Fix the process before you buy the AI tool

Episode 363 of the WealthTech Today podcast, published September 17, asks advisory firms to spend their automation money in a different order than the purchase order suggests: fix the foundation before you automate the mess. The guest is Jennifer Maruca, a senior consultant with Ezra Group and founder of Platform Value Group, and the episode frames her as having spent years helping wealth management firms see a problem that rarely shows up as a line item on any budget.

The episode description calls that problem a gap and leaves the second half of the thought unfinished, which hands the argument to the title; the title is enough. Automation runs whatever process it is handed, faster and in greater volume, and where the process is the mess the tool scales the mess. It may also be the easiest advice in the industry to nod past, since foundation work produces no demo.

The shelf is loud enough to make skipping ahead tempting, and Ezra Group scored 69 AI tools for advisors on an autonomy scale in a directory this publication covered this month, a checklist built precisely because the language vendors use in this category has outrun what a firm can verify from a sales call. PWD has argued that the adoption gap is now a supervision split: firms cannot buy back client time until they can assure the output. Maruca's sequencing pushes that argument one step earlier, because assurance begins with a process somebody can describe, and a firm that cannot describe its own workflow has nothing to hand a reviewing principal, never mind a model.

Sequence matters because the invoice arrives either way: this publication reported last month that wealthtech contracts pass AI compute costs through on usage terms, and the meter turns whether or not anyone uses the feature. A firm that automates a process it never fixed pays for the compute and then pays again to unwind what the automation did with it. The point is not unfamiliar at smaller scale: the AI Prompt Lab made it about seasonal client email, where a fixed skeleton has to exist before per-client variables or a compliance rail mean anything.

The judgment call is where the readiness work belongs on the budget, and it is not the technology line: data cleanup, workflow documentation, exception routing, decisions about which outputs a human reviews—that is operations work priced in staff hours, and no vendor sells a license for it. Watch whether firms fund it in this planning season alongside the software, or meet it a year from now as a remediation project, with the vendor who sold the license in the room.