← All talks

AI Engineer NYC 2026 Online track · 9 min

We Let Our Financial Planners Write the Agents

Matt RagoneRange

At Range, financial planners write and maintain the skills that draft their work. Matt Ragone explains how to turn professional judgment into a repeatable process without handing client communication to an agent.

The constraint was being in two places at once

3:26:48 ↗

A financial planner can ask a chatbot to make a client message clearer and save a few minutes. But while the planner is on a call, somebody else's cash-flow analysis still waits. Faster typing barely touches that constraint.

At Range, an agent prepares that draft while the planner works with another client. Matt Ragone, a certified financial planner leading the team's AI adoption, describes what changed in his own day. After the call, he opened the analysis, reviewed it, changed what needed changing, and sent the recommendations himself. The agent never sent them. In a regulated advice business, that review remained part of the work. 3:28:43

For this arrangement to work, the agent needed access to the place where planners actually worked. Copying material into a separate chat window left the planner coordinating every step. It also needed Range's way of doing a task: what to inspect first, what to flag, and how to phrase a recommendation. General knowledge about financial planning could not supply those firm-specific choices.

The planners wrote those methods as skills the agent could follow. Their advantage was knowing why a draft was wrong and how to fix it. They did not need to become programmers first.

Pick, teach, pack, test, ship

3:30:03 ↗

Ragone describes a five-step process, beginning with work small enough to judge and frequent enough to matter.

  1. Pick a repeated workflow. Start with something done every week, where a knowledgeable person can quickly distinguish good output from bad output. Range chose cash-flow analysis and insurance reviews: familiar work with substantial manual preparation. The most impressive or complex task is a poor first choice if nobody can evaluate the result quickly.
  2. Teach through real cases. Walk through several plans already delivered to clients. Narrate why a pay stub is being opened, which detail matters, and what would trigger a question. A list of clicks records the procedure; the explanation records the professional judgment that makes the procedure useful.
  3. Pack the method into a skill. Write down the questions the agent must ask before starting and the limits it must respect. Missing information should produce a clarification instead of a confident guess. At Range, a central limit is that the agent drafts and a planner decides whether anything goes to a client.
  4. Test on real work. Run the skill five to ten times before sharing it. Each run can expose a missing question, an edge case, or an answer that sounds right but is wrong. Update the skill and repeat. Ragone compares that effort with the much larger cleanup required after distributing an unfinished process to a whole team.
  5. Ship with an owner and a reviewer. An owner keeps the skill current as processes and tax rules change. Someone other than the author reviews it before colleagues depend on it. A method that worked in March may no longer be right in December.

The critical teaching material often sits between the visible steps. Why stop here? Which missing document changes the answer? What must the agent never decide by itself? Those details make a skill more than a checklist. They also make review possible: another planner can examine whether the written process reflects the firm's practice. 3:31:15

Maintenance follows the work

3:33:50 ↗

Six months into this effort, Ragone reports 24 skills across the company, with 17 in planning. He says adoption spread to sales, product, and marketing without a company-wide mandate. Colleagues saw planners getting useful drafts and started building methods for their own work.

The skills became more specific as specialists took ownership. The person who understood insurance best could write the insurance process; the person closest to tax work could maintain that process. Reviewing one another's skills also fit an existing professional habit. Planners already reviewed one another's plans, so the new artifact entered a familiar cycle of authorship and checking.

Once colleagues depend on a skill, someone has to keep it correct. Who updates the insurance method? Who catches a changed tax rule? Range's ownership and peer review answer those questions using habits the planning team already had. The model makes the draft; the planners maintain the practice behind it.

Full transcript

Timestamps follow the original livestream. Select a time to watch that moment. Transcripts are machine-generated with reviewed name and terminology corrections; occasional recognition errors may remain.

3:26:48

Hi everyone, I'm Matt Ragone and I'm a financial planner at Range.

3:26:52

I want to show you how our financial planners ended up writing the agents that we use every day.

3:26:59

And none of us, the financial planners for that matter, write code.

3:27:03

This is the build guide.

3:27:05

How we did it step by step so you can take it back to your own teams, companies, etc.

3:27:11

So a quick bit about me because it matters for how you'll hear this talk.

3:27:16

I'm not an engineer.

3:27:17

I'm a certified financial planner.

3:27:19

and I spend my days with clients talking about cash flow, taxes, insurance, retirement, estate, you name it.

3:27:25

I also lead the AI adoption for the planning team at Range, which is an all-in-one wealth management company.

3:27:33

Most agent talks you'll hear come from the people who built the platform.

3:27:38

This talk comes from someone on the other side, the people you hand the platform to.

3:27:43

We work in a regulated business. The output is financial advice to real clients. And the review step is not optional. So this isn't a vendor pitch. It's a method and it's what actually worked for us.

3:27:59

We started where everyone starts, AI as a chatbot.

3:28:03

I'd write a client message, paste it in, and ask it to make it a little clearer or friendlier.

3:28:08

And that was useful.

3:28:09

It saved me a few minutes here and there, but I was still doing all of the work manually.

3:28:15

I was just doing it a little bit faster.

3:28:17

The change that actually mattered came later.

3:28:22

Here's what we figured out.

3:28:23

A planner's constraint was never typing speed, for example.

3:28:28

It's that we can only be in one place at a time.

3:28:31

If I'm on a call with a client, I'm not building someone else's plan.

3:28:36

Writing faster doesn't fix that problem.

3:28:39

What fixes it is work that happens while I'm somewhere else.

3:28:43

So here's what a day looks like now.

3:28:46

I'm on a call with one client at the same time an agent is drafting a full cash flow analysis or an insurance analysis for a different client.

