We Let Our Financial Planners Write the Agents — Matthew Ragone, Range
https://www.youtube.com/watch?v=gpmze-6ueGU
AI Engineer · 8 min · 120 sentences · sentence-level transcript from YouTube's caption track; each line begins at the time shown

[00:01] Hi everyone, I'm Matt Rigon and I'm a financial planner at Range.
[00:05] I want to show you how our financial planners ended up writing the agents that we use every day and none of us the financial planners for that matter write code.
[00:16] This is the build guide.
[00:18] How we did it step by step so you can take it back to your own teams, companies, etc. So, a quick bit about me because it matters for how you'll hear this talk.
[00:28] I'm not an engineer.
[00:30] I'm a certified financial planner and I spend my days with clients talking about cash flow, taxes, insurance, retirement, estate, you name it.
[00:38] I also lead the AI adoption for the planning team at Range, which is an all-in-one wealth management company.
[00:46] Most agent talks you'll hear come from the people who built the platform.
[00:51] This talk comes from someone on the other side, the people you hand the platform to.
[00:56] We work in a regulated business.
[00:58] The output is financial advice to real clients.
[01:02] And the review step is not optional.
[01:05] So, this isn't a vendor pitch.
[01:07] It's a method and it's what actually worked for us.
[01:12] We started where everyone starts AI as a chatbot.
[01:16] I'd write a client message, paste it in, and ask it to make it a little clearer or friendlier.
[01:21] And that was useful.
[01:22] It saved me a few minutes here and there, but I was still doing all of the work manually.
[01:28] I was just doing it a little bit faster.
[01:30] The change that actually mattered came later.
[01:35] Here's what we figured out.
[01:36] A planner's constraint was never typing speed, for example.
[01:40] It's that we can only be in one place at a time.
[01:44] If I'm on a call with a client, I'm not building someone else's plan.
[01:49] Writing faster doesn't fix that problem.
[01:51] What fixes it is work that happens while I'm somewhere else.
[01:56] So, here's what a day looks like now.
[01:59] I'm on a call with one client.
[02:01] At the same time, an agent is drafting a full cash flow analysis or an insurance analysis for a different client.
[02:09] When the call ends, I open that draft.
[02:12] I fully review it, adjust what needs to be adjusted, and send the client their recommendations.
[02:18] I want to be really clear about that last step.
[02:21] The agent doesn't send anything.
[02:23] A planner reviews every single recommendation before a client sees it.
[02:29] What changed is that the first draft no longer needs me sitting in front of it.
[02:35] Getting there took a lot more than a good prompt.
[02:38] However, two things had to be true.
[02:41] First, the agent has to work where the work lives, not in a separate chat window I'm copying and pasting it into.
[02:49] Second, it has to know how we do it, not how financial planning works in general.
[02:54] The model already knows that, but the agent needs to know how range does it.
[03:00] What we look at first, what we always flag, how we phrase a specific recommendation.
[03:05] That combination is what a skill is.
[03:08] Our firm's way of doing one specific piece of work written down so an agent can follow it.
[03:16] The bulk of this talk is how you build one.
[03:21] Five steps.
[03:22] Pick, teach, pack, test, and then ship.
[03:26] None of them require writing code.
[03:29] Every one of them requires knowing the work really well, which is exactly why the people doing the work should also be the ones building it.
[03:40] Step one, pick a workflow you repeat.
[03:43] Your initial instinct is to go for the flashiest work, whatever's taking the most amount of time and the most complex.
[03:49] Don't pick something you do every week where you already know what good looks like, where you can spot a bad draft at a glance.
[03:58] Additionally, we recommend choosing some of the least complex workflows to start, but some that require the most manual work.
[04:07] For us, that meant things like cash flow analysis and insurance reviews.
[04:12] Just for us here at range, that required the most manual work.
[04:16] And if you can't judge the output quickly, you can't test it and you can't trust it.
[04:21] So look for two things.
[04:23] You do it often and you know how to judge it.
[04:28] Step two, show it several real cases for previous plans you've delivered.
[04:33] Take several actual client cases and walk through the work start to finish.
[04:39] And as you do it, narrate your judgment out loud.
[04:42] Not just next I open the pastub, why you're opening it, what you're looking for, what would make you stop and ask a question.
[04:51] That narration is the whole game.
[04:54] The steps are easy to write down.
[04:55] The judgment is what makes it your firm's work instead of generic advice.
[05:03] Step three, pack it into a skill.
[05:06] A skill isn't just a lip a list of steps.
[05:09] Two things matter most.
[05:10] The questions it must ask before it starts.
[05:14] What does it need to know?
[05:15] What should it confirm with me instead of simply guessing?
[05:19] And the limits it must respect.
[05:21] What it never does on its own.
[05:23] In our world, the big one is that it never sends anything to a client.
[05:28] It drafts and then a planner decides if it gets sent or needs to be adjusted.
[05:33] If you only write the steps, you get something that simply just runs on its own.
[05:38] If you write the questions and set the limits, you can get something you can trust.
[05:43] Think of a skill as a recipe.
[05:45] Once Claude knows how to do the skill, it will repeat your process on command.
[05:51] Step number four, run it five or 10 times on real cases before anyone else sees it.
[05:59] This is the step everyone wants to skip.
[06:02] Every run teaches you something the skill didn't know yet.
[06:05] An edge case, a question it should have asked, a spot where it sounded confident and was wrong.
[06:10] You need to fix it, save the skill, and run it again and again and again.
[06:16] Five to 10 runs sounds like a lot, but it's a lot less work than cleaning up after you hand a halfbaked skill to 15, 20, 100 people in your team.
[06:28] Step five, ship it with an owner and a reviewer.
[06:32] The owner keeps it current.
[06:34] Our processes change, tax rules change, and a skill that was right in March can be wrong by December.
[06:41] The reviewer checks it before the team relies on it.
[06:45] And it's someone other than the author.
[06:48] Same as a plan, you don't grade your own work.
[06:51] Once other people depend on a skill, somebody has to be accountable for it.
[06:56] That's the difference between a clever personal shortcut and something a team can build on.
[07:03] So, where are we at currently?
[07:05] Six months in, we have 24 skills across the company and 17 of them are in planning.
[07:11] And here's the part I didn't expect.
[07:13] It spread to sales, product, marketing, all without a mandate.
[07:18] Nobody sent a memo.
[07:19] People saw what the planners were doing and started building their own agents.
[07:24] Every one of those skills was written and is maintained by planners, marketers, and salespeople, not developers or engineers.
[07:34] A couple of things changed once subject matter experts started building for their own niche.
[07:40] The skills got a lot more specific.
[07:42] The person who knows insurance, for example, best builds the insurance skill.
[07:47] The person who best knows tax, tax code, tax law should be building the tax one.
[07:53] and a review came naturally.
[07:55] We already review each other's plans and that's just how planning works.
[07:59] So now we review each other's skills in that same way.
[08:03] It's the same habit, just a new thing to review.
[08:08] So why does this matter right now?
[08:10] Because the bottleneck has moved.
[08:13] Whether a model can draft a cash flow analysis, that question's already been settled.
[08:17] It can.
[08:18] What's not settled?
[08:20] Who writes down how your firm does it?
[08:22] who tests it before it ships, who owns it once 15 other people depend on it.
[08:28] If you're being asked to push agents out past the engineering org, you're about to hit these exact questions.
[08:35] There isn't much field data on this yet, but this is how we've done it here at Range.
[08:43] So, that's it.
[08:44] Thank you so much for listening to this talk.
[08:46] I'm Matt Rigon from Range and you can reach me for at mattrange.com.
[08:50] Have a great day.