OpenAI just told the enterprise world what many financial advisory firms are learning the hard way: AI agents can rack up unpredictable bills if you don’t set guardrails. The company announced new cost-control features for businesses after seeing token consumption spike as agents handle more complex workflows. For advisory firms running AI tools to draft statements of advice, prep client meetings, or automate compliance documentation, this admission is a warning sign. If you don’t know what each client interaction costs in tokens, you’re flying blind.

The problem isn’t that AI is expensive. The problem is that most firms have no idea what they’re spending per task. An agent that pulls portfolio data, drafts a meeting brief, and logs file notes might burn through 50,000 tokens for one client review. Multiply that by 200 client meetings a month and you’re looking at 10 million tokens. At current rates, that’s $300 to $600 a month for one workflow. Add in advice document generation, onboarding fact-finds, and ad-hoc queries from advisers, and the bill climbs fast. Firms that rolled out AI tools six months ago are now seeing monthly invoices they can’t explain.

The real issue is visibility. Most advisory firms don’t track which workflows consume the most tokens or which advisers are triggering high-cost queries. They see a lump sum on the invoice and assume it’s the cost of doing business. But when you break it down by agent and by task, patterns emerge. Meeting prep might cost $2 per client. Advice document generation might cost $15. A complex onboarding workflow with document parsing and risk profiling might hit $40. Without that granularity, you can’t make informed decisions about where to set usage caps or which workflows to optimize.

Why Token Spend Spirals in Advisory Firms

Financial advisory firms are particularly vulnerable to runaway AI costs because the workflows involve large documents, repeated queries, and context-heavy tasks. A Meeting Prep Agent doesn’t just pull a few data points. It reads the client’s last three meeting notes, reviews portfolio performance, checks goal progress, and generates a one-page brief. That’s a lot of input tokens. If the agent re-reads the entire client file every time instead of caching key data, the cost doubles.

Advice document generation is even worse. An Advice Document Agent drafts a statement of advice from a meeting transcript, the firm’s compliance template, and the client’s fact-find. A typical SOA runs 20 to 40 pages. The agent might process 30,000 input tokens and generate 15,000 output tokens for a single document. If the adviser asks for revisions, the agent re-processes the entire context. Three revision rounds can push the cost to $50 or more for one SOA. Multiply that by 50 advice documents a month and you’re spending $2,500 just on drafting. That’s not counting the paraplanner time saved, but it’s a real line item that needs oversight.

Client onboarding is the third cost driver. A Client Onboarding Agent runs a guided fact-find, collects KYC documents, parses PDFs, and prepares an onboarding pack. Document parsing is token-intensive. A 10-page super statement might require 8,000 tokens to extract balances and asset allocations. If the client uploads five documents, the agent burns through 40,000 tokens before it even starts drafting the onboarding summary. Firms that onboard 20 new clients a month can see $800 to $1,200 in token costs just for this workflow.

The common thread is that these workflows involve large context windows and multiple steps. Every time an agent reads a document or generates a long output, the token count climbs. If the firm hasn’t set usage caps or optimized the agent’s prompts, costs spiral. OpenAI’s admission that enterprises need better controls is a signal that the industry is waking up to this reality. The firms that audit their AI spending now will have a cost advantage over those that wait until the bill becomes unsustainable.

What Cost Control Actually Looks Like

Cost control for AI agents isn’t about cutting features. It’s about knowing what you’re spending and making intentional trade-offs. The first step is tracking token usage by workflow and by user. Most AI platforms don’t surface this data by default. You need to instrument your agents to log every query, the input and output token counts, and the user who triggered it. Once you have that data, you can see which workflows are expensive and which advisers are heavy users.

The second step is setting usage caps. You don’t need to throttle every query, but you do need guardrails for high-cost tasks. For example, you might allow unlimited Meeting Prep Agent queries because they’re cheap and high-value. But you might require approval for Advice Document Agent revisions beyond the second round. Or you might cap the Client Onboarding Agent at three document uploads per client unless the adviser requests more. These caps prevent runaway costs without blocking legitimate work.

