Ep. 331: The Bot Wrangler: How Hamachi Plans to Govern the AI Chaos in Wealth Management

Podcast: Play in new window | Download | Embed

Subscribe: RSS

This episode features a rare “full house” collaboration between four heavyweights of the wealth management technology space who have joined forces to launch Hamachi.

• Eric Clarke — The former founder and CEO of Orion, Eric brings decades of experience in scaling enterprise-grade advisor technology and portfolio management systems.
• Brian McLaughlin — The former founder and CEO of Redtail CRM, Brian is a pioneer in advisor workflow and client relationship management software.
• Mike Wilson — The former CEO of Advisory World, Mike has a deep background in proposal generation and investment analytics.
• Mustapha Baassiri — The CTO of Hamachi and former co-founder of Advisor Software, Mustapha leads the technical vision for Hamachi’s context engineering and AI orchestration.

Six Things Worth Writing Down

1. Hamachi’s mission is to help firms “safely say yes” to AI — By providing enterprise guardrails that handle PII issues and compliance requirements, the platform allows broker-dealers to finally embrace AI capabilities instead of blocking them.
2. The Model Context Protocol (MCP) was the “Aha Moment” — The team realized they could build a sophisticated overlay using MCP to give AI the “skills” to programmatically navigate the fragmented landscape of advisor tools, from CRMs to portfolio systems.
3. Context Engineering beats generic prompting — Rather than relying on a “throw it into ChatGPT” approach, Hamachi uses specialized preprocessing pipelines and narrow-scope agents to ensure high accuracy when dealing with complex financial tables, charts, and figures.
4. The “Bot Wrangler” approach centralizes industry expertise — Hamachi acts as an orchestrator for a “bot store,” allowing advisors to tap into out-of-the-box expertise from asset managers, practice management coaches, and industry thought leaders like FP Pathfinder and Morningstar.
5. “Super Prompts” remove the engineering burden from advisors — To prevent advisors from needing to become prompt engineers, Hamachi uses pre-programmed, curated “super prompts” that are tested and validated to interact perfectly with specific data sources and use cases.
6. Observability is the key to AI compliance — Unlike generic platforms, Hamachi logs every prompt and action for audit purposes and forces agents to provide direct links to source material so that every answer can be verified by the user.

In Their Own Words

“Hamachi is all about allowing advisory firms and broker-dealers to finally say yes to their advisors engaging and using all the capabilities that AI has to bring to the table.”

— Eric Clarke

“If we can do a bot layer that can be expert bots, programmatically trained where they’re built for a specific purpose… it’s not just throw into ChatGPT and ask it a question, hope I get the right answer. It’s more curated than that.”

— Brian McLaughlin

“We look at ourselves as being additive… We are performing governance and logic on top of [CRMs and portfolio systems] to produce things like a daily brief and a household brief.”

— Mike Wilson

“Our goal is definitely to remove the prompt engineering away from the financial advisor. We want to handle and deal with all the complexity.”

— Mustapha Baassiri

“Anytime an agent responds to a question, it’s always required to provide the source material… they can click through and verify for themselves that the data is accurate.”

— Mustapha Baassiri

What We Cover in This Episode

Here at Ezra Group, we’re experts on everything wealthtech, including CRM, portfolio management, trading, rebalancing, performance reporting, just to name a few. When we start working with an RIA or broker dealer, the first thing we do is a  comprehensive tech stack assessment. This provides a top to bottom view of all systems and processes, and it’s a critical part of the firm’s growth plan, since the tech stack is the foundation for building towards the future.

So if you’d like to see your tech stack converted from a liability into an asset you need to run not walk to our website, EzraGroup.com, and click on the golden Contact Us button at the top of the homepage, the experienced team at Ezra Group will conduct a detailed tech stack assessment for you, delivering targeted recommendations that will optimize your existing software platforms. Or we can run an RFP process and help you select and then implement a new solution to help take your firm to the next level. You can schedule a free consultation by going to EzraGroup.com.

