Palantir and Nvidia just launched an air-gapped AI stack, and the headline isn’t the technology. It’s the economics. Token-based billing is cracking under the weight of continuous-use enterprise applications, and financial advisory firms running AI for portfolio monitoring, compliance documentation, or client onboarding are about to feel it.

If you’re paying per API call or per thousand tokens, the math stops working the moment you automate something that runs all day. A Meeting Prep Agent that pulls portfolio data and recent comms before every client meeting might query your CRM, custodian feeds, and document store fifty times a day per adviser. At $0.02 per call, that’s a dollar a day per adviser. Across ten advisers over a year, you’re at $3,600 before you’ve touched compliance documentation or onboarding workflows. Scale that to a firm with thirty advisers running three agents each, and you’re looking at $30K-50K in token costs alone, not counting the platform subscription or the engineering time to wire it all together.

The firms we work with in the $1M-25M revenue band are starting to ask the right question: why are we renting compute by the sip when we could own the tap? This article walks through the token billing problem, what it means for the three highest-value AI use cases in advisory firms, and how to evaluate flat-rate or self-hosted alternatives before per-query costs make automation uneconomical.

Why Token Billing Worked Until It Didn’t

Token-based pricing made sense in 2023. You paid for what you used. If you ran ten queries a month to summarize client emails, you paid ten cents. If you scaled to a thousand queries, you paid ten dollars. The model was elastic, transparent, and easy to budget.

Then firms started automating continuous workflows. Portfolio monitoring agents that check for rebalancing triggers every morning. Compliance agents that draft file notes after every meeting. Onboarding agents that run a guided fact-find with every new client and pull KYC documents from three different systems. These aren’t one-off queries. They’re always-on processes that touch multiple data sources, run inference loops, and generate structured outputs dozens of times a day.

The token meter starts spinning, and the bill climbs faster than the value. One wealth management firm in our network describes their first quarter running an advice document agent: $4,800 in API costs to draft forty SOAs. That’s $120 per document, on top of the $3K-8K in paraplanner time they were trying to reduce. The agent worked. The economics didn’t.

The problem isn’t the technology. It’s the pricing model. Token billing assumes episodic use. Continuous automation breaks that assumption, and firms are waking up to the fact that they’re paying SaaS rent on something that should be infrastructure.

The Three Use Cases Where Token Costs Spiral

Not every AI workload hits the token wall. Summarizing a single email or answering a one-off client question costs pennies. The pain shows up in three specific places where advisory firms automate high-frequency, multi-step workflows.

Meeting Prep and Notes

Advisers spend five to ten hours a week preparing for client reviews and writing them up afterwards. A Meeting Prep Agent pulls portfolio performance, recent comms, goal progress, and compliance flags into a one-page brief the adviser reads before every meeting. The agent queries the CRM, custodian API, document store, and compliance log. That’s four to six API calls per meeting. An adviser running twenty client meetings a month generates 80-120 calls. At $0.02 per call, that’s $1.60-2.40 per adviser per month, or $20-30 a year.

Multiply that by thirty advisers, and you’re at $600-900 annually just for meeting prep. Add post-meeting file notes, which query the same systems plus a transcription service, and the token cost doubles. Now you’re at $1,200-1,800 a year for one agent doing one job.

The value is real. Advisers get back two to three hours a week. But the token cost scales linearly with the number of advisers, and the savings don’t. A firm with fifty advisers hits $3K in token costs before they’ve automated compliance documentation or onboarding.

Compliance Documentation

SOAs, ROAs, and file notes consume paraplanner time. Cycle times stretch into weeks. An Advice Document Agent drafts these documents from meeting transcripts, the firm’s compliance template, and client data. The agent queries the transcript service, pulls client facts from the CRM, checks the compliance template library, runs a draft through a reasoning model, and formats the output. That’s eight to twelve API calls per document.

A firm producing sixty advice documents a quarter generates 480-720 calls. At $0.02 per call, that’s $9.60-14.40 per quarter, or $38-58 annually. Not a budget-breaker on its own. But stack it on top of meeting prep, add the cost of the reasoning model (which charges per token, not per call), and you’re at $2K-3K a year for two agents.

