# The Repo: The AI Operating System SteinLaw has been running without.

**Ad-Apt · Proposal · v1.0 · End-to-end engagement · Confidential**

| Prepared for | Issued | Validity | Strategists |
| --- | --- | --- | --- |
| SteinLaw | May 2026 | 30 days | Chase · Lucas · Bryan |

We build you *The Repo*, the substrate that ties Actionstep, your supervisors, and your case-banking pipeline into one detection-and-recommendation layer. **Hybrid infrastructure**: cloud for the workload, a local Mac mini for privileged data.

One staff-facing chat surface (**Carcin**) on top. Underneath: *three Phase 1 agents* running daily across enforcement, inactivity detection, and settled-case banking. A human strategist setting the play, supervisors keeping authority. **The Repo is yours from the moment of creation** — source code, prompts, skills, memory, configurations. *Work-for-hire.*

> **Scope note.** This proposal covers Foundation + Phase 1 workflows only. Two adjacent sprints from earlier scoping — the *Case Vision migration sprint* and the *LawScore revival sprint* — have been **tabled at SteinLaw's request**. Both can be re-scoped in a future engagement once Phase 1 is in production. The architecture and Repo built here are designed to absorb them later without rework.

| | |
| --- | --- |
| **1** | The Repo · SteinLaw-owned |
| **3** | Phase 1 workflow agents |
| **3** | Strategists on call |
| **6 mo** | Sprint engagement |

---

## 1. The Vocabulary

*Five terms. One "SteinLaw AI" language.*

Before we go deeper, here's how we use a few words. The proposal makes more sense once we share the vocabulary. **The Repo** contains everything else; the rest are the components inside it.

- **05 · The Repo.** The durable, accretive infrastructure where every Agent, Workflow, Model, and Connection lives. **Everything inside The Repo is yours from the moment of creation.**
- **01 · Agent.** AI with skills, soul, and memory — spawned, tasked, and held accountable for outcomes.
- **02 · Workflow.** A client-specific task or process an Agent completes — repeatable, measured, and ownable.
- **03 · Model.** A tuned AI from companies like Anthropic and OpenAI, or open source — what an Agent calls when it needs to think.
- **04 · Connection.** An authenticated read/write path to a particular system or piece of software.

---

## 2. The Problem

*Missed tasks, stalled cases, and slow settled-case banking are symptoms of a greater "Actionstep alone isn't watching" problem.*

Actionstep is the system of record. It is not a system of *detection*. Tasks go overdue without anyone noticing. Cases go inactive for weeks before a supervisor catches it. Settlements sit in CLS for 30+ days because nobody's tracking the milestone clock. The original SteinLaw Workflow Build Spec Pack identified these as ten distinct workflows. We start with the three that hurt most today — the Phase 1 enforcement layer — and we operationalize the fix continuously rather than discovering it in a monthly review.

### The signals are already in Actionstep.

Every overdue task, every stalled case, every settled-but-unbanked matter is sitting in Actionstep right now, waiting to be detected. What's missing is a layer that scans for them three times a day, routes the alert to the right supervisor, escalates if uncorrected, and produces a digest at end of day. The data exists. **The accountability ladder doesn't.** Phase 1 builds that ladder for the three workflows where the cost of a missed signal is highest — missed deadlines, stalled files, and delayed banking.

> "A daily review meeting catches yesterday's problem. The Repo catches this morning's, before it becomes Friday's escalation."

### The framework is SteinLaw-specific.

This isn't a generic legal-AI playbook. The decision logic is built against SteinLaw's actual operating model — the supervisor hierarchy, the no-noise alert discipline, the Florida Bar UPL boundary, the IOLTA trust-account rules, the existing role of the Unit Operations Head. AI never crosses into legal advice or client-facing communication. AI never moves money. **Detection and recommendation only.** Humans keep enforcement authority. The Phase 1 build hardens that boundary into the architecture from day one.

> "AI as a detection layer, not a decision layer. Brandon's words. The Repo enforces the line in code, not in policy."