A few quick housekeeping tasks before we continue.

• Please subscribe to the show wherever you listen to podcasts so you don’t miss an episode.
• Check out our sponsors, the Invest in Others Charitable Foundation. You can find them at InvestinOthers.org.
• If you are a wealthtech vendor, please register for the Ezra Group WealthTech Integration Score Portal where you can update your integration data in real-time and improve your score.

Now let’s kick this thing off!

Which AI Productivity Tools Are Right for Your Firm?  Ask Ezra Group!

Integrate smarter tools. Streamline workflows.
Boost advisor confidence.

Talk to an Expert →

Podcast Transcript

Craig: I’m excited to introduce our next guests for this episode. We have a collaboration here. We’ve got four guests, the four co-founders of a new startup called Hamachi. We have Eric Clarke, Brian McLaughlin, Mike Wilson, and Mustapha Baassiri, all from Hamachi. Guys, thanks for being here. I appreciate it. This is a full house.

Eric: Thanks for having us. We’re excited, Craig.

Craig: Excellent. We’ve got a lot of questions and a lot of guests. I’m going to go one round robin so everyone can hear your voices. We know who’s who. For the people listening on the podcast, obviously people watching on YouTube will be able to see it. For the podcast, let’s start a quick round. Normally we ask for a 30-second elevator pitch for the product. For Hamachi, let’s start with Eric. Eric Clarke, former founder and CEO of Orion. What’s Hamachi about?

Secure AI Adoption for Enterprises

Eric: Thanks for the question, Craig. At Hamachi, we allow advisory firms and broker-dealers to finally say “yes” to using AI with all the industry guardrails in place to alleviate all the compliance concerns, the concerns around PII, and yet at the same time allow these firms to implement and get all the benefit and efficiencies that AI has to offer. That’s the elevator pitch. It has everything to do with compliance and everything to do with allowing firms to embrace all the capabilities that AI has to bring to the table.

Craig: Very cool. This question is for Brian McLaughlin, former founder and CEO of Redtail. Before we get to the specific features, what’s the specific moment in an advisor’s day that made you say, this is broken, and we need to build a new product around it?

Solving the Advisor Tool Fragmentation

Brian: I want to speak a little bit to myself, but I think we all feel the same way, which is when MCP servers came out. MCP servers, Model Context Protocol, enabled AI to have skills, or to have insights into where to get information that you can programmatically do. You had this background history we all have of watching advisors use so many different tools to achieve their job, or having to use so many different tools to do research, to even communicate with a client, or provide feedback or insights to everything from portfolios to planning to prospecting. When you realize they’re using so many tools, and we have this capability in technology to write overlays that can read and know where to go get contact information from a CRM, where to go get a document, where to go look for different things, it was like an aha moment. If we can do a bot layer that can be expert bots, programmatically trained where they’re built for a specific purpose, they have the regulation pipelines and the custom built tools to make it very highly accurate and insightful. That was the moment about a year ago where we’re like, this would be so cool if we could do X, Y, and Z, but the technology was there. We could do that. We started building. I think that was one of the powerful moments that got us going. There’s been a number of iterations as we played with this and the technology changes so fast. We’ve been keeping up with the changes and making iterations to make it even better. The way we did our pipelines to control compliance regulatory checks, marketing checks, everything’s well built so that it’s not just throw into ChatGPT and ask it a question, hope I get the right answer. It’s more curated than that. I love that part of the feature set.

Craig: Are you sure we can’t just throw things in ChatGPT? Everyone’s—

Brian: Of course you can. There’s a million of those. I just watched the Super Bowl. That’s when we’re airing this is the day after the Super Bowl and we’re recording this.

Craig: We’re doing a little bit of a round robin here. We have four guests. That was Brian. Up for Mike Wilson, former CEO of Advisory World and a co-founder of Hamachi. Mike, you guys describe Hamachi as a communication layer rather than a traditional advisor tool. I’m a little confused. Where does it sit in the tech stack? How does it interact with systems that advisors are already using?

