Schwab's Claude AI deal sidelines 11,000 small firms due to seat minimums
A new 20-seat minimum for Anthropic's Claude enterprise plan effectively excludes smaller investment firms from a major new AI partnership.
The short version
- Anthropic requires a 20-seat minimum for its Claude enterprise plan, sidelining over 11,000 small RIAs.
- The annual cost floor for the software is $4,800, which may be too high for firms with fewer than $100 million in assets.
- Industry experts suggest the tool may be 'read-only,' meaning it can view data but cannot perform actions.
- Anthropic is exploring potential seat-count reductions for smaller firms, but no immediate path exists.
Why it matters
This shift reveals the practical barriers to AI adoption in finance, where high entry costs can create a 'digital divide' between large and small firms. It also highlights the technical limitations of current AI connectors that can view data but cannot yet execute complex workflows.
The pricing mismatch
Schwab recently announced a major partnership with Anthropic to provide its Claude AI model to 16,000 registered investment advisers (RIAs). The deal was initially presented as a universal access point for all firms within the Schwab network. However, new details regarding the enterprise plan have complicated that promise.
Anthropic has imposed a 20-seat minimum for its Claude enterprise software. This requirement effectively sidelines more than 11,000 of the firms Schwab originally intended to include. Most of these smaller firms manage less than $100 million in assets and often consist of only one or two advisors.
Because of this minimum, small firms face an annual cost floor of $4,800 for software features they may not fully utilize. Sources suggest the initial announcement of universal access may have been rushed because Anthropic named all of its charter partners in a single news dump on September 14.
Technical hurdles for advisors
Beyond the high entry cost, industry experts warn that the Claude-for-Financial-Advisors (CFFA) tool may have limited functionality. Some analysts point out that the current connector is 'read-only,' meaning the AI can look at data but cannot act on it. This creates a gap between the marketing hype and the actual utility for advisors seeking automation.
The technical challenge lies in how different systems speak to one another. Because there is no shared data layer underneath, the AI has to reconcile different client IDs and account structures from various vendors on the fly. If the data from a CRM, a planning software, and a custodian do not match perfectly, the AI may either guess or stop entirely.
Some large firms are choosing to build their own internal data layers first to solve this problem. By integrating data below the AI level, these companies can remain agnostic to specific custodians and software. This approach allows them to eventually put AI on top of a unified system to achieve true intelligence and automation benefits.
Throttling the demand
Some observers believe that the high seat minimum and costs serve a strategic purpose for Schwab and Anthropic. By setting a high bar for entry, the companies can avoid a 'stampede' of demand that they might not be able to satisfy immediately. This allows them to manage the rollout and ensure the infrastructure can handle the load.
Schwab CEO Rick Wurster noted that RIAs are already in a pilot phase with Claude and that the firm is looking to expand to other large language models (LLMs) in the coming months. This suggests a cautious approach to integrating AI into the financial workflow. The focus appears to be on the large RIAs who are further along in their AI journey.
While Anthropic does offer a cheaper 'teams' plan, the company does not recommend it for registered investment advisers. They state that the enterprise plan is necessary because it includes the audit logs required for proper recordkeeping. This requirement reinforces the high cost for smaller firms who do not have the scale to absorb the $4,800 annual fee.
What comes next for RIAs
The situation remains uncertain for the 11,000 smaller firms currently excluded from the primary bundle. Anthropic executives have expressed a willingness to explore lowering the seat count for smaller firms in the future. However, as of now, there is no confirmed path or timeline for making the tool more accessible to solo practitioners.
Advisors who do manage to secure the tool may find that the actual costs could be significantly higher than the starting price. Experts warn that token costs—the way AI models charge for the amount of data processed—could cause the annual fee to double, triple, or even increase tenfold. This could lead to a shift in sentiment once firms see the final bills.
For now, the focus remains on the large 'mega-firms' who have the resources to navigate these enterprise requirements. They will likely be the first to see the benefits of the Claude integration, while smaller firms may have to wait for more affordable, specialized tools to emerge. The reality of manual work remaining will be the ultimate test of the tool's value.
What to watch
- Whether Anthropic announces a specific 'small firm' tier for the CFFA plan.
- How many large RIAs actually find the 'read-only' functionality useful enough to justify the costs.
Try it
Check if your firm's current software stack has a unified data layer or if you have to manually reconcile data between your CRM and custodian.
Sources and links
- Officialclaude.com/resources/articles/claude-for-financial-advisors claude.com
- ReportingRIABiz: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by RIABiz · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