The firms we talk to don’t balk at $3K. They balk at the trajectory. If token costs grow with usage, what happens when the firm doubles in size? What happens when they add a third agent for client onboarding? The math stops being about this year’s bill and starts being about whether the cost structure is sustainable.

Client Onboarding and KYC

New clients lose momentum during onboarding. Document collection, fact-finding, and risk profiling drag on for 30-60 days. A Client Onboarding Agent runs a guided fact-find, collects KYC docs, and prepares a clean onboarding pack for the adviser. The agent queries the CRM, sends document requests, pulls data from third-party verification services, and compiles everything into a structured file. That’s ten to fifteen API calls per new client.

A firm onboarding forty new clients a quarter generates 400-600 calls. At $0.02 per call, that’s $8-12 per quarter, or $32-48 annually. Again, not material on its own. But onboarding agents touch more external services than meeting prep or compliance agents, and those services charge their own API fees. The token cost is the visible line item. The total cost of automation is higher.

When you add up meeting prep, compliance documentation, and onboarding, a thirty-adviser firm running all three agents is looking at $5K-8K in token costs annually. That’s before platform fees, before engineering time to maintain the integrations, and before the cost of the reasoning models that power the agents. The $70K-200K in annual leakage we see across advisory firms isn’t just wasted paraplanner hours. It’s also the hidden cost of automation that bills by the query.

What Flat-Rate and Self-Hosted Options Look Like

The alternative isn’t to stop automating. It’s to stop renting compute by the sip. Flat-rate AI platforms charge a fixed monthly or annual fee regardless of query volume. Self-hosted models run on your own infrastructure and bill by the server, not the token. Both options shift the cost structure from variable to fixed, which changes the economics for continuous-use workflows.

Flat-rate platforms work best for firms that want to automate fast without standing up infrastructure. You pay a subscription, plug your data sources into pre-built agents, and the platform handles inference, orchestration, and updates. The cost is predictable. The trade-off is less control over the model, the data pipeline, and the compliance posture.

Self-hosted models work best for firms that need air-gapped environments, custom compliance logic, or the ability to fine-tune models on proprietary data. You run the model on your own servers or a private cloud instance. The upfront cost is higher, the engineering lift is real, but the marginal cost of each query drops to near zero. For a firm running hundreds of thousands of queries a year, the break-even point can hit in six to twelve months.

The Palantir-Nvidia stack is a self-hosted option built for enterprises that can’t send client data to a third-party API. It’s overkill for most advisory firms, but the signal is clear: the market is moving away from token billing for continuous workloads. If you’re evaluating AI agents now, ask your vendor how they charge. If the answer is per-token or per-call, model out what your bill looks like at 10x, 50x, and 100x your current query volume. If the number makes you uncomfortable, start looking at flat-rate or self-hosted alternatives.

We built Omni as a flat-rate platform because we kept hearing the same story from advisory firms: the agents worked, the token bill didn’t. Omni Ops agents for meeting prep, compliance documentation, and client onboarding run on a fixed subscription. You don’t pay per query. You don’t pay per adviser. You pay for the platform, and the platform handles the rest. If you want to see what that looks like for your firm, book a 60-min Omni Audit. We’ll map your current workflows, estimate the token cost of automating them, and show you the flat-rate alternative.

How to Evaluate Your Current AI Spend

Most advisory firms don’t know what they’re paying in token costs because the bills are buried in platform subscriptions or billed to a general IT line item. If you’re running AI agents now, or evaluating vendors, here’s how to surface the real cost.

Pull your last three months of invoices from every AI vendor you use. Look for line items labeled API usage, token consumption, or inference credits. Add them up. Divide by three to get your average monthly spend. Multiply by twelve to annualize it. That’s your baseline.

Now model what happens if you double your usage. If you’re running meeting prep for ten advisers, what does the bill look like for twenty? If you’re drafting forty SOAs a quarter, what does eighty cost? If the vendor charges per token, ask them for a cost calculator or a sample bill at 2x, 5x, and 10x your current volume. If they won’t give you one, that’s a signal.

Compare that trajectory to a flat-rate option. If a flat-rate platform charges $2K a month and your token bill is $500 today but $3K at 2x usage, the break-even point is obvious. If your token bill is $200 today and you’re not planning to scale, flat-rate might not make sense yet. The key is to model the future state, not the current state.

