Anthropic’s Claude models disappeared for 19 days in early 2025. No warning, no staged rollback. Export controls kicked in, the API went silent, and thousands of firms that had built meeting prep, compliance drafting, and client onboarding workflows on Claude suddenly had nothing.

If your advisory firm runs any part of client service through an AI tool, that shutdown was a wake-up call. The lesson isn’t to stop using AI. It’s to treat AI as infrastructure and build accordingly. That means fallback processes, multi-provider strategies, and a clear map of which workflows break if a single vendor goes offline.

The 19-Day Shutdown and What It Revealed

Claude’s Fable 5 and Mythos 5 models were pulled on short notice due to export restrictions. For 19 days, firms that relied on Claude for summarizing meeting transcripts, drafting SOAs, or generating client briefs had to revert to manual work or scramble to remap their prompts to OpenAI or Gemini.

The firms that felt the least pain were the ones that had already documented their workflows, stored prompt templates outside the vendor platform, and maintained a secondary provider account. The firms that felt the most pain were running everything through a single API key with no written process for what happens when that key stops working.

Financial advisory firms typically don’t think of AI as infrastructure. You think of it as a tool, a productivity boost, maybe a paraplanner assistant. But if losing access to Claude means your advice documentation cycle time jumps from three days back to two weeks, it’s infrastructure. If your advisers can’t walk into a client meeting without the AI-generated brief they’ve come to rely on, it’s infrastructure.

Infrastructure requires redundancy. It requires documentation. It requires a plan for when it fails.

Where Advisory Firms Rely on AI Without Realizing It

Most advisory firms we work with have adopted AI in three places, often without a formal rollout or vendor evaluation. An adviser tries ChatGPT for meeting prep, likes it, and shares the prompt with the team. A paraplanner uses Claude to draft file notes. The client services team uses an AI transcription tool to summarize discovery calls.

Within six months, those ad hoc experiments become load-bearing. The firm’s advice documentation process now assumes the paraplanner will use Claude to turn a 90-minute meeting transcript into a structured SOA draft. The adviser’s pre-meeting routine now includes pulling up the AI-generated client brief. The onboarding workflow now includes an AI-summarized fact-find that the adviser reviews before the first strategy session.

None of this shows up in the firm’s risk register. There’s no vendor contract, no business continuity plan, no fallback process. When Claude went dark, firms discovered they’d built critical workflows on a platform they didn’t control and couldn’t replace in real time.

The three places this shows up most often are meeting prep, compliance documentation, and client onboarding. Let’s walk through what each one looks like when it breaks.

Meeting Prep Without AI: Back to the Spreadsheet

Advisers spend 5 to 10 hours a week preparing for client meetings. You’re pulling portfolio performance, reviewing recent emails, checking goal progress, and writing a one-page brief that reminds you where the client is and what you need to discuss.

When firms start using AI for this work, they build a Meeting Prep Agent. It’s usually a prompt template that takes the client’s name, pulls data from the CRM and portfolio system, and generates a structured brief. The adviser reads it 10 minutes before the meeting and walks in ready.

When Claude went offline, firms that had built this workflow on Claude’s API had to revert to manual prep. That means the adviser or a support team member is back to opening five browser tabs, copying data into a Word doc, and writing the brief by hand. The 10-minute pre-meeting routine becomes a 45-minute task, and the adviser either skips it or blocks out more time in the calendar.

The firms that had documented their Meeting Prep Agent workflow, stored the prompt template in a shared drive, and maintained an OpenAI account were able to remap the workflow in a few hours. The firms that hadn’t done that lost the capability entirely until Claude came back online.

This is what treating AI as infrastructure looks like. You don’t just build the workflow. You document it, test the fallback, and make sure the team knows what to do when the primary provider fails.

Compliance Documentation: The $8K Bottleneck

SOAs, ROAs, and file notes are the most expensive part of advice delivery. A paraplanner spends 8 to 15 hours drafting a single SOA, and the firm pays $3,000 to $8,000 in labor cost per document. Cycle times stretch into weeks because the paraplanner is juggling multiple clients and the adviser’s review feedback creates another round of edits.

Firms that use AI for compliance documentation typically build an Advice Document Agent. It takes the meeting transcript, the client’s fact-find, and the firm’s compliance template, then drafts the SOA in the firm’s voice. The paraplanner reviews it, makes edits, and sends it to the adviser for sign-off. What used to take two weeks now takes three days.

When Claude went offline, firms that had built their Advice Document Agent on Claude’s models saw cycle times balloon back to pre-AI levels. The paraplanner couldn’t generate the first draft, so they went back to writing it from scratch. The adviser’s review took longer because the draft quality dropped. Clients waited an extra week for their advice document.

