Build agents or buy the layer: the math that rarely works in-house

Six months ago, the board approved "building an AI team." They hired two engineers, provisioned GPUs, started building. Today they have a prototype that works on Tuesdays, a roadmap that only grows, and a fixed cost that never stops. And across the street, the competitor who chose to buy is already serving clients.

The "build it in-house" math looks good on the slide and rarely closes in practice. Not because building an agent is impossible — models are accessible. It's that the cost isn't building. It's maintaining. An agent that acts on money demands continuous evaluation, model updates, hallucination monitoring, and an audit trail — all under regulation that keeps moving. You don't hire that once; you carry it forever.

The numbers tell the story: only 11% of companies got agents into production (Deloitte, 2026). Most projects die between the prototype and daily use — exactly in the part no one budgets: running safely, at scale, under compliance.

Buying the layer flips the math. Instead of building AI engineering, MLOps, and governance from scratch, the brokerage embeds a ready layer — infra, models, orchestration, and guardrails included — under its own license. It integrates with what already runs; it doesn't replace. The internal team goes back to what generates revenue, not maintaining AI plumbing.

And here's the detail that makes the layer cheaper than it looks: it doesn't serve only you. The same infrastructure serves the direct investor and the brokerage that white-labels it. Direct use proves the product; distribution through brokerages spreads the cost. You pay for a layer amortized across many — not a project only your firm funds.

That doesn't mean building never makes sense. It does — when AI is your end product. For a brokerage, AI is a means: the end is serving and retaining clients. Spending scarce capital rebuilding what you could embed is trading time-to-market for an engineering trophy.

For anyone deciding, the test is honest: over the next three years, do you want to be maintaining an AI team — evaluating models, answering the regulator, chasing hallucinations — or serving more clients? The answer usually decides between build and buy.

→ 20-min diagnostic of your operation: lumes.io/broker

Lumes provides technology and educational content. Past performance is not a guarantee of future results. Lumes is not a securities advisor or a brokerage. Consult a qualified professional before investing.