We Let Our Financial Planners Write the Agents — Matthew Ragone, Range

AI Engineer · 8 min · 120 sentences · from YouTube's caption track

Each timecode opens YouTube at the start of that sentence. Line anchors (#s42) are the cue ids in the WebVTT, and every line carries its start and end seconds. All transcripts has every talk, and the whole corpus as one file.

1. 00:01Hi everyone, I'm Matt Rigon and I'm a financial planner at Range.
2. 00:05I 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.
3. 00:16This is the build guide.
4. 00:18How 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.
5. 00:28I'm not an engineer.
6. 00:30I'm a certified financial planner and I spend my days with clients talking about cash flow, taxes, insurance, retirement, estate, you name it.
7. 00:38I also lead the AI adoption for the planning team at Range, which is an all-in-one wealth management company.
8. 00:46Most agent talks you'll hear come from the people who built the platform.
9. 00:51This talk comes from someone on the other side, the people you hand the platform to.
10. 00:56We work in a regulated business.
11. 00:58The output is financial advice to real clients.
12. 01:02And the review step is not optional.
13. 01:05So, this isn't a vendor pitch.
14. 01:07It's a method and it's what actually worked for us.
15. 01:12We started where everyone starts AI as a chatbot.
16. 01:16I'd write a client message, paste it in, and ask it to make it a little clearer or friendlier.
17. 01:21And that was useful.
18. 01:22It saved me a few minutes here and there, but I was still doing all of the work manually.
19. 01:28I was just doing it a little bit faster.
20. 01:30The change that actually mattered came later.
21. 01:35Here's what we figured out.
22. 01:36A planner's constraint was never typing speed, for example.
23. 01:40It's that we can only be in one place at a time.
24. 01:44If I'm on a call with a client, I'm not building someone else's plan.
25. 01:49Writing faster doesn't fix that problem.
26. 01:51What fixes it is work that happens while I'm somewhere else.
27. 01:56So, here's what a day looks like now.
28. 01:59I'm on a call with one client.
29. 02:01At the same time, an agent is drafting a full cash flow analysis or an insurance analysis for a different client.
30. 02:09When the call ends, I open that draft.
31. 02:12I fully review it, adjust what needs to be adjusted, and send the client their recommendations.
32. 02:18I want to be really clear about that last step.
33. 02:21The agent doesn't send anything.
34. 02:23A planner reviews every single recommendation before a client sees it.
35. 02:29What changed is that the first draft no longer needs me sitting in front of it.
36. 02:35Getting there took a lot more than a good prompt.
37. 02:38However, two things had to be true.
38. 02:41First, the agent has to work where the work lives, not in a separate chat window I'm copying and pasting it into.
39. 02:49Second, it has to know how we do it, not how financial planning works in general.
40. 02:54The model already knows that, but the agent needs to know how range does it.
41. 03:00What we look at first, what we always flag, how we phrase a specific recommendation.
42. 03:05That combination is what a skill is.
43. 03:08Our firm's way of doing one specific piece of work written down so an agent can follow it.
44. 03:16The bulk of this talk is how you build one.
45. 03:21Five steps.
46. 03:22Pick, teach, pack, test, and then ship.
47. 03:26None of them require writing code.
48. 03:29Every 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.
49. 03:40Step one, pick a workflow you repeat.
50. 03:43Your initial instinct is to go for the flashiest work, whatever's taking the most amount of time and the most complex.
51. 03:49Don'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.
52. 03:58Additionally, we recommend choosing some of the least complex workflows to start, but some that require the most manual work.
53. 04:07For us, that meant things like cash flow analysis and insurance reviews.
54. 04:12Just for us here at range, that required the most manual work.
55. 04:16And if you can't judge the output quickly, you can't test it and you can't trust it.
56. 04:21So look for two things.
57. 04:23You do it often and you know how to judge it.
58. 04:28Step two, show it several real cases for previous plans you've delivered.
59. 04:33Take several actual client cases and walk through the work start to finish.
60. 04:39And as you do it, narrate your judgment out loud.
61. 04:42Not 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.
62. 04:51That narration is the whole game.
63. 04:54The steps are easy to write down.
64. 04:55The judgment is what makes it your firm's work instead of generic advice.
65. 05:03Step three, pack it into a skill.
66. 05:06A skill isn't just a lip a list of steps.
67. 05:09Two things matter most.
68. 05:10The questions it must ask before it starts.
69. 05:14What does it need to know?
70. 05:15What should it confirm with me instead of simply guessing?
71. 05:19And the limits it must respect.
72. 05:21What it never does on its own.
73. 05:23In our world, the big one is that it never sends anything to a client.
74. 05:28It drafts and then a planner decides if it gets sent or needs to be adjusted.
75. 05:33If you only write the steps, you get something that simply just runs on its own.
76. 05:38If you write the questions and set the limits, you can get something you can trust.
77. 05:43Think of a skill as a recipe.
78. 05:45Once Claude knows how to do the skill, it will repeat your process on command.
79. 05:51Step number four, run it five or 10 times on real cases before anyone else sees it.
80. 05:59This is the step everyone wants to skip.
81. 06:02Every run teaches you something the skill didn't know yet.
82. 06:05An edge case, a question it should have asked, a spot where it sounded confident and was wrong.
83. 06:10You need to fix it, save the skill, and run it again and again and again.
84. 06:16Five 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.
85. 06:28Step five, ship it with an owner and a reviewer.
86. 06:32The owner keeps it current.
87. 06:34Our processes change, tax rules change, and a skill that was right in March can be wrong by December.
88. 06:41The reviewer checks it before the team relies on it.
89. 06:45And it's someone other than the author.
90. 06:48Same as a plan, you don't grade your own work.
91. 06:51Once other people depend on a skill, somebody has to be accountable for it.
92. 06:56That's the difference between a clever personal shortcut and something a team can build on.
93. 07:03So, where are we at currently?
94. 07:05Six months in, we have 24 skills across the company and 17 of them are in planning.
95. 07:11And here's the part I didn't expect.
96. 07:13It spread to sales, product, marketing, all without a mandate.
97. 07:18Nobody sent a memo.
98. 07:19People saw what the planners were doing and started building their own agents.
99. 07:24Every one of those skills was written and is maintained by planners, marketers, and salespeople, not developers or engineers.
100. 07:34A couple of things changed once subject matter experts started building for their own niche.
101. 07:40The skills got a lot more specific.
102. 07:42The person who knows insurance, for example, best builds the insurance skill.
103. 07:47The person who best knows tax, tax code, tax law should be building the tax one.
104. 07:53and a review came naturally.
105. 07:55We already review each other's plans and that's just how planning works.
106. 07:59So now we review each other's skills in that same way.
107. 08:03It's the same habit, just a new thing to review.
108. 08:08So why does this matter right now?
109. 08:10Because the bottleneck has moved.
110. 08:13Whether a model can draft a cash flow analysis, that question's already been settled.
111. 08:17It can.
112. 08:18What's not settled?
113. 08:20Who writes down how your firm does it?
114. 08:22who tests it before it ships, who owns it once 15 other people depend on it.
115. 08:28If you're being asked to push agents out past the engineering org, you're about to hit these exact questions.
116. 08:35There isn't much field data on this yet, but this is how we've done it here at Range.
117. 08:43So, that's it.
118. 08:44Thank you so much for listening to this talk.
119. 08:46I'm Matt Rigon from Range and you can reach me for at mattrange.com.
120. 08:50Have a great day.