---

## 3. The Frame

*Hub & Spoke. Decades old. New target.*

For SteinLaw, the hub is **Actionstep**: the system of record where every matter, task, note, and case event lives today. Phase 1 doesn't replace it. Phase 1 reads from it on a schedule, mirrors what we need into a **local Postgres store on a Mac mini inside the firm** (the privileged-data residency layer), and runs detection from the **cloud orchestrator**, then writes alerts and digests back through the proper accountability ladder. Hybrid by design: *sensitive data stays in the building, elastic compute runs in the cloud.*

**Hub:** Actionstep + Mirror (source of truth).

**Six spokes radiating out:**

| # | Domain | Carries |
| --- | --- | --- |
| 01 | Case Ops | tasks · matters |
| 02 | Supervisors | enforcement |
| 03 | Banking | CLS · IOLTA |
| 04 | Compliance | FL Bar · privilege |
| 05 | Reporting | digests · alerts |
| 06 | Audit Trail | eval · lineage |

*Phase 1 lights up Case Ops, Supervisors, and Banking first. Same pattern, every time.*

---

## 4. The Pillars

*Six SteinLaw domains. Phase 1 lights up three.*

These are the six domains the firm's AI ecosystem will eventually span, mapped from the original Workflow Build Spec Pack. Phase 1 ships agents inside three of them: **Case Operations, Supervisor Enforcement, and Banking**. The other three (Compliance scaffolding, Reporting digests, Audit Trail) get the substrate built in foundation but the workflow agents inside them are deferred to later phases. *The framework absorbs the rest later without rework.*

### 01 · Case Ops — *Tasks, caught.* [PHASE 1]

- **Hub:** **Actionstep** — the system of record for every matter, task, due date, and case event. Phase 1 reads from it three times a day.
- **Spokes:** Workflow 1 (Missed Task Enforcement) · Workflow 2 (Case Inactivity Detection) · Postgres mirror · Inactivity scoring · Severity tiers (Yellow/Orange/Red) · No-noise digest discipline.
- **Phase 1 deliverables:** Daily overdue-task sweep (3x/day) routing to direct supervisor · correction-cycle timer · daily inactivity queue tagged Yellow/Orange/Red · Red goes same-day to Unit Case Manager and Unit Leader. *Both ship with labeled eval set + acceptance gate.*

### 02 · Supervisors — *Accountability, laddered.* [PHASE 1]

- **Hub:** **The escalation ladder** — every alert routes to a named supervisor with a documented correction window, then transfers ownership upward if unresolved.
- **Spokes:** Direct supervisor routing · Unit Leader oversight · Unit Operations Head · Managing Partner (executive risk only) · Same-day directive logs · First-cycle resolution tracking.
- **Phase 1 deliverables:** Wired escalation ladder per workflow · one-active-issue-per-matter alert discipline · end-of-day unresolved digest · correction-cycle timer with auto-escalation. *No real-time alert noise except critical failures.*

### 03 · Banking — *Settlement to bank, tracked.* [PHASE 1]

- **Hub:** **The CLS pipeline** — settlement event → preliminary CLS → release request → release receipt → trust deposit → final CLS. Today, milestones drift. Phase 1 makes them visible.
- **Spokes:** Workflow 3 (Settled Case Banking) · 24h notice tracking · 5-day release windows · 20-day draft receipt clock · Day-30 banking breach flag · Day-90 finalization breach · IOLTA discipline (track, never move).
- **Phase 1 deliverables:** Daily settled-case pipeline digest · milestone countdown status · late-case escalation queue · Managing Partner sees day-20+, day-30, day-90 only. *AI tracks; humans execute every trust-account action.*

### 04 · Compliance — *Privilege, preserved.* [SUBSTRATE]

- **Hub:** **Hybrid architecture + compliance scaffolding.** Privileged content lives on a **local Mac mini inside the firm**, never in the cloud. Non-sensitive workloads (orchestrator, AI calls with redacted prompts, dashboards) run in the cloud. BAA-tier model contracts, redaction layer, UPL boundary enforcement, conflict screening. Built in foundation; protects every Phase 1 workflow.
- **Spokes:** BAA-tier Anthropic + OpenAI · Redaction layer (PII / PHI / privileged) · UPL guardrails · FL Bar review gate · IOLTA discipline (Workflow 3) · Conflict-of-interest screen · Data classification taxonomy · Retention + deletion policy.
- **Built in foundation:** Compliance scaffolding ships before any workflow does. Every Phase 1 agent inherits the redaction layer, BAA-tier model selection, and UPL boundary by default. *No workflow goes to pilot without compliance counsel sign-off on its prompt library.*

