009 --> 00:03:20.090 -model chat interface. This is a huge efficiency 00:03:20.090 --> 00:03:22.210 boost. I mean, it solves that problem of trying 00:03:22.210 --> 00:03:24.949 to force one AI model to do everything badly. 00:03:25.449 --> 00:03:27.889 This interface lets you switch between large 00:03:27.889 --> 00:03:29.689 language models right in the middle of a thread. 00:03:29.870 --> 00:03:31.849 You don't lose your context. You don't have to 00:03:31.849 --> 00:03:33.909 start over. It's like having a small specialized 00:03:33.909 --> 00:03:37.300 team on call. You can start with GPT -5 for reasoning. 00:03:37.400 --> 00:03:39.139 Yeah, for the big picture stuff. But then you 00:03:39.139 --> 00:03:41.020 realize you need to analyze a huge document, 00:03:41.099 --> 00:03:43.360 like 50 ,000 words, so you just... switch over 00:03:43.360 --> 00:03:45.840 to Claude, which is known for handling that huge 00:03:45.840 --> 00:03:48.120 context. Or you need to check a stock price right 00:03:48.120 --> 00:03:50.379 now, so you flip to Gemini or a model with a 00:03:50.379 --> 00:03:52.759 live web search tool, you get real -time info. 00:03:52.939 --> 00:03:54.879 No more knowledge cut -offs. What's fascinating 00:03:54.879 --> 00:03:57.400 here is the ability to use these specialized 00:03:57.400 --> 00:04:00.580 models, but how does ensuring consistency across 00:04:00.580 --> 00:04:03.759 those multi -model interactions affect the outcome? 00:04:04.000 --> 00:04:06.780 You use each model where it shines, leading to 00:04:06.780 --> 00:04:10.479 higher quality and more reliable analysis. Okay, 00:04:10.539 --> 00:04:13.300 so the second feature is custom personal agents. 00:04:13.939 --> 00:04:17.480 Think of these as the lightweight players on 00:04:17.480 --> 00:04:20.220 the team. They're simple agents with custom instructions 00:04:20.220 --> 00:04:23.800 and very specific limited tool access. And you 00:04:23.800 --> 00:04:26.300 can create them without building a whole complex 00:04:26.300 --> 00:04:28.459 workflow. So for someone just starting, this 00:04:28.459 --> 00:04:30.740 is probably the easiest way in. Absolutely. You 00:04:30.740 --> 00:04:32.600 could make a content editor agent. The instruction 00:04:32.600 --> 00:04:35.699 is just always use cloud for consistency. And 00:04:35.699 --> 00:04:38.639 that's it. Simple. But the real power, the heavy 00:04:38.639 --> 00:04:41.339 machinery, that's the third feature. workflow 00:04:41.339 --> 00:04:44.079 agents. This is the game changer. This is where 00:04:44.079 --> 00:04:47.420 you connect your big, complex NN workflows directly 00:04:47.420 --> 00:04:49.939 to the chat using the new chat trigger node. 00:04:50.120 --> 00:04:52.600 So you're merging conversation with actual execution. 00:04:52.899 --> 00:04:55.259 That's it. Instead of manually triggering a data 00:04:55.259 --> 00:04:57.540 analysis workflow, you just chat with it. You 00:04:57.540 --> 00:05:00.120 say, analyze last month's sales data. And the 00:05:00.120 --> 00:05:01.800 workflow just runs in the background, it connects 00:05:01.800 --> 00:05:03.800 to the database, does the analysis. And then 00:05:03.800 --> 00:05:05.480 replies to you right there in the chat window, 00:05:05.680 --> 00:05:08.720 right away your next question. Whoa. Imagine 00:05:08.720 --> 00:05:12.079 scaling this capability to manage a billion queries 00:05:12.079 --> 00:05:15.019 across diverse model architectures. That's a 00:05:15.019 --> 00:05:19.040 serious leap forward. Sponsor. So