Unlocking Data Through Integrated Governance

Mike: Thanks, Craig. We look at ourselves as being additive. I’ll tell you why we’ve been so anchored on communication is because the advisory business is all about relationships. We see trying to have communications with more and more clients in a highly informed manner. It’s what advisors do day in and day out. we were looking for ways to help that scale as more investors look and seek out advice. When I say we’re additive, we’ve hooked into the CRM and the portfolio management systems that are widely used throughout the industry and looked for ways to unlock the data that’s in those systems. CRMs are data systems. Portfolio management systems run calculations. We are performing governance and logic on top of those to produce things like a daily brief and a household brief. Both of those materials help give advisors an immediate insight into what their book looks like, what their day is going to consist of, who they need to talk to. It does so in a manner that is highly informed in the sense that we’ve got a series of what we refer to as expert bots that come in over the top of Hamachi and say, these are the type of things you should be looking at. Those extra bots might come from the parent RIA, it might come from your broker-dealer, it might be from a practice management coach, it might be from the asset management firms that you know and trust and follow. All of that content comes together on top of our platform and allows advisors to better interact with their clients in the form of communication.

Craig: So it’s sort of a bot wrangler?

Mike: I don’t know if we’re bot wranglers, I guess you could say that, it’s the first I’ve heard it.

Craig: You’re organizing and managing a lot of other bots from other companies, right? Is that what you’re implying here? You’re putting your governance on top. You’re managing all the other bots. This is something I’ve thought about. With every application, I was on a panel at a RIA conference about introducing AI into your RIA. When I first started up, my first thing that I said was, you already have AI in your RIA. Any software you’ve got is launching AI, and they’re not even telling you. They’re just pushing it out. There’s no way to control that. Every one of your vendors has got it. They’re pushing it out. They’re building it on who knows what LLMs, who knows how they’re training it, where is it coming from, where is it going? It seems like having a governance system on top of all that would be valuable.

Mike: Thanks, Craig. Eric mentioned that as the first pillar of our mission statement, of our elevator pitch. That’s all around guardrails and compliance. One of the subcomponents of that that I’m sure you’re familiar with is observability. Every single time a user of Hamachi runs a prompt or does anything within our platform that’s logged and available for audit, and we also go back and check to make sure that any of the answers that were produced by an LLM were accurate. That’s something that Mustapha could go in a lot deeper on, and maybe that’s a topic for another conversation, but it is an important component of what we’re bringing to the table, all this guardrails and governance.

Craig: I just made a note to come back to that, because I want to get to Mustapha, because he hasn’t had a chance yet. The fourth pillar, the fourth guest here is Mustapha Baassiri, who’s the CTO of Hamachi, former co-founder of Advisor Software. Question for you, sir, you guys have emphasized that you’re doing expert-trained agents or expert bots, which I think you’re saying they’re going to replace generic AI prompts. What goes into training these agents and bots and what work is required from advisors or your clients to keep it accurate and compliant?

