Skip to Content

Vet and Choose Tools

This is the fourth step in The Adoption Path. Steps one through three produced a use-case list, a duty map, and a data matrix; this step is where a specific tool either clears the bar those three already set, or doesn’t. Vetting applies what the firm has already decided — it doesn’t decide anything fresh for each new tool. A firm that vets tools before doing the first three steps ends up evaluating features instead of fit, and picks tools it can’t actually justify using later.

Why it matters

Comparing tools by capability or price, before the use case and the data question are settled, leads to a policy in Write the AI Policy that has to justify tools already bought rather than the other way around. It also means every tool gets vetted on its own, inconsistent basis — one adviser reads the terms carefully, another goes on reputation, and the firm ends up with no single standard it can point to for how a tool was approved. This step exists to make that standard the same one every time: the matrix, the duty map, and the use case, checked in the same order.

What actually happens

Start from the matrix, not the marketplace

Before looking at any tool, know which cell of the data matrix from Sort Your Data it needs to fill. A tool that only needs to handle public or internal information clears most of the bar before you have read a single terms page; a tool meant to touch client-identifying information starts the evaluation somewhere much stricter. Starting from the matrix instead of a tool’s feature list keeps the question narrow: not “is this a good tool,” but “is this tool allowed to do the one thing on the use-case list it is meant to do.”

Read the tool’s own terms, and get them in writing

Read the tool’s own terms-of-service, data-handling, and privacy pages — not a summary of them, and not a vendor’s sales page characterizing them. Look for five things specifically: whether inputs are used to train the model, the retention period, whether the vendor will sign an agreement acting as the firm’s service provider, whether subprocessors are disclosed, and what happens to data on cancellation. The Vetting a Vendor checklist walks through these five one at a time. A public policy page can change unilaterally, whenever the vendor updates it; a signed agreement is what the firm can actually enforce and point to in an exam. If a tool is going to touch client-identifying or restricted information, a signed agreement is not optional — a public terms page alone does not qualify.

Weigh advisor-specific platforms against general-purpose tools with a signed agreement

This guide’s default scenario is a general-purpose tool used under a signed agreement, since that is the more common starting point for a small firm. That does not mean it is always the better answer. An advisor-specific platform that has already cleared these questions as part of its own product — built specifically to handle client-identifying advisory data — can be the simpler choice for a given use, and this step should weigh it on that basis rather than favoring the general-purpose path on principle. Compare the two the same way: against the matrix, against the terms, against whether the commitments are actually in writing.

Decide who can approve a new tool

Name who in the firm can actually approve a new tool for use, and require the approval to reference the use-case list, the duty map, and the data matrix by name — not just “looks fine to me.” An approval that does not point to those three documents is not really an approval; it is a guess made to sound like one.

Set a re-check trigger

Approval is not permanent. Set a trigger for re-checking a tool: a material change to its terms, a new use case proposed for a tool already approved for something else, or a fixed review interval — whichever comes first. A tool that was fine for summarizing internal notes is not automatically fine once someone starts using it for client-facing drafts; that is a new use case against the matrix, not an extension of the old approval.

Questions to ask yourself

  • For the tool in front of you, which cell of the matrix does it actually need to fill?
  • Have you read the tool’s own terms yourself, or are you relying on someone else’s summary of them?
  • Is there a signed agreement, or only a public policy page that could change tomorrow?
  • Would an advisor-specific platform clear these same questions more simply for this particular use?
  • If this tool’s use expanded next month, would that trigger a new check, or would nobody notice?

Common mistakes

  • Approving a tool because a colleague or a review site said it was compliant, rather than checking the tool’s own terms.
  • Treating a public terms page as equivalent to a signed agreement.
  • Approving a tool for one use and letting its use quietly expand without a new check.
  • Choosing a tool for its features before checking what cell of the matrix it actually needs to fill.
  • Never re-checking a tool’s terms after the initial approval.

Which Firm Functions apply

  • Vendor Oversight: This step is where vendor oversight actually starts — the initial check a tool has to pass before it is in use at all.
  • Data Privacy and Security: The matrix this step starts from is a data-privacy tool, and the questions about training and retention are privacy questions.
  • Compliance and Supervision: Naming who can approve a tool, and requiring the approval to reference the firm’s own documents, is what makes this step supervisable.

At a larger firm, tool approval usually runs through a vendor management or IT security process the adviser does not own. The job is not to vet the tool independently; it is to bring the specific use case, the data tier involved, and the duties it touches to that existing process, and get its answer in writing before using anything it has not cleared.

Next step

Write the AI Policy: With tools approved against the matrix and the duty map, put the firm’s rules for using them in writing.