AI is moving into the RIA back office, and the pressure is real: grow AUM without growing headcount. Operations teams, the people who open accounts, move data between the CRM and custodian, build client reports, and triage the inbox, are being asked whether tools like Claude Code can take the repetitive work off their plate. It's a fair question with an honest, two-part answer.
Claude Code is genuinely capable. For analysis, drafting, and technical tasks an advisor or ops person reviews, it's a real boost. But "reliable enough to lean on" and "reliable enough to run your operations unattended" are different bars, and for a fiduciary handling client money and data, the second bar is the one that matters. This guide is written for it.
Where RIAs are trying to use Claude Code in operations
Most firms aren't trying to get Claude Code to give advice, they're trying to recover the hours their teams spend moving data between systems. The usual candidates:
- Account opening & onboarding. Pulling household, beneficiary, and account data from forms and IDs into the CRM and custodian with no rekeying.
- PDF-to-CRM data entry. Getting statements, applications, and forms into Wealthbox, Redtail, or Salesforce accurately.
- Reconciliation. Checking positions and balances across the custodian, portfolio system, and CRM.
- Client reporting. Pulling performance and positions and assembling recurring review packets.
- Inbox triage. Sorting client requests, custodian notices, and service tasks and routing each to the right place.
The reliability frustrations ops teams run into with Claude Code
If you've piloted Claude Code on this kind of work, these will feel familiar. They don't mean you're using it wrong, they're properties of the tool. Here's each one and what's actually happening.
Claude Code isn't following your instructions
You give clear instructions and it honors most, then drops one. Instructions are a prompt the model interprets and weighs against everything else in context, not a rule it's bound to. The more detailed your instructions, the more likely a step gets dropped, which is a problem when one of those steps is validating an account number.
Claude Code keeps changing the format
The same task comes back formatted differently each run, a date format here, a column order there, a field labeled a new way. The model regenerates output rather than filling a fixed template, so structure drifts even when content is right. For data flowing into a CRM or a client report, that drift breaks things downstream.
Claude Code gives inconsistent results
Run the same task twice and you can get two different answers. A large language model generates the most likely output rather than executing a fixed procedure, and small context differences nudge the result. Fine for exploring; a blocker when reconciliation or reporting has to come out the same way every time.
Claude Code produces different output every time
Even when the answer is correct, its shape changes between runs. Anything that consumes a fixed shape, a reconciliation sheet, a CRM import, a report template, breaks when the shape moves.
Claude Code won't follow your rules
You document the rules and it still goes off-script. A rules file is context the model weighs, not an enforcement layer. When rules seem to conflict, the model silently picks how to resolve them, and you don't control the choice.
Claude Code makes mistakes
It produces plausible output, and plausible isn't correct. Usually it's right; occasionally it's confidently wrong in an easy-to-miss way, with nothing flagging it. On client money and data, those misses ship unless a person checks every run.
Claude Code hallucinations
Sometimes the most plausible output is invented, an account value, a figure, a detail that reads as normal but isn't real. Hallucinations can be reduced but not guaranteed away, which is why client-data work shouldn't depend on a model generating the answer each run.
Why this happens: Claude Code is non-deterministic by design
Every frustration above traces to one root cause. Claude Code is built on a large language model, and language models are non-deterministic: they generate output by predicting what's most likely, sampling from a range of possibilities. The same input can produce different output on different runs. That's not a defect, it's what makes them flexible and good at language.
It's also why they shine for analysis and drafting and struggle with unattended operations. A person reviewing each result absorbs the variability; an automation running unattended at volume cannot. For a fiduciary, "mostly right" on a reconciliation or a client report is not a standard you can stand behind. The reliability gap is structural, not a missing prompt.
| Claude Code | Caddi | |
|---|---|---|
| What runs in production | A model generates fresh output on every run | Deterministic code, generated once at setup |
| Same input, same output? | Not guaranteed, output can vary run to run | Yes, identical inputs yield identical results |
| Following your rules | Instructions are a prompt the model may reinterpret | Rules are compiled into the workflow, not re-read each time |
| Auditability | Hard to prove what happened or why | Full run-by-run audit trail (SOC 2) |
| Who maintains it | You re-prompt and babysit it | Built and maintained for you |
Is Claude Code too complicated for a non-technical ops team?