### 05 · Reporting — *One number. One narrative.* [PHASE 1]

- **Hub:** **Carcin's daily digest** — end-of-day rollup of overdue tasks, inactivity-flag changes, settled-case pipeline status. Supervisors open it in the morning, leadership scans it at end of week.
- **Spokes:** Daily unresolved digest · Settled-case pipeline view · Inactivity heat map · Severity-tagged queue (Y/O/R) · Same-day-action flag rate · Power BI / Metabase.
- **Phase 1 deliverables:** Carcin daily digest · per-role views (paralegal / supervisor / Unit Leader / Ops Head / Managing Partner) · weekly scorecard for Phase 1 metrics. *Phase 3 supervisor scorecard and Phase 4 War Room dashboard are deferred.*

### 06 · Audit Trail — *Lineage, defensible.* [SUBSTRATE]

- **Hub:** **Append-only audit log** — every AI inference, every alert, every supervisor action recorded with model version, prompt version, and outcome. Daily Merkle root mirrored to GitHub for tamper evidence.
- **Spokes:** Per-decision lineage · Model + prompt versioning · Eval set per workflow · Acceptance-gate logs · Cost instrumentation · Token meter per workflow · Quarterly re-labeling pass.
- **Built in foundation:** Audit harness ships in foundation. Every Phase 1 workflow logs through it from day one. Eval framework, regression gates, cost instrumentation — all live before W1 ships. *Required for legal defensibility and for our own debugging.*

---

## 5. The Engine

*Priority is computed, not opined.*

The original Workflow Build Spec Pack ranked all ten workflows by complexity, time-to-value, and dependency. Phase 1 selects the three with the highest weighted score given today's pain at SteinLaw — missed deadlines, stalled files, and slow banking. The same engine produces the ranking when Phase 2 (Workflows 4-6) and Phase 3 (Workflows 7-9) come up for re-scoping. *Output is a deterministic ranked list. We work top-down.*

### Candidate priority — weighted score

```
priority(candidate) =
  0.30 · impactOnNumber
+ 0.25 · timeToValue
+ 0.15 · stageFit
+ 0.10 · strategyFit
+ 0.10 · verticalFit
+ 0.10 · compoundingValue

// then adjusted for reality
× budgetFit(cost)
× capacityFit(complexity)
× timelineFit(ttv)
```

### Inputs into the engine

| Input | Source / role |
| --- | --- |
| Domain | Every system you run — common ones plus whatever's specific to you. |
| Business model | How you actually make money — shapes which work compounds. |
| Vertical | Learned during intake — every config tailored to your industry. |
| Stage | Where you are in the lifecycle — drives priority calibration. |
| Budget posture | What you can spend — drives intensity per system. |
| Team capacity | Your team's bandwidth — drives owned-vs-delegated split. |
| Timeline | Your urgency — drives sequencing. |
| Existing stack | Audited per system — we meet you where you are. |

---

## 6. The Deliverable

*The Repo & the Phase 1 roadmap that runs inside it.*

Before any workflow agent is deployed, SteinLaw receives The Repo, the substrate (hybrid cloud + local), the Actionstep mirror, the compliance scaffolding, the audit trail, the eval harness, and the Carcin daily digest. The **6-month sprint** covers building all of that plus shipping the three Phase 1 workflows, then a short operate-and-tune window. *Workflows 4-10 from the original Build Spec Pack, plus the Case Vision migration and LawScore revival sprints, are tabled for a future engagement.*

### Here's the 6-month plan for SteinLaw.

The original Spec Pack defined ten workflows across four phases. Phase 1 ships the three enforcement-layer workflows: Missed Task Enforcement, Case Inactivity Detection, Settled Case Banking. Foundation gets built first (M1–M2); workflows roll out M2–M5; M5–M6 is operate, tune, and Phase 2 scoping decision.

| | |
| --- | --- |
| **3** | Phase 1 workflows |
| **186** | Build hours — foundation + workflows |
| **6** | Month sprint |
| **1** | Source of truth (Actionstep) |