moving into 00:05:19.040 --> 00:05:21.279 implementation, there are a few really crucial 00:05:21.279 --> 00:05:24.060 details you have to get right for this to actually 00:05:24.060 --> 00:05:28.259 work. You must use the newest version of the 00:05:28.259 --> 00:05:30.579 chat trigger node. That is the number one pitfall 00:05:30.579 --> 00:05:32.759 we see. People are using the old legacy node. 00:05:32.980 --> 00:05:34.860 It looks similar, right? Very similar. But the 00:05:34.860 --> 00:05:37.360 new one is what handles that back and forth conversational 00:05:37.360 --> 00:05:39.300 flow. If you're stuck, check that node version 00:05:39.300 --> 00:05:42.199 first. Okay. And the other key thing is enabling 00:05:42.199 --> 00:05:45.439 streaming on the AI agent node. Jargon alert. 00:05:45.759 --> 00:05:48.360 What's streaming? Streaming just means the AI's 00:05:48.360 --> 00:05:51.000 answer appears in real time, word by word, like 00:05:51.000 --> 00:05:53.259 someone typing. Not a big block of text after 00:05:53.259 --> 00:05:55.149 a long wait. And that's important because it 00:05:55.149 --> 00:05:57.350 feels more like a conversation. It feels collaborative. 00:05:57.709 --> 00:05:59.810 Exactly. It's about that feeling of interaction. 00:06:00.129 --> 00:06:02.610 And what about knowledge cutoffs? I mean, most 00:06:02.610 --> 00:06:04.810 models don't know what happened last week. Right. 00:06:04.949 --> 00:06:06.949 So you have to connect external tools. You can 00:06:06.949 --> 00:06:09.589 plug in something like Jaina AI for real -time 00:06:09.589 --> 00:06:12.589 web search or connect to your own knowledge bases. 00:06:12.949 --> 00:06:15.209 It keeps the agent current. So now for the power 00:06:15.209 --> 00:06:19.079 move. Turning an existing complex workflow into 00:06:19.079 --> 00:06:22.800 an agent your team can talk to. It's surprisingly 00:06:22.800 --> 00:06:24.860 simple. You just go into the chat trigger and 00:06:24.860 --> 00:06:27.660 you turn on one setting. Make available an 8 00:06:27.660 --> 00:06:29.620 a .m. chat. And you have to give it a clear descriptive 00:06:29.620 --> 00:06:32.720 name, not workflow three. Please don't. Something 00:06:32.720 --> 00:06:35.180 like sales data analyzer. So your team knows 00:06:35.180 --> 00:06:37.600 exactly what it does. You know, I have to admit, 00:06:37.740 --> 00:06:40.639 I still wrestle with prompt drift myself sometimes. 00:06:40.939 --> 00:06:43.060 Oh, everyone does. Especially when you're dealing 00:06:43.060 --> 00:06:45.779 with these little setup details. Like just remembering 00:06:45.779 --> 00:06:48.480 to replace that old chat trigger node. It sounds 00:06:48.480 --> 00:06:51.120 so obvious. But it's an easy mistake to make 00:06:51.120 --> 00:06:52.899 when you're moving fast and just trying to get 00:06:52.899 --> 00:06:54.819 it working. It really happens to everyone. But 00:06:54.819 --> 00:06:57.000 when you get those details right, the use cases 00:06:57.000 --> 00:07:00.360 are just, they're transformative. Like an interactive 00:07:00.360 --> 00:07:03.120 customer support bot. Perfect example. Your support 00:07:03.120 --> 00:07:06.240 team can just ask, check status for order, hashtag 00:07:06.240 --> 00:07:09.829 12345. And they don't have to log into three 00:07:09.829 --> 00:07:11.790 different systems to get that one answer. Right. 