3:28:56

When the call ends, I open that draft, I fully review it, adjust what needs to be adjusted, and send the client their recommendations.

3:29:05

I want to be really clear about that last step.

3:29:08

The agent doesn't send anything.

3:29:11

A planner reviews every single recommendation before a client sees it.

3:29:16

What changed is that the first draft no longer needs me sitting in front of it.

3:29:22

Getting there took a lot more than a good prompt, however.

3:29:26

Two things had to be true.

3:29:28

First, the agent has to work where the work lives,

3:29:32

not in a separate chat window I'm copying and pasting it into.

3:29:36

Second, it has to know how we do it,

3:29:39

not how financial planning works in general.

3:29:42

The model already knows that,

3:29:44

but the agent needs to know how range does it,

3:29:47

what we look at first, what we always flag,

3:29:50

how we phrase a specific recommendation.

3:29:52

That combination is what a skill is.

3:29:55

our firm's way of doing one specific piece of work written

3:30:00

down so an agent can follow it.

3:30:03

The bulk of this talk is how you build one. Five steps. Pick, teach, pack, test, and then ship. None of them require writing code.

3:30:16

Every one of them requires knowing the work really well,

3:30:20

which is exactly why the people doing the work should also be the ones building it.

3:30:27

Step one, pick a workflow you repeat.

3:30:30

Your initial instinct is to go for the flashiest work,

3:30:34

whatever's taking the most amount of time and the most complex.

3:30:37

Don't. Pick something you do every week where you already know what good looks like,

3:30:42

where you can spot a bad draft at a glance.

3:30:45

Additionally, we recommend choosing some of the least complex workflows to start, but some that require the most manual work.

3:30:54

For us, that meant things like cash flow analysis and insurance reviews.

3:30:59

Just for us here at Range, that required the most manual work.

3:31:03

And if you can't judge the output quickly, you can't test it and you can't trust it.

3:31:09

So look for two things. You do it often and you know how to judge it.

3:31:15

Step two, show it several real cases for previous plans you've delivered.

3:31:20

Take several actual client cases and walk through the work start to finish.

3:31:26

And as you do it, narrate your judgment out loud.

3:31:29

Not just, next I open the pay stub, why you're opening it.

3:31:34

What you're looking for.

3:31:35

What would make you stop and ask a question.

3:31:38

That narration is the whole game.

3:31:41

The steps are easy to write down.

3:31:43

the judgment is what makes it your firm's work instead of generic advice.

3:31:50

Step three, pack it into a skill. A skill isn't just a list of steps. Two things matter most.

3:31:58

The questions it must ask. Before it starts, what does it need to know? What should it confirm with

3:32:04

me instead of simply guessing? And the limits it must respect. What it never does on its own.

3:32:10

In our world, the big one is that it never sends anything to a client.

3:32:15

It drafts, and then a planner decides if it gets sent or needs to be adjusted.

3:32:21

If you only write the steps, you get something that simply just runs on its own.

3:32:25

If you write the questions and set the limits, you can get something you can trust.

3:32:30

Think of a skill as a recipe.

3:32:33

Once Claude knows how to do the skill, it will repeat your process on command.

3:32:39

Step number four, run it five or 10 times on real cases before anyone else sees it.

3:32:46

This is the step everyone wants to skip.

3:32:49

Every run teaches you something the skill didn't know yet, an edge case, a question it should have asked, a spot where it sounded confident and was wrong.

3:32:57

You need to fix it, save the skill, and run it again and again and again.

3:33:03

5-10 runs sounds like a lot, but it's a lot less work than cleaning up after you hand a half-baked skill to 15, 20, 100 people in your team.

3:33:15

Step 5. Ship it with an owner and a reviewer.

3:33:19

The owner keeps it current. Our processes change, tax rules change, and a skill that was right in March can be wrong by December.

3:33:28

Remember, the reviewer checks it before the team relies on it, and it's someone other than the author.

3:33:35

Same as a plan. You don't grade your own work.

3:33:38

Once other people depend on a skill, somebody has to be accountable for it.

3:33:43

That's the difference between a clever personal shortcut and something a team can build on.

3:33:50

So where are we at currently?

3:33:52

Six months in, we have 24 skills across the company and 17 of them are in planning.

3:33:58

And here's the part I didn't expect.

3:34:00

It spread to sales, product, marketing, all without a mandate.

3:34:05

Nobody sent a memo.

3:34:06

People saw what the planners were doing and started building their own agents.

3:34:11

Every one of those skills was written and is maintained by planners, marketers, and salespeople, not developers or engineers.

3:34:21

A couple of things changed once subject matter experts started building for their own niche.

3:34:27

The skills got a lot more specific.

3:34:30

The person who knows insurance, for example, best builds the insurance skill.

3:34:34

The person who best knows tax, tax code, tax law should be building the tax one.

3:34:40

And a review came naturally.

3:34:42

We already review each other's plans, and that's just how planning works.

3:34:46

So now we review each other's skills in that same way. It's the same habit, just a new thing to review.

3:34:55

So why does this matter right now? Because the bottleneck has moved. Whether a model can draft a cash flow analysis, that question's already been settled. It can.

3:35:06

What's not settled, who writes down how your firm does it? Who tests it before it ships? Who owns it once 15 other people depend on it?

3:35:15

If you're being asked to push agents out past the engineering org, you're about to hit these exact questions.

3:35:23

There isn't much field data on this yet, but this is how we've done it here at Range.

3:35:30

So that's it. Thank you so much for listening to this talk.

3:35:33

I'm Matt Ragone from Range, and you can reach me at mattatrange.com.

3:35:38

Have a great day.