### Timeline

| Month | Focus |
| --- | --- |
| M1 | Foundation |
| M2 | Carcin + W1 build |
| M3 | W1 ships + W2 (PEAK) |
| M4 | W2 ships + W3 |
| M5 | W3 ships + UAT |
| M6 | Operate + handoff |

### Plan steps

| # | Item | Window | Hours | Status |
| --- | --- | --- | --- | --- |
| 01 | Foundation — **hybrid infrastructure** (cloud VPS + local Mac mini for privileged data), OpenClaw orchestrator, Postgres mirror, Actionstep API integration · *substrate* | M1 | 60 | Ready |
| 02 | Compliance scaffolding, BAA-tier model contracts, redaction layer, UPL guardrails, retention + deletion policy · *substrate* | M1–M2 | 18 | Ready |
| 03 | Audit trail + eval harness + cost instrumentation, append-only log on local box, GitHub-mirrored Merkle root, regression gates · *substrate* | M1–M2 | 22 | Ready |
| 04 | Carcin daily digest live, end-of-day rollup of all Phase 1 signals to role-scoped views · *reporting* | M2 | 8 | Ready |
| 05 | **Workflow 1, Missed Task Enforcement** — 3x daily overdue sweep, supervisor routing, escalation ladder · *case ops* | M2–M3 | 23 | Ready |
| 06 | **Workflow 2, Case Inactivity Detection** — daily inactivity scoring, Yellow/Orange/Red severity tiers, Red → same-day to Unit Case Manager · *case ops* | M3–M4 | 29 | Ready |
| 07 | **Workflow 3, Settled Case Banking** — settlement-to-banking pipeline tracking with day-30 / day-90 breach flags · *banking* | M4 | 26 | Ready |
| 08 | UAT + pilot rollout per workflow, acceptance gates, supervisor onboarding, training sessions · *operate* | M5 | 24 | Ready |
| 09 | Operate & tune, eval drift monitoring, prompt iteration, supervisor friction loop, handoff · *operate* | M6 | 25 | Operate |
| 10 | Phase 2 scoping (Workflows 4–6), re-engagement decision point, Case Vision sprint re-scope if approved · *planning* | M6 | included | Optional |

> **Tabled at client request — not in this engagement:** Workflows 4–10 (Back-End Review Generator, Deep Dive Review, Litigation Scheduling Enforcement, Case Valuation Review, Supervisor Enforcement Scorecard, Gross Profit + Root Cause, Executive War Room Dashboard) · *Case Vision migration sprint* · *LawScore revival sprint*. All can be re-scoped in a future engagement; the foundation built here absorbs them without rework.

---

## 7. The Operator

*We don't just recommend the work. We build The Repo and run it.*

The Repo is the substrate, the owned codebase, the Actionstep API integration, the Postgres mirror, the compliance scaffolding, the audit trail, the eval harness. On top of it sits **Carcin**, the staff-facing chat surface for the daily digest and ad-hoc queries. Underneath: **three Phase 1 agents** running autonomously across the firm's task, matter, and banking pipelines, with a human strategist setting the play and supervisors keeping enforcement authority. Cancel or scale back any agent at any time.

```
ad-apt-agents · SteinLaw Phase 1 delivery loop · ● running

# The Repo dispatches the work, humans & agents execute against Actionstep
$ ad-apt run --client "SteinLaw" --scope "phase-1"

# Carcin, staff-facing chat surface, sits on top of The Repo
Carcin                  → daily digest · role-scoped queries · production

# Phase 1 enforcement layer · 3 workflow agents
W1 · Missed Tasks       → Actionstep · 3x daily sweep · supervisor routing · escalation ladder
W2 · Inactivity         → daily inactivity scoring · Yellow/Orange/Red tiers · Red same-day
W3 · Settled Banking    → CLS pipeline tracking · day-30 / day-90 breach flags · IOLTA-safe (track only)

# Substrate agents that protect every Phase 1 workflow
Compliance Layer        → BAA-tier model selection · redaction · UPL guardrail · conflict screen
Audit Trail             → per-decision lineage · daily Merkle root mirrored to GitHub
Eval Runner             → labeled set per workflow · regression gate · drift detection

# Tabled at client request, not in this engagement
Case Vision Bridge      → DEFERRED
LawScore Healer         → DEFERRED

# Human gate, supervisors and strategists approve before anything ships
Strategist              → APPROVED
[ DEPLOY ]              → LIVE
```

