CSM
- · Case nomination
- · Participant completion
- · Implementation follow-through
- · Outcome capture
- · Normal customer implementation support
Internal preview · not linked publicly
AI creates the hypothesis. Omar validates the diagnosis. Everything below is seeded prototype data.
Case TC-0148 · Jacob (CSM) · May 2
Firm
Ellis & Co. CPAs
Primary contact
Jordan Ellis — Managing Partner
Problem the firm wants solved
They believe they need another preparer before next season.
Why now
Hiring decision lands in June. Partner review hours are already the ceiling on throughput.
Workflow
Tax preparation
Case rationale
Clear owner sponsorship, one workflow, and a measurable decision attached to it. If the constraint isn't headcount, the transformation is visible within one season.
Potential value
Case TC-0148 · Prepared for session on May 14
Firm context
Tax-first firm, Southeast. Assigned CSM: Jacob. Respondent: Jordan Ellis, Managing Partner. Workflow in scope: tax preparation.
Stated problem
Preparation takes too long and review is a bottleneck. Leadership believes they are short-staffed heading into season.
Why it matters now
A hiring decision is due in June, and the partner cannot absorb more review hours next season.
Current workflow
Organizer sent → client returns partial documents → admin files what arrives → preparer starts on an incomplete file → missing items surface mid-preparation → return parks in an open queue → partner reviews and often re-does sections.
Actual operators
Flag — respondent is not an operator
The respondent does not perform this workflow. The described process is leadership's understanding; the closest operator is not yet in the session. Closest to the work: Dana Ruiz — Tax Manager. Omar and the CSM decide whether that person needs to join the session.
Firm's perceived constraint
Firm's AI expectation
Work is entering technical production incomplete. Preparation and review absorb the cost of an unfinished intake step, which reads inside the firm as a staffing shortage.
Supporting evidence
Three validation questions
Potential transformation direction
Likely direction, not a prescription: an enforced completeness gate before preparation, with an agent reconciling document sets against a return-type requirement list and chasing gaps.
Proposed success metric
Firm-stated evidence: average review time 50 → 25 minutes, and the partner out of routine review.
Implementation owner
Dana Ruiz — Tax Manager
Assigned CSM
Jacob
Pattern tags
Captured live by Omar with the firm
Agreed future-state workflow
Requirement list per return type → agent reconciles received documents against the list → automated gap requests with a deadline → file flips to 'Prep Ready' only when complete → preparer picks up only Prep Ready work → reviewer scores technical items, not completeness.
Role of AI
Reconcile documents against requirements, draft and send gap requests, track outstanding items, flag files stalled beyond 5 days.
Role of humans
Own the requirement standard, approve exceptions, hold the gate, handle client escalations.
Primary implementation action
Publish requirement lists for the top 3 return types and turn on the completeness gate for one pod.
Implementation owner
Dana Ruiz — Tax Manager
Success metric
% of returns entering preparation complete, and average review minutes per return.
Target date
June 30
Assigned CSM
Jacob
A case isn't ready for CSM implementation until all seven are present
The CSM should be able to act on this without Omar reinterpreting the session afterward.
30 days after session · captured by Jacob
What was actually implemented
Requirement lists live for 1040 and 1120-S. Gate enabled for one pod of 4. Automated gap requests running.
What happened to the success metric
Complete-at-intake 55% → 78%. Average review time 50 → 34 minutes.
What blocked progress
Partner still pulls incomplete files forward for a handful of legacy clients.
Needs to return to coaching
No
One required check at 30 days. Earlier updates only when something materially changes. Target time to complete: about two minutes.
Who owns what across the loop
Visibility, not approval. Leadership is never a required approver on an individual case.
Illustrative aggregate across pilot cases — seeded data
Chain compared across cases: nominated problem → firm-stated problem → firm-perceived constraint → AI hypothesis → Omar-confirmed constraint → expected AI role → actual transformation → implementation status → outcome.
63%
of firms named a different constraint than the one confirmed in session
44%
of intakes were completed by someone who does not perform the workflow
71%
of implemented cases moved their agreed success metric
Most common nominated problems
Most common firm-stated problems
Confirmed root constraints
Workflow areas
AI expectation vs appropriate AI role
Who actually performs the workflow
Friction categories
Case outcomes
Reusable insight cards
Firms often frame labor shortages as preparation problems; the recurring root constraint is incomplete work entering technical production.
Adoption holds when a single named owner controls one gate. Multi-owner rollouts stall inside three weeks.
A blocking readiness status is requested far more often than more automation. Firms want enforcement, not more suggestions.
Discovery converts better when the conversation starts at intake completeness rather than software replacement.
Prototype only. No real firm or client data. Nothing on this page is persisted.