The firms that had a fallback process didn’t lose the capability. They’d already tested their Advice Document Agent on OpenAI’s GPT-4 and knew the output quality was comparable. When Claude went dark, they updated the API key, ran a few test documents, and kept moving.

The lesson here isn’t that Claude is unreliable. The lesson is that any single-vendor dependency is a risk. If your compliance workflow depends on one AI provider, you need a documented process for switching to another one. That means storing your prompt templates outside the vendor platform, testing your workflow on at least two providers, and making sure your team knows how to execute the fallback.

We walk firms through this in the AI audit for financial advisory firms. You bring your existing workflows, we map the dependencies, and we build a multi-provider strategy that keeps your advice delivery running even when one vendor goes offline.

Client Onboarding: The 60-Day Drag

New clients lose momentum during onboarding. The typical advisory firm takes 30 to 60 days to move a new client from signed engagement letter to first strategy session. That’s because onboarding involves a lot of manual work: collecting KYC documents, running the fact-find, entering data into the CRM, and preparing the onboarding pack for the adviser.

Firms that use AI for onboarding typically build a Client Onboarding Agent. It runs a guided fact-find with the new client, collects KYC documents, and prepares a clean onboarding pack that includes the client’s goals, current position, and risk profile. The adviser reviews the pack, books the strategy session, and the client moves from signed to active in two weeks instead of two months.

When Claude went offline, firms that had built their Client Onboarding Agent on Claude’s API had to revert to manual onboarding. That means the client services team is back to sending email requests for documents, manually entering data into the CRM, and writing the onboarding pack by hand. The 60-day cycle time comes back, and new clients start to disengage.

The firms that had a fallback process didn’t lose the capability. They’d already tested their Client Onboarding Agent on Gemini and knew it could handle the same workflow. When Claude went dark, they switched providers and kept onboarding new clients at the same pace.

This is the difference between treating AI as a tool and treating it as infrastructure. If your onboarding workflow depends on one AI provider, you need a plan for when that provider fails. That means documenting the workflow, testing it on multiple providers, and making sure your team can execute the switch without waiting for IT support.

What a Multi-Provider Strategy Looks Like

A multi-provider strategy doesn’t mean you run every workflow on three different AI platforms at the same time. It means you’ve tested your workflows on at least two providers, documented the differences in output quality, and built a process for switching between them when one goes offline.

Here’s what that looks like in practice. You build your Meeting Prep Agent on Claude because you like the output quality. You test the same workflow on OpenAI’s GPT-4 and find that the output is 90 percent as good but takes a bit more prompt engineering to get the structure right. You document both versions of the prompt, store them in a shared drive, and make sure your team knows how to switch the API key if Claude goes offline.

When Claude goes dark, you don’t lose the capability. You switch to OpenAI, run a few test briefs to confirm the output quality, and keep moving. Your advisers still get their pre-meeting briefs, your cycle time doesn’t change, and your clients don’t notice the difference.

The same logic applies to your Advice Document Agent and your Client Onboarding Agent. You build them on your preferred provider, test them on at least one backup, and document the process for switching. That way, when export controls or API outages or vendor shutdowns happen, you’re not scrambling to rebuild your workflows from scratch.

We help firms build this multi-provider strategy in the Omni Audit. You bring your existing workflows, we test them on multiple providers, and we document the fallback process so your team can execute it without waiting for external support. Book a 60-min Omni Audit and we’ll map your dependencies, test your fallbacks, and build a continuity plan that keeps your advice delivery running no matter which vendor goes offline.

The Dollar Reality of Single-Vendor Risk

Financial advisory firms in the $1M to $25M revenue range typically leak $70,000 to $200,000 a year to manual work that AI could handle. That’s meeting prep, compliance documentation, and client onboarding work that takes 10 to 20 hours a week per adviser and another 15 to 25 hours a week of paraplanner time.

When you adopt AI and cut that manual work by 60 to 80 percent, you recover most of that leakage. Your advisers spend less time on prep and more time in client meetings. Your paraplanners draft SOAs in three days instead of two weeks. Your onboarding cycle drops from 60 days to 15 days, and new clients stay engaged.

But if you build all of that on a single AI provider and that provider goes offline, you lose the efficiency gain overnight. Your advisers go back to manual prep, your paraplanners go back to writing SOAs from scratch, and your onboarding cycle stretches back out to 60 days. The $150,000 in recovered capacity disappears, and you’re back to pre-AI productivity levels.

The cost of a multi-provider strategy is minimal. You spend a few hours testing your workflows on a second provider, you document the fallback process, and you maintain a backup API account. That’s maybe 10 hours of setup time and $200 a month in additional API costs.

The cost of not having a multi-provider strategy is the full leakage amount. When your primary provider goes offline and you have no fallback, you lose the entire efficiency gain until the provider comes back online. For a firm leaking $150,000 a year to manual work, a 19-day outage costs roughly $8,000 in lost capacity. That’s eight grand you’ve already paid to recover, now gone because you didn’t spend 10 hours building a fallback.

