Maintaining Your Vault - AdviserGPT-Adviser Class AI for Investment Managers

Maintaining Your Vault

Most RFP tools make you the content manager

Without automation, keeping a shared answer library usable and up-to-date with your firm’s most recent compliance-approved language is manual, invisible work: scrolling through hundreds of Q&A pairs to catch near-duplicate answers before they drift out of sync, remembering which questionnaires are old enough to retire, and hunting down every document that mentions an AUM or performance number the moment it changes. It’s tedious to fall behind, and the problem gets exponentially worse with more strategies, more clients, and more documents.

AdviserGPT keeps your Vault up-to-date and compliant with three background Vault AI agents. You describe the rule or the change once; the agent does the repetitive part and queues its work for your approval - you stay the reviewer, not the content manager.

Where to find your vault agents
AdviserGPT Vault \u2014 Agents menu open

Open the Vault AI Agents from the top right corner of your web app Vault view. Agent Activities, next to Agents, is where you review proposed changes generated by your agents. No changes are made to your Vault content until you approve.

Three agents, three kinds of busywork removed

Instead of keeping a mental list of which DDQs are “probably too old to trust,” describe the rule once and let the Archive agent apply it consistently — on a one-time run or a recurring schedule.

AdviserGPT Vault — archive agent123

Click any numbered pin on the screenshot — or any card below — to explore that feature

1

Describe the Rule

Type a rule like “Archive questionnaires older than 18 months” — and the agent generates the matching documents for you. No manual filtering through documents one by one.

2

Archive Automatically & Daily Auto Run

Turn on Archive Automatically to route matches to Agent Activities for approval, and Enable Daily Auto Run so the rule works on its own — you don’t need to run it manually.

3

Preview, then Run

Preview Matching Documents shows exactly what the rule will catch, so you can confirm scope before you hit Run.

Two analysts answering the same question six months apart is how libaries quietly drift out of sync. Rather than eyeballing hundreds of pairs for overlap, the Find Duplicates agent scores similarity for you and flags what’s worth a second look.

AdviserGPT Vault — duplicates agent123

Click any numbered pin on the screenshot — or any card below — to explore that feature

1

Document & Tag Filters

Optionally filter the scan to specific documents or tags — or leave both empty to scan your entire Vault. Either way, the agent does the cross-referencing, not you.

2

Similarity Threshold

Drag the slider to set how close two answers need to be (e.g. 0.92) to be grouped as duplicates — one setting applies consistently instead of judging “close enough” pair by pair.

3

Run Once or Schedule

Run a one-time scan now with Run Now, or set a future Run Date — duplicate detection runs in the background even when no one remembers to check. Duplicate are added to Agent Activities for review.

A change in AUM, new performance numbers, a portfolio manager joins the team — normally that means manual updates across hundreds of Q&A pairs. The Firm Update agent finds and updates each instance from a single description so you can review and approve.

AdviserGPT Vault — firmupdates agent12

Click any numbered pin on the screenshot — or any card below — to explore that feature

1

Describe the Change

Describe the change in plain language — the agent finds and suggests replacement text everywhere it applies, instead of you searching through Q&A pairs and documents for every instance.

2

Run Firm Update

Suggested edits are queued in Agent Activities for your review — one update propagates firm-wide instead of dozens of manual edits, but nothing changes without your approval.

What stays manual? Every agent above stops at the review step in Agent Activities so you can review and approve. What disappears is the busywork of finding what needs attention in the first place.

You review - the agents do the janitorial work

ArchiveNo more tracking which questionnaires are too old to trust — describe the rule once, schedule it, done.
Find DuplicatesNo more eyeballing hundreds of pairs for drift — similarity scoring surfaces the ones worth merging.
Firm UpdatesNo more hunting document by document for one changing fact — one description updates every instance.

Every agent’s output lands in Agent Activities for your approval — automation removes the librarian work, not your final say.

See AdviserGPT in action