The third step is optimizing prompts and caching strategies. Many agents re-read the same context on every query. A Meeting Prep Agent that pulls the client’s entire file history every time is wasting tokens. A better approach is to cache the client’s profile and only pull recent updates. Similarly, an Advice Document Agent that re-processes the compliance template on every draft is inefficient. Cache the template once and reference it. These optimizations can cut token usage by 30% to 50% without changing the output quality.

The fourth step is approving high-cost queries before they run. For tasks that involve large documents or complex reasoning, you can configure the agent to estimate the token cost and ask for confirmation before proceeding. An adviser who knows a query will cost $20 might choose to simplify the request or handle it manually. This puts the cost decision in the hands of the person who understands the business value.

Firms that implement these four steps typically see their AI costs stabilize or drop even as usage grows. The key is treating AI spend like any other operational cost. You wouldn’t let advisers expense $500 dinners without approval. You shouldn’t let them trigger $50 AI queries without visibility. See Omni for financial advisory firms to understand how we build cost tracking into every agent workflow.

The Omni Approach to Token Spend

We build cost visibility into Omni agents from day one. Every query logs the token count, the user, the workflow, and the timestamp. Firms get a dashboard that shows spending by agent, by adviser, and by client. You can drill down to see which tasks are expensive and set caps at the workflow level. This isn’t an add-on feature. It’s how we design every agent because we know cost control matters as much as automation.

Our Meeting Prep Agent uses a caching strategy that reduces token usage by 40% compared to a naive implementation. Instead of re-reading the client’s full file history, it caches the profile and only pulls updates since the last meeting. The agent still generates a comprehensive brief, but it does so with fewer input tokens. For a firm running 200 meetings a month, this optimization saves $200 to $300 in token costs without any change to the adviser experience.

The Advice Document Agent includes a revision budget. The first draft is always approved. The second revision is automatic. The third revision requires the adviser to confirm the change is worth the cost. Most advisers never hit the cap because two rounds of revisions are enough for a clean SOA. But the cap prevents edge cases where an adviser asks for five rounds of tweaks and racks up $100 in token costs for one document. This simple guardrail has saved our clients thousands of dollars over six months.

The Client Onboarding Agent caps document uploads at three per client by default. If a client needs to upload more, the adviser can approve it with one click. The cap doesn’t block legitimate work, but it prevents scenarios where a client uploads 15 PDFs and the agent burns through 150,000 tokens parsing them all. For firms onboarding 20 clients a month, this cap alone saves $400 to $600 in unnecessary token costs.

We also surface cost estimates before high-cost queries run. If an adviser asks the Advice Document Agent to regenerate an SOA with major changes, the agent estimates the token cost and asks for confirmation. The adviser sees “This query will cost approximately $18” and can decide whether to proceed or simplify the request. This transparency puts cost control in the hands of the people who understand the business value of each task.

These features aren’t theoretical. They’re live in every Omni deployment. Firms that run the AI audit for financial advisory firms get a breakdown of their current AI spending and a projection of what it will look like with Omni’s cost controls in place. The typical savings range from 30% to 50% of current token spend, even as the firm automates more workflows. That’s the difference between AI being a manageable line item and AI being a budget risk.

What an Audit Reveals About Your Current Spend

Most advisory firms don’t know what they’re spending on AI tokens until they run an audit. The invoice shows a lump sum, but it doesn’t break down which workflows are expensive or which advisers are heavy users. An Omni Audit takes 60 minutes and produces three outputs: a workflow map, a cost breakdown, and a savings projection. No deck, no sales pitch. Just the numbers.

The workflow map shows which tasks your firm is automating and how much each one costs in tokens. Meeting prep might be $400 a month. Advice document generation might be $2,000. Client onboarding might be $800. The map also shows which workflows are inefficient. If your Meeting Prep Agent is burning through 80,000 tokens per query, that’s a red flag. A well-optimized agent should use 30,000 to 40,000 tokens for the same output.