How Omni Builds Continuity Into Every Workflow

When we build AI agents for advisory firms, we treat continuity as a design requirement from day one. That means every agent we build is tested on at least two providers, documented with a fallback process, and deployed with a clear plan for what happens when the primary provider fails.

Here’s what that looks like for the three agents we mentioned earlier.

The Meeting Prep Agent pulls portfolio data, recent comms, and goal progress into a one-page brief the adviser reads before every client meeting. We build it on your preferred provider, test it on a backup, and document the prompt template so your team can switch providers without losing the workflow. When your primary provider goes offline, you update the API key, run a few test briefs, and keep moving.

The Advice Document Agent drafts SOAs, ROAs, and file notes from meeting transcripts and the firm’s compliance template. We build it on your preferred provider, test it on a backup, and document the differences in output quality so you know what to expect when you switch. When your primary provider goes offline, you switch to the backup, run a few test documents, and keep delivering advice at the same pace.

The Client Onboarding Agent runs a guided fact-find with new clients, collects KYC docs, and prepares a clean onboarding pack for the adviser. We build it on your preferred provider, test it on a backup, and document the fallback process so your client services team can execute the switch without IT support. When your primary provider goes offline, you switch to the backup and keep onboarding new clients at the same pace.

This is what treating AI as infrastructure looks like. You don’t just build the workflow. You document it, test the fallback, and make sure the team knows what to do when the primary provider fails. You can explore more about how we approach agent design at Omni Ops, where we walk through the full build process.

What the Audit Delivers

The Omni Audit is a 60-minute session where we map your firm’s manual work, identify the workflows that AI can handle, and build a multi-provider strategy that keeps your advice delivery running even when one vendor goes offline.

You walk away with three outputs. First, a workflow map that shows where your firm relies on AI, which providers you’re using, and where the single-vendor dependencies are. Second, a fallback plan that documents the process for switching providers when your primary vendor goes offline. Third, a cost model that shows how much capacity you’re recovering with AI and how much you’d lose if your primary provider failed.

We don’t deliver a deck. We don’t schedule a follow-up. You get the three outputs in the session, and you decide what to do next. Most firms take the workflow map and fallback plan, test the backup providers themselves, and come back when they’re ready to build the agents. Some firms want us to build the agents in the audit session and deploy them the same week.

The audit costs nothing if you’re a financial advisory firm doing $1M to $25M in revenue and you’re serious about building AI workflows that don’t break when one vendor goes offline. Book my Omni Audit and we’ll map your dependencies, test your fallbacks, and build a continuity plan in 60 minutes.

The Infrastructure Mindset

The 19-day Claude shutdown taught every firm the same lesson. If you’re building workflows on AI, you need to treat AI as infrastructure. That means redundancy, documentation, and a clear plan for when the primary provider fails.

Financial advisory firms don’t typically think this way. You think of AI as a tool, a productivity boost, maybe a paraplanner assistant. But when losing access to one AI provider means your advice documentation cycle time jumps from three days back to two weeks, it’s infrastructure. When your advisers can’t walk into a client meeting without the AI-generated brief, it’s infrastructure. When your onboarding workflow depends on an AI-powered fact-find, it’s infrastructure.

Infrastructure requires a different approach. You don’t just adopt it and hope it keeps working. You document it, test the fallbacks, and make sure the team knows what to do when it fails. You maintain backup providers, store your prompt templates outside the vendor platform, and build a process for switching between providers without losing the workflow.

This isn’t complicated. It’s 10 hours of setup time, a few hundred dollars a month in backup API costs, and a documented process that your team can execute without IT support. The cost of not doing it is the full leakage amount when your primary provider goes offline.

We’ve built this continuity process into every agent we deploy for advisory firms. You can see the full approach at Omni for financial advisory firms, or book the audit and we’ll walk you through it in 60 minutes. You’ll leave with a workflow map, a fallback plan, and a cost model that shows exactly how much capacity you’re recovering and how much you’d lose if your primary provider failed.

The firms that came through the Claude shutdown with the least pain were the ones that had already built this infrastructure mindset into their AI workflows. They had documented their prompts, tested their fallbacks, and maintained backup provider accounts. When Claude went dark, they switched providers and kept moving.

The firms that felt the most pain were the ones that had built everything on a single vendor with no fallback plan. They lost the capability entirely, reverted to manual work, and spent 19 days waiting for Claude to come back online.

You get to choose which firm you are. The choice is a 10-hour investment in documentation and testing, or an $8,000 loss the next time your primary provider goes offline. We’ll help you build the fallback plan in the audit. Bring your workflows, we’ll map the dependencies, and we’ll build a continuity strategy that keeps your advice delivery running no matter which vendor goes offline.