Context Engineering and the Expert Bot Framework

Mustapha: Thanks, Craig. When it comes to these experts, AI agents, we basically do several things to train these agents and make sure they’re accurate. One of the things we do is when you create an agent, you provide it with source material, the domain you want this agent to be an expert on. You can upload PDFs, point it to an API or several types of data sources. We take care of taking that data and making sure that that agent becomes an expert on that topic or that data that you’re giving it. We do that in a few ways. First, we have our own pre-processing workflow. We take all of this data that could be in many different formats and we pre-process it in a way that optimizes the data to make an LLM understand it deeply. Compare that to when you’re using ChatGPT, you might just attach a file in your conversation and ask ChatGPT about that file. It’ll probably answer you correctly most of the time. But when that file contains complex data, like tables, charts, figures, it might not respond accurately. We have this preprocessing pipeline that we take the data through to make sure it’s understandable by the LLM. Another thing is we’ve engineered all the prompts and the context behind these agents to make sure that these agents are not distracted by any other types of data that they don’t need. It’s all about context engineering. When you have an agent that has a narrow scope, it knows what it’s supposed to know. It’s always fed the right data when it’s trying to answer a question, then it’s likely to answer you correctly, especially if you’re using these frontier models that have reasoning built in, like GPT 5.2 or OPUS 4.6. We always make sure we’re using the latest reasoning models, and we’re always making sure that the context engineering is there and it’s optimized. Another thing we do is anytime an agent responds to a question, it’s always required to provide the source material, like where did it get this answer from? It provides the link, we take it all the way to the front end to the user to see that link, and they can click through and verify for themselves that the data is accurate. I think it’s a host of things, but I would say the biggest thing is making sure we do a lot of the context engineering and making sure that these agents have a narrow scope and they have all the data they need to answer the questions that they’re being asked.

Craig: I’m going to open it up and at the risk of this becoming a free-for-all with all four of you guys, I want to talk about agents and the engineering. You mentioned context engineering, Mustapha, and I’m concerned about context and prompting. I feel that a lot of the AI apps we see every day involve an advisor being able to ask any question they want, which really is telling the advisor, you need to be a good prompt engineer to get anything out of this system. Is that what you’re saying here, or is it something else entirely going on? How are we going to avoid that, forcing the advisors to be prompt engineers?

Beyond Prompting: The Super Prompt Era

Mustapha: That’s a great question. We were talking about this earlier this morning. Our goal is definitely to remove the prompt engineering away from the financial advisor. We want to handle and deal with all the complexity. We do the context engineering with these AI agents, these expert bots, and we also do the prompt engineering with these super prompts. I’m happy to let Mike talk more about super prompts and our idea behind creating these super prompts, but I think that should answer your question, Craig.

Craig: What’s a super prompt, and is it better than an expert bot?

Mike: No, it’s similar. They work with one another. Imagine a super prompt as being a prompt that knows how to talk to our orchestrator or your wrangler, if you will.

Craig: I’m going to trademark that, bot wrangler.

Mike: Our super prompts are highly curated prompts that are pre-designed for very specific use cases, and they understand how to interact with multiple bots in the scope of our orchestrator. For example, as part of our household brief, we can call our Morningstar bot and pull down Morningstar content about a portfolio. We can interact with other practice management bots that are part of our library so that we can get specific content related to that household brief. These prompts are predefined canned prompts that allow advisors to handle specific use cases, and they understand how to interact with LLMs and our bot store perfectly.

Craig: That can also get complicated. Brian and I were talking about this two weeks ago about how we’re both working with AI agents. When I’m online doing reading on Reddit or on Twitter X, people are posting about great prompts to use with agents. I’m getting overwhelmed. Which prompt should I use? Which one’s better? How are advisors going to know? Are you pre-defining these super prompts and telling the advisor, here’s the prompt to use with this particular bot?

Mike: That’s right. This particular bot or this particular use case. For example, if you need to get ready for an annual review. Are you familiar with the firm name, FP Pathfinder, Craig?

Craig: Yes, I am.

Mike: They’re an example of one of the firms we’re working with to create a bot. If you think about a super prompt that interacts with your CRM data, Redtail, Salesforce, whatever that be, FP Pathfinder, and Morningstar, an advisor could go in and give a full prompt. That would be great, but that’s hard to do, whereas inside of Hamachi, click one button, and it generates you this beautiful report that you can then talk to.

Brian: Part of the thing you were talking about, Craig, is that the super prompts are already curated prompts. You can go to God of Prompt or all these places. I can’t tell you how many times I see the seven prompts that make it best. It’s overwhelming. The super prompts are pre-programmed, tested, validated by us and others, and we just keep refining those as time goes on. That provides the context engineering requirements to know how to work with this type of data that a financial advisor works with, not something else. Our tool specifically, you can’t just go on there and say find me the best flights for my next trip. That’s not the point of the tool.