There's a second barrier. Claude Code is a developer tool, terminal, prompts, configuration, and maintenance. For an engineer that's natural; for the operations associate or client-service lead who actually runs the workflow, it's a steep climb, and the integration work to reach your CRM and custodian is real engineering on top.
The result is that the people closest to the work usually can't build for it, so automation stalls or lands on one technical person. Most ops teams don't want to become prompt engineers; they want the repetitive work to stop. A tool that requires the first to deliver the second rarely makes it past the pilot.
Can you trust Claude Code with client data?
As a fiduciary, the bar isn't only "is the model secure", it's whether you can scope and control access, prove exactly what happened with client data on every run, and get the same handling every time. Prompt-driven, run-to-run-variable execution is hard to audit and constrain, which is the opposite of what compliance and clients expect.
The real risk is the quiet one: a hallucinated value in a record, inconsistent handling that surfaces in an exam, or credentials sitting in an ad-hoc script no one can account for. Those are the failures that create regulatory and client-trust problems, so the operating model has to make them structurally hard.
Why firms give up, and what reliable automation actually looks like
This is why so many RIAs prototype something promising in Claude Code and then quietly abandon it. The demo works; production doesn't. Connecting the CRM, custodian, and portfolio system is real engineering, reliability takes constant prompt-tuning, and every tool change breaks something the firm then has to fix. Giving up isn't a failure of effort, it's the wrong tool for the hardest 80% of the job.
The fix is to change the order of operations. Caddi uses AI once, at setup, to understand a workflow you demonstrate over a screen share, the way you'd train a new associate. From then on it runs that workflow as deterministic code over real connections to your CRM, custodian, and documents: the same inputs produce the same outputs every time, every run is audit-logged, and genuine exceptions are routed to a person. The reasoning happens at design time, where variability is fine; production is just code, where it isn't. No terminal, no prompts to maintain, and Caddi keeps it working as your stack changes.
Keep reading
See deterministic automation in action
Caddi builds reliable automations from a screen recording and runs them across 70+ tools as deterministic code. Explore real workflows for law firms and RIAs & financial advisors, or book a demo to see your own workflow built live.
Do more with less
See Caddi in action
Tell us where to reach you and the calendar opens right here. In 30 minutes we'll show you how Caddi automates the back-office work that grows with your clients—built, run, and maintained for you.
Frequently asked questions
Is Claude Code reliable enough for RIA operations?
For supervised work like analysis and drafting, yes. For unattended, rule-bound back-office workflows it's limited by non-determinism, the same input can yield different output run to run, so it's hard to run without a person checking every result.
Why does Claude Code give inconsistent results?
Because it's a large language model that generates output probabilistically rather than executing a fixed procedure, so the same task can produce different results on different runs.
Why won't Claude Code follow my firm's rules?
Rules in a prompt or rules file are guidance the model weighs, not an enforcement layer it must obey, so compliance isn't guaranteed, especially with many rules or long context.
Can I trust Claude Code with client data?
For unattended work the concerns are auditability, scoped access, and consistent handling, where prompt-driven execution is weak. Deterministic automation with scoped access and a full audit trail (like Caddi, which is SOC 2 attested) is a stronger fit for client data.
Do I need to be technical to use Claude Code?
Effectively yes, it's a developer tool driven from a terminal with prompts and maintenance. Non-technical ops teams typically get more from a done-for-you platform like Caddi, where you demonstrate the workflow and it's built and maintained for you.
What should an RIA use for reliable back-office automation?
Caddi. It captures a workflow once and runs it as deterministic code across your CRM, custodian, and documents, identical every run, audit-logged, and maintained for you.