What five interviews revealed about how Secondary's work really flows — and where AI assistance can remove the manual load without touching the judgment calls.
A small, senior team running Encompass-centric, deadline-driven workflows.
▸ Click any card to read the full insight.
Phyllis estimates 9 out of 10 guideline questions are already answered in the GMFS intranet guideline, yet closers and retail call her instead of looking. Molly maintains a fragmented set of 4–5+ product guides with no single owner and format-constrained PDFs, and is already building an investor-overlay shared drive. The shared gap isn't knowledge — it's a single, current, searchable place to find it.
Molly works with 22 investors who disseminate changes inconsistently — emails get lost, misrouted, or blocked by suspected spam filters — forcing monthly website spot-checks to catch what was missed. The same unreliable stream drives rate-sheet LLPA changes, overlay rules, and guide updates. A reliable capture-and-flag layer over investor announcements would de-risk several downstream workflows at once.
James's investor recertification packets and report creation, Phyllis's trail-docs chasing, and Molly's announcement processing all share the same shape: most of the time goes to collecting data from systems and chasing people, while the actual decision is quick. James calls bothering executives for data that "should be quickly and easily available" a top pain point. AI that does the gathering and produces a first draft would remove the bulk of the effort while leaving the decision with the human.
Phyllis's weekly aged warehouse list and daily pipeline monitoring, James's ad-hoc report creation, and Kyle's rate sheets and active broker list all follow the same loop: pull from Encompass (or Optimal Blue), filter by stable rules in Excel, and email a fixed recipient list. These pulls have explicit, written-downable rules (e.g., remove portfolio loans and anything ≤ 20 days) and are prime automation candidates.
Kyle estimates over 50% of Optimal Blue profile-change requests could be made by the requester themselves but get routed to the lock desk anyway. James's report requests routinely arrive missing key parameters (time period, units vs UPB, geography, channel, lock/funded), forcing back-and-forth. Smart intake — triage, clarification prompts, and routing — would reclaim significant time on both desks.
Molly has no automated way to compare daily investor rate sheets day-over-day to catch LLPA changes, and when investors don't spell out a change she does full side-by-side grid comparison. Kyle must fully restart the nine-rate-sheet build whenever rates move intraday. Automated change detection between document versions is a repeatable capability multiple people need.
Knowing each investor's overlays, quirks, presentation preferences, and tolerance for shipping with minor missing docs is unwritten and experience-based — Phyllis, Molly, and James all flag it as where new people struggle. Capturing this into a structured, maintained reference is both a knowledge-base play and a succession-risk mitigation.
▸ Click a row for detail. Dots mark who raised each thread.
| Thread | Prevalence | Phyllis | Molly | Kyle | James |
|---|---|---|---|---|---|
| Encompass/OB → Excel → filter → email | MOST | ||||
| Keeping guides, rate sheets & overlays in sync | MOST | ||||
| Chasing people & documents to assemble deliverables | MOST | ||||
| Manual day-over-day / version comparison | SEVERAL | ||||
| Under-specified or self-servable inbound requests | SEVERAL | ||||
| Unreliable investor dissemination (upstream risk) | SEVERAL | ||||
| Comfort with AI already present | A FEW |
Each cluster recurs across multiple people — so one build helps several workflows at once.
All Encompass → Excel → filter → email loops with stable rules.
One need: a single, current, searchable source of overlays, guidelines & guides.
Spotting deltas and routing/gathering — not judgment.
Flags changes that are currently caught only by manual side-by-side grid review.
The aged warehouse list, trail-docs 90-day chasing, and committed-not-shipped lock-expiration alerts — reusable for James's recurring reports.
Directly powers Phyllis's Q&A and Molly's announcement/guide workflow — building on Molly's overlay drive and James's guides folder.
Ship-vs-hold, credit-committee risk, ambiguous compliance toggles — AI drafts and gathers; people decide. And: reschedule Brandon Genre to complete coverage of the MCT/trading and margin side.
Two connected capabilities for the whole desk: a single grounded investor knowledge layer, plus scheduled AI skills that run the recurring pulls, comparisons, and follow-ups. Four skills: the pipeline-lists skill, the rate-sheet delta skill, the recert-packet drafter, and the intake-triage skill.
The strategy: a structured, maintained folder per investor (overlays, guidelines, current guides, rate-sheet tabs, recert requirements, contacts) — consolidating Molly's overlay drive and James's guides folder into one source the AI skills read from and write back to.
Stand up the structured investor folder; point a grounded Q&A skill at the GMFS guideline and overlays — serving Phyllis's questions and Molly's announcement work.
Ship the rate-sheet day-over-day comparison and the scheduled pipeline lists with drafted follow-ups — the fastest, lowest-risk wins.
Add the recert packet drafter and request-intake triage, then layer announcement-to-guide-and-overlay drafting on the now-populated folder.
Reschedule and fold in Brandon's MCT/trading and margin workflows so the desk's coverage is complete.
This synthesis is a mirror, not a mandate. Before anything gets built, we want the team's read on it.
Anything we overstated, missed, or got backwards about how the work actually flows?
Rate-sheet / LLPA comparison, or the scheduled pipeline lists with follow-up emails — where's the most pain right now?
Who should own it, and what belongs in the folder-per-investor structure on day one?
Did we draw the AI-drafts / you-decide boundary in the right places?