Brian: The point of the tool is to talk to your data from different sources, aggregate data from different expert bots and provide you a comprehensive result. That is the biggest challenge. We’re past all the fake AI outputs because of all the reasoning that goes on with LLMs now, but it’s still a challenge to have context management. I’m seeing the advertisements that context management is king, which is absolutely right because it gets you the best result every single time consistently. That’s the goal of the super prompts.

Craig: Indeed. The time is flying by. I’m going to jump ahead in a couple of the questions to try to ask ones I think will be the most interesting. I want to talk about differentiation. We all know how fast and beneficial AI is to programming and building. A lot of larger RIAs and asset managers and broker-dealers are building their own systems now because it’s so easy to do. Why shouldn’t a sophisticated RIA or broker-dealer or asset manager look at what you guys are doing and say we can build a lot of this ourselves? What are they underestimating in terms of what you’re building and delivering in this product?

Eric: Mustapha can chime in here too, but I would say that those compliance guardrails are ready to go relative to having to say to an enterprise, how do we engineer all these guardrails that we need so our advisors can leverage this tech? Whether it’s in regards to removing the PII before we share the information with the language model, we’re doing that in a very unique way where we redact and tag all of the PII so that the context is still understood by the language model so that Robert, Bob and Bobby are all mentioned, say in a correspondence, as being the same person. We tag the PII in a way that the AI can still respond appropriately with that context in mind, but yet share the information in a safe manner. We provide complete auditing of the audit trail of not only the prompts that we send over, but the responses that come back so that at any time if you want to use those responses to retrain the agents, you can do that. Or, for compliance purposes, you can go back and show how the advice was derived, we have a complete audit trail in place. Everything that comes back through the Ai platform we run through agents that are trained on SEC and FINRA regulations so the responses that come back are appropriate.

Now, those compliance views are a nice start but certainly from a technical perspective, it took a significant lift for us to get those items in place for those firms and we’re giving them a head start. They can take what we’ve built and move it forward from there as opposed to having to worry about putting that foundation layer of compliance in place. I think, Craig, more often than not, the biggest thing that prevents firms from fully embracing AI is the concerns that they have from the compliance teams. We wanted to address those head-on and make sure that compliance could give the green light in allowing firms to leverage the tech.

Mustapha: Yes, I’d be totally with Eric. I would say, while it’s easier to build software these days with AI, it’s still difficult to build an AI system that is reliable, compliant, accurate, and we’ve spent countless hours working on these specific problems and we feel like we have a solid solution that deals with PII and accuracy and compliance. To Eric’s point, we will be giving clients a head starts overall.

Craig: Excellent. So one thing we haven’t spoken about is the asset manager part of your business. One of the core parts of the business model as I understand it is, delivering bots from asset managers to advisors or asset managers sponsoring a platform. How is that going to work at scale and what’s going to be available through your platform that wealth management firms can’t already get from their asset manager partners?

Mike: I’ll answer the second part of your question first if you don’t mind, Craig, around what an advisor can’t get from their asset manager without Hamachi. What we’ve designed at Hamachi is the ability for asset managers to build their expertise in an agent and with that, all of their white papers, back sheets, research materials, I’ll use the term “religion” again just because it came up on a call recently with an asset manager who talks about their “religion” and how advisors preach a certain methodology when they’re talking to investors. All of that domain goes into an agent on Hamachi so every time an advisor needs to answer a question about a specific investment, product or strategy, any time they need to take a look at an investor’s portfolio and make a recommendation for a certain product or investment strategy, all of that corpus gets immediately accessible in the context of the investor and his or her CRM data.