The cost breakdown shows spending by adviser and by client. You might discover that one adviser is triggering 40% of your AI queries because they’re using the tool for every client interaction. That’s not necessarily a problem, but it’s information you need. You might also discover that a handful of high-complexity clients are driving 20% of your token costs because their files are large and their workflows are custom. Again, that’s not a problem if the revenue justifies it. But you need visibility to make that call.

The savings projection estimates what your costs would look like with Omni’s caching, caps, and approval workflows in place. For a firm spending $4,000 a month on AI tokens, the projection might show $2,500 with optimizations. That’s $18,000 a year in savings without cutting any automation. The projection also shows what your costs would look like if you added new workflows. If you’re not automating advice documents yet, the projection estimates the token cost and compares it to the paraplanner time saved.

Firms that run the audit typically make one of three decisions. Some realize their current AI spend is fine and they just need better visibility. They implement cost tracking and move on. Others realize they’re overspending on inefficient workflows and they switch to Omni to cut costs. A third group realizes they’re underspending because they’re not automating enough high-value tasks. They add new agents and the cost increase is more than offset by the time saved. All three outcomes are good because they’re informed decisions based on real data.

Book a 60-min Omni Audit to see your current AI spending broken down by workflow and adviser. You’ll walk away with a cost projection and a plan to optimize your token usage.

The Real Cost of Not Auditing

The firms that don’t audit their AI spending fall into one of two traps. The first trap is runaway costs. They roll out agents without usage caps, and six months later they’re spending $8,000 a month on tokens. The ROI is still positive because the agents save more in labour than they cost in tokens, but the spend is unpredictable and it’s growing faster than revenue. Eventually the CFO asks why the AI bill is higher than the software bill, and the firm scrambles to implement controls after the fact.

The second trap is underutilization. The firm is so worried about costs that they throttle agent usage or avoid rolling out new workflows. Advisers stop using the Meeting Prep Agent because they don’t want to trigger high-cost queries. Paraplanners avoid the Advice Document Agent because they’re not sure if it’s worth the token spend. The firm ends up paying for AI tools but getting minimal value because the team is afraid to use them. This is worse than overspending because the firm is leaving time savings on the table.

Both traps are avoidable with an audit. The audit gives you the data to set intelligent caps and the confidence to roll out new workflows. You know what each task costs, you know what the ROI is, and you can make trade-offs based on business value. Advisers use the tools without hesitation because they know the firm has guardrails in place. The CFO is happy because the AI budget is predictable and justified by labour savings.

The typical financial advisory firm in the $1M to $25M revenue range spends $70K to $200K a year on manual work that AI agents could handle. Meeting prep, advice document generation, and client onboarding account for the bulk of that cost. If you automate those workflows with Omni agents and implement cost controls, you save 60% to 80% of that labour cost while keeping token spend under $30K a year. That’s a net savings of $50K to $150K. The firms that audit their AI spending now will capture that savings. The firms that wait will either overspend on tokens or underspend on automation.

Next Steps

If you’re running AI agents in your advisory firm, you need to know what you’re spending per workflow and per adviser. OpenAI’s admission that enterprises need better cost controls is a signal that the industry is moving toward transparency and accountability. The firms that get ahead of this trend will have a cost advantage over those that treat AI as a black box.

Start with an audit. Sixty minutes, three outputs, no deck. You’ll see your current spending broken down by task, a projection of what it would look like with Omni’s cost controls, and a plan to optimize your token usage. From there, you can decide whether to implement tracking on your current tools or switch to Omni agents that have cost visibility built in. Either way, you’ll have the data to make an informed decision.

The alternative is to keep paying invoices you can’t explain and hope the costs don’t spiral. That’s not a strategy. That’s a risk. Book my Omni Audit and take control of your AI spending before it becomes a budget problem.

For more on how we build cost-efficient agents for advisory firms, explore our Omni Ops platform and read through our insights on AI implementation. The firms that treat AI spend like any other operational cost will be the ones that scale automation profitably.