### Behind the agents — Three strategists. One mission.

Carcin and the agent stack do the production. Three humans set the play, approve before anything ships, and own the outcome.

| Name | Role | Responsibilities |
| --- | --- | --- |
| Chase Bradley West | CTO | Architecture · sys admin · security · hardware · schema · libraries · uptime |
| Lucas Sommer | CEO / CFO | Strategy · creative · business model · leadership · use cases |
| Bryan Lozano | CMO + BI | Dashboards · notifications · marketing efficiency · conversion & event tracking |

---

## 8. Phase 1 Workflows in Detail

*Three workflows. Every detail.*

These three workflows come directly from SteinLaw's original Workflow Build Specification Pack. We are not inventing them. We are *building exactly what was scoped* — with every required field, decision rule, alert discipline, and escalation path preserved. Below is each workflow expanded against the spec, so the firm sees precisely what gets shipped.

### W1 — Missed Task Enforcement

*Same-day routing through the accountability ladder.*

**Phase 1 · Owner: Direct supervisor · Oversight: Unit Leader / Unit Operations Head**

- **Purpose.** Detect overdue tasks and route same-day enforcement through the correct accountability ladder.
- **Trigger.** Any Actionstep task with a due date in the past and no completed or cleared status. Scanned **3 times per business day.**
- **Outputs.**
  - Create enforcement alert record
  - Create same-day action request for supervisor
  - Start correction-cycle timer
  - If unresolved, place in escalation queue
- **Alert rule.** One active issue alert per matter/task. End-of-day unresolved digest. **No duplicate alerts** unless the correction deadline passes or severity changes.
- **Escalation.** After one documented correction cycle, transfer enforcement ownership upward if the issue remains unresolved. Escalation transfers ownership, it is not a reminder mechanism.
- **Decision logic.**
  1. Confirm task has a due date
  2. Confirm due date has passed
  3. Confirm task is not completed or cleared
  4. Confirm the issue is not already in an active enforcement cycle
  5. Route to the proper supervisor based on operations or litigation ladder
- **Required fields.** matter ID · matter name · task ID · task name · task category · due date · assigned attorney · delegated paralegal (if stored) · current task status · last status change date · legal unit · team · direct supervisor · escalation path.

### W2 — Case Inactivity Detection

*Non-moving files surfaced before they become value loss.*

**Phase 1 · Detection: Unit Case Manager · Enforcement: Unit Leader**

- **Purpose.** Detect non-moving files before they become value-loss or movement failures.
- **Trigger.** Daily scan of all active matters in the pilot unit.
- **Outputs.**
  - Daily inactivity queue for Unit Case Manager and Unit Leader
  - Severity-tagged movement risk status (Yellow / Orange / Red)
  - Red-status same-day review item
- **Alert rule.** **Yellow** is dashboard only. **Orange** goes into the daily digest. **Red** receives a same-day notice to Unit Case Manager and Unit Leader only. *Never route system instructions directly to legal teams.*
- **Escalation.** If the Unit Leader does not act within the review window, escalate to the Unit Operations Head.
- **Decision logic.**
  1. Exclude intentionally paused, discharged, and post-settlement files
  2. Calculate inactivity score based on days since activity, overdue tasks, and posture-specific movement expectations
  3. Classify file as Yellow, Orange, or Red
- **Required fields.** matter ID · matter name · case posture · legal unit · assigned attorney · assigned paralegal · last user note date · last substantive document date · last completed task date · open task count · case track date (if stored) · inactivity exception flag · Unit Case Manager · Unit Leader.