There’s no way an advisor today, advisors are wonderful people and incredibly talented, but no way any advisor can say, I can take a white paper that I get on LinkedIn and a fact sheet I get at a conference and put those together and make it relevant for my 500 clients. Sure, you could send it to them, give them a one-liner, but making it super personalized and relevant is nearly a fool’s errand at that point. However with our platform, every conversation that you have with a client or a prospect about a specific investment strategy can be highly personalized.

Craig: I like it. That’s important. We talk about personalization all the time, it’s kind of a buzzword but I love to see an application like Hamachi be able to do that at scale across all conversations. We’re still running out of time, I’m going to keep squeezing a couple questions in. All of you guys have successfully launched companies, scaled them, and sold them, and now you’re starting all over again. What is harder this time? The market’s changed, technology’s changed, the industry’s changed, what’s harder: is it building the product, earning trust, or navigating perceptions of your prospects on what you’re doing?

Eric: I would say that AI is the biggest advantage that we have this go round, and it’s also the biggest disadvantage because just as soon as we get an idea in place it can be quickly commoditized because Ai is advancing at such a rapid clip. So not only are we embracing AI to create the offering, but AI itself is becoming exponentially better as we’ve spent the last year together building Hamachi. It’s been unbelievable to look back and see the advances that AI out of the box has been able to provide to firms. We’re super excited about it, obviously we wouldn’t be anywhere near where we are today without it, but we’re still trying to make sure that our offering is on the cutting edge of allowing advisors to use all fo the latest and greatest that AI has to offer.

Mustapha: I would just add that AI is still relatively new, so the challenge for us and for lots of AI startups is we’re not only trying to deliver a great product, we’re also trying to change advisors’ habits to use AI and leverage it. Which is an added level of difficulty compared to if we were building just another SaaS product like all of us did before.

Craig: Mike, zooming out say 12 months from now you’ve got traction, you’ve got some clients, things are working. These new clients are now taking advantage of Hamachi’s platform. How does it change their business model, how does it change their division of labor in their staff versus firms that are not using Hamachi?

Mike: AI isn’t going to replace advisors. AI is going to get rid of all the stuff that’s slowing advisors down. What we seek to do is add a ton of efficiency to advisors day in and day out. Some of what we’re producing in the daily brief for example will help isolate what certain people on your team should be doing from the advisor to the paraplanner, to the assistant, the admin, everybody. It’s really designed to help folks understand where to act in their daily operations. I’ll use the term efficiency again, that’s exactly where we’re headed with Hamachi, is trying to open up advisor practice to serve more clients. That can come in the form of organic growth, it might come in the form of an advisor and his or her assistants get to play more golf during the week. Whatever it is that a practice is trying to do, we’re trying to create some space in their day.

Craig: Excellent. Last question, Brian, putting your future vision glasses on, where do you see AI-driven advisor communications, is it a built in feature or are advisors pulling in separate tools like Hamachi for these specific use cases?

Brian: I think it’s going to be specialized tools on top of existing platforms. The existing platforms are going to take some time to incorporate AI into everything, they have a lot of work to do that is very valuable for their applications. But it’s still having multiple tools that an advisor has to implement in their business. So many different tools from email, to Google services, or maybe Microsoft Office365 apps, to all of that incorporated into one. Being able to interface all of that in one application like Hamachi I think is the future state of these tools while people are building out the internal AI use cases for internal workflows in a CRM or in a portfolio management system could make those individual workflows even better. That’s where I think the power is in the current state.

Craig: Awesome, wow. Time has flown, gentlement, I really appreciate it. We are out of time. Where can people find more information about Hamachi?

Mike: Our homepage is Hamachi.ai. Maybe soon, even, BotWrangler.ai.

Craig: Yes, quick, get that one before I do. Great, guys. Eric, Brian, Mike, Mustapha, thanks so much. I really appreciate talking to you.

Brian: Thank you, Craig.

Which AI Productivity Tools Are Right for Your Firm?  Ask Ezra Group!

Integrate smarter tools. Streamline workflows.
Boost advisor confidence.

Talk to an Expert →