00:07:11.850 --> 00:07:15.290 It pulls from the CRM, inventory, shipping, all 00:07:15.290 --> 00:07:17.310 at once. The psychological win there is huge. 00:07:17.569 --> 00:07:20.149 Or the data analysis assistant. Instead of pulling 00:07:20.149 --> 00:07:22.930 raw data, you just query the database in chat. 00:07:23.189 --> 00:07:25.709 What were our top 10 products last quarter? And 00:07:25.709 --> 00:07:28.689 you get the answer already summarized. So given 00:07:28.689 --> 00:07:30.769 that the workflow agent is the heavy machinery 00:07:30.769 --> 00:07:33.829 here, what's the fastest win a learner can get 00:07:33.829 --> 00:07:35.810 when they're starting out with a custom agent? 00:07:36.199 --> 00:07:38.680 Focus on building one workflow agent that reliably 00:07:38.680 --> 00:07:41.860 answers a single painful question your team asks 00:07:41.860 --> 00:07:45.480 every day. Solve one problem perfectly, then 00:07:45.480 --> 00:07:47.920 you can scale. Now, when we talk about rolling 00:07:47.920 --> 00:07:50.959 this out to a bigger team, especially in a company, 00:07:51.199 --> 00:07:54.399 control and security are everything. And this 00:07:54.399 --> 00:07:56.339 is where the chat user role comes in. This is 00:07:56.339 --> 00:07:59.139 a vital guardrail. It's designed for non -technical 00:07:59.139 --> 00:08:01.660 people, your sales team, your marketers, so they 00:08:01.660 --> 00:08:03.699 can use this power safely. So what do they see? 00:08:04.220 --> 00:08:06.339 They only see the chat screen and the agents 00:08:06.339 --> 00:08:08.259 you've chosen to expose to them. They can't see 00:08:08.259 --> 00:08:10.160 the workflows. They can't see the credentials. 00:08:10.459 --> 00:08:12.660 And can't break the engine. Exactly. It's a perfect 00:08:12.660 --> 00:08:15.120 separation of duties. They get the power without 00:08:15.120 --> 00:08:17.319 the risk. We should probably touch on limitations, 00:08:17.540 --> 00:08:19.860 too. Yes. Remember that personal agents are the 00:08:19.860 --> 00:08:22.000 lightweight ones. They're great for quick text 00:08:22.000 --> 00:08:24.480 tasks, but they can't do the heavy lifting like 00:08:24.480 --> 00:08:26.500 reading through a huge knowledge base. For that, 00:08:26.639 --> 00:08:28.699 you need a RAG, right? Retrieval Augmented Generation. 00:08:29.120 --> 00:08:31.660 Correct. And for proper RAG, great, you must 00:08:31.660 --> 00:08:35.230 use a workflow agent. Why is that? Because RGRI 00:08:35.230 --> 00:08:38.289 needs a vector store node to handle all the indexing 00:08:38.289 --> 00:08:41.090 and retrieval of that big document. A personal 00:08:41.090 --> 00:08:43.490 agent is just too lightweight to manage that 00:08:43.490 --> 00:08:46.029 complexity. So to make sure these are reliable, 00:08:46.210 --> 00:08:48.870 you really have to treat the rollout like a product 00:08:48.870 --> 00:08:51.529 launch. It helps to think of it like you're hiring 00:08:51.529 --> 00:08:54.399 a new team. You need to give them clear names, 00:08:54.539 --> 00:08:57.820 not Agent 1. Call it Customer Order Lookup. And 00:08:57.820 --> 00:08:59.919 write a good description that answers, what does 00:08:59.919 --> 00:09:02.299 this do and when should I use it? And then the 00:09:02.299 --> 00:09:04.820 system prompts. That's like the employee handbook. 