For firms that want to explore self-hosted models, the calculus is different. You’re trading subscription costs for infrastructure costs. A private cloud instance running a mid-size language model costs $500-1,500 a month depending on the provider and the instance size. You’ll also need engineering time to set it up, maintain it, and fine-tune it. Budget $10K-20K for the first year, then $5K-10K annually after that. If your token bill is running $5K-8K a year and climbing, self-hosted starts to pencil. If it’s under $2K, it probably doesn’t.

The firms we work with in the $1M-25M band usually land on flat-rate platforms. They don’t have the engineering team to self-host, and they don’t want to become infrastructure managers. They want agents that work, costs that don’t surprise them, and a vendor that understands advisory workflows. That’s the design brief for the AI audit for financial advisory firms. We map your workflows, show you what automation looks like, and give you a fixed-cost option to implement it.

What an Omni Audit Uncovers in 60 Minutes

An Omni Audit isn’t a sales pitch. It’s a structured review of your current operations, the manual work that’s leaking time and money, and the agents that can close the gap. We run it in 60 minutes and deliver three outputs: a workflow map, a leakage estimate, and a build plan.

The workflow map diagrams your current process for meeting prep, compliance documentation, and client onboarding. We ask what systems you use, where data lives, how many steps each process takes, and who owns each step. The map shows you where the bottlenecks are and where an agent can intervene.

The leakage estimate quantifies the cost of manual work. If your advisers spend eight hours a week on meeting prep and notes, that’s $40K-80K in opportunity cost annually depending on what an adviser’s time is worth. If your paraplanners spend three weeks drafting an SOA, that’s $3K-8K per document in direct labor cost. If onboarding takes 45 days and you lose 20% of new clients to friction, that’s $50K-150K in lost revenue. We don’t invent these numbers. We ask you what your time costs and what your conversion rates are, then we do the math.

The build plan shows you what agents we’d deploy, what data sources they’d connect to, and what the timeline looks like. A Meeting Prep Agent typically takes two weeks to configure and test. An Advice Document Agent takes three to four weeks because it needs to learn your compliance templates. A Client Onboarding Agent takes two to three weeks depending on how many external services it needs to integrate with. The build plan includes cost, timeline, and expected ROI.

We’ve run this audit with over 200 advisory firms in the past year. The median leakage estimate is $120K. The median build plan costs $30K-50K to implement and pays back in six to nine months. The firms that move fastest are the ones that already know their processes are broken and just need someone to show them what good looks like.

If you want to see what your firm’s leakage looks like, book my Omni Audit. We’ll spend an hour mapping your workflows, estimating your costs, and showing you what flat-rate automation looks like. No deck, no pitch, just three outputs you can use whether you work with us or not.

Why This Matters Now

Token billing isn’t going away overnight. The hyperscalers have built their businesses on it, and most AI vendors pass those costs through to customers. But the pressure is building. Palantir and Nvidia didn’t launch an air-gapped stack because enterprises love token billing. They launched it because enterprises are tired of variable costs that scale faster than value.

Advisory firms are early in the automation curve. Most firms haven’t deployed agents yet. The ones that have are still figuring out what works. That’s the window. If you evaluate AI options now, while the market is still sorting out pricing models, you can lock in a cost structure that makes sense for continuous-use workflows. If you wait until token billing is the default and flat-rate options are niche, you’ll have less leverage and fewer alternatives.

The firms that win in the next three years won’t be the ones with the most AI. They’ll be the ones with the most economical AI. The ones that automate meeting prep, compliance documentation, and client onboarding without bleeding $50K a year in token costs. The ones that own their compute instead of renting it by the query.

We built Omni to be that option. Flat-rate pricing, pre-built agents for advisory workflows, and a 60-minute audit that shows you what automation looks like before you commit a dollar. If you’re ready to see what that looks like for your firm, start with the audit. If you want to read more about how AI is reshaping advisory operations, explore our insights library or dive into Omni Ops to see the agents we’ve already built.

Token billing is breaking. The firms that move now will own the economics. The ones that wait will rent them.