00:09:05.000 --> 00:09:07.440 That's where you set the rules, the tone, the 00:09:07.440 --> 00:09:10.840 limits. Things like, always verify customer identity 00:09:10.840 --> 00:09:13.700 before sharing order info. It makes the agent 00:09:13.700 --> 00:09:16.080 dependable. And finally, you have to test it. 00:09:16.100 --> 00:09:17.919 You have to try and break it before you share 00:09:17.919 --> 00:09:21.009 it. Absolutely. Run edge cases. Ask confusing 00:09:21.009 --> 00:09:23.509 questions. If it can handle the stress test, 00:09:23.750 --> 00:09:26.110 it's ready for the team. Okay, so a quick recap 00:09:26.110 --> 00:09:28.990 of the comparison. Direct chat tools like GPT 00:09:28.990 --> 00:09:32.169 are great for thinking. Brainstorming, drafting. 00:09:32.629 --> 00:09:35.330 And fully custom chatbots are incredibly powerful. 00:09:35.590 --> 00:09:39.049 A massive engineering project to build and maintain. 00:09:39.389 --> 00:09:41.450 So ChatHub sits right in that sweet spot in the 00:09:41.450 --> 00:09:44.629 middle. It's where your thinking turns into direct 00:09:44.629 --> 00:09:47.490 automated action. It gives you that custom behavior 00:09:47.490 --> 00:09:50.250 without the insane cost of building a whole system 00:09:50.250 --> 00:09:52.809 from scratch. So if security and cost control 00:09:52.809 --> 00:09:56.049 are paramount, which administrative control should 00:09:56.049 --> 00:09:59.070 an admin prioritize when setting up ChatHub for 00:09:59.070 --> 00:10:01.840 a big team? they should focus on the chat user 00:10:01.840 --> 00:10:04.960 role restrict credential management and enable 00:10:04.960 --> 00:10:07.960 or disable specific potentially expensive ai 00:10:07.960 --> 00:10:10.639 models the essential insight we found here is 00:10:10.639 --> 00:10:12.519 that chat hub is really more than just a feature 00:10:12.519 --> 00:10:16.139 it's a a fundamental paradigm shift toward conversational 00:10:16.139 --> 00:10:18.820 automation It just removes the guesswork. You 00:10:18.820 --> 00:10:21.700 can debug in plain language. You can guide complex 00:10:21.700 --> 00:10:24.580 processes in real time. You stop waiting for 00:10:24.580 --> 00:10:26.440 an output and start engaging with the pipeline. 00:10:26.960 --> 00:10:29.539 The teams that are going to move fastest are 00:10:29.539 --> 00:10:31.779 the ones that upgrade from sending tickets to 00:10:31.779 --> 00:10:34.399 having a live analyst, a system they can talk 00:10:34.399 --> 00:10:36.639 to and make decisions with in the moment. And 00:10:36.639 --> 00:10:38.340 here's where it gets really interesting for me. 00:10:38.740 --> 00:10:42.399 This shift. means automation is no longer this 00:10:42.399 --> 00:10:44.460 invisible engine in the dark. It's an active 00:10:44.460 --> 00:10:46.799 teammate. It's an active teammate ready for you 00:10:46.799 --> 00:10:50.120 to delegate to. For anyone focused on fast, high 00:10:50.120 --> 00:10:53.000 -quality knowledge, this tool means you can now 00:10:53.000 --> 00:10:55.820 interrogate your data pipelines. You can question 00:10:55.820 --> 00:10:57.940 your workflows. You're not just waiting for the 00:10:57.940 --> 00:11:00.120 final report anymore. You're an active participant. 00:11:00.399 --> 00:11:02.460 An active participant in your own business logic. 00:11:02.659 --> 00:11:04.279 So if we accept that the future of automation 00:11:04.279 --> 00:11:06.659 is conversational, the final thought for you 00:11:06.659 --> 00:11:10.299 to consider is this. If your workflows could 00:11:10.299 --> 00:11:12.980 actually talk to each other, if the data analyst 00:11:12.980 --> 00:11:15.639 agent and the content generator agent could hold 00:11:15.639 --> 00:11:19.419 a meeting, what's the first task they would solve 00:11:19.419 --> 00:11:21.259 together? Go build something amazing.