### W3 — Settled Case Banking

*Settlement-to-banking-to-finalization timeline tracked end-to-end.*

**Phase 1 · Executes: Assigned attorney · Enforces: Unit Leader · Risk view: Managing Partner**

- **Purpose.** Track the complete settlement-to-banking-to-finalization timeline.
- **Trigger.** Matter status changes to settled, or a settlement event is logged.
- **Outputs.**
  - Daily settled-case pipeline digest
  - Milestone countdown status
  - Late-case escalation queue
- **Alert rule.** Daily digest only. **Managing Partner** receives late-stage risk only: **day-20+, day-30, and day-90** matters.
- **Escalation.** Missed milestone routes first to Unit Leader / Department Head, then to Managing Partner when delay becomes material or persistent.
- **Decision logic.**
  1. Track notice within 24 hours
  2. Track preliminary CLS within 24 hours
  3. Track proposed release within 5 days
  4. Track client-executed release return within 5 days of send
  5. Track draft receipt within 20 days of executed release
  6. Flag day-30 banking breach and day-90 finalization breach
- **Required fields.** matter ID · matter name · assigned attorney · legal unit · settlement date · settlement amount · SteinLaw fee amount · Operations Manager notice date · preliminary CLS dates · release request date · release receipt date · executed release date · draft receipt date · trust deposit date · final CLS draft date · final balances confirmed date · final CLS signed date · delay/exception reason.

> **IOLTA discipline:** Workflow 3 *tracks but never moves money*. Trust-account actions stay 100% human-executed. AI surfaces the milestone clock; humans make every transfer.

### Applied to all three workflows — SteinLaw's no-noise discipline, in code.

- **No alert without action.** Dashboard-only conditions stay dashboard-only.
- **One issue, one active alert.** Duplicate reminders are prohibited unless a deadline passes or severity changes.
- **Severity tiers:** Critical for same-day action, Warning for digest-based action, Informational for dashboards.
- **Most workflows route through daily digests**, not real-time alerts.
- **Escalation transfers ownership upward.** It is not a reminder mechanism.
- **Role boundaries are mandatory.** AI never directs legal teams. AI never changes operating rules.

---

## 9. Investment

*One retainer. Six months of build, ship, operate & tune.*

One ongoing retainer covers everything for Phase 1 — building The Repo, shipping the three Phase 1 workflow agents, running Carcin, iterating on prompts, monitoring evals, supervisor onboarding, monthly review. **Six-month sprint cadence.** The retainer ramps up M1–M3, peaks at M3–M4 during build + UAT, then declines through M5–M6 as the system stabilizes. Cancel or scale back any workflow at any time — hours saved, retainer reduced, AI/API spend lowered.

### Inside the retainer — what the retainer actually covers.

SteinLaw owns The Repo from the moment of creation. Source code, prompts, skills, memory, configurations — work-made-for-hire. Hardware (if local path is chosen) is SteinLaw's. Cloud accounts (Hetzner, AWS, Anthropic, OpenAI) registered to and paid by SteinLaw directly. The retainer covers all of the work below, end to end: building, tuning, running, and iterating.

1. **Foundation + Carcin (hybrid infra).** Hybrid infrastructure: cloud VPS for orchestrator, AI calls, dashboards · **local Mac mini inside the firm** for privileged Postgres mirror, audit log, scraper archives (one-time ~$2K hardware purchase in M3, *optional, as needed*) · Actionstep API integration · **compliance scaffolding** (BAA-tier models, redaction, UPL guardrails, FL Bar review gate) · **audit trail** · **eval harness** · cost instrumentation · plus **Carcin**, the staff-facing daily digest.
2. **Phase 1 workflow agents (3).** **Missed Task Enforcement**, **Case Inactivity Detection**, **Settled Case Banking**. Each ships with success metric + baseline, labeled eval set, privilege check, and explicit pilot-to-production acceptance gate. *Workflows 4–10, the Case Vision sprint, and the LawScore sprint are tabled at client request.*
3. **Strategists + forward-deployed engineers.** Three strategists on call (Chase, Lucas, Bryan) · forward-deployed engineers embedded in SteinLaw's stack · weekly working sessions, on-demand Slack/phone access, monthly performance review · the humans who set the play and approve before anything ships.

### Cost components

| Stream | Average / mo | Range / mo | Notes |
| --- | --- | --- | --- |
| Retainer | **$3,333** | $2,000 – $4,500 | Ramps M1–M4 as agents come online, peaks while tuned, declines as system matures. |
| Infrastructure | **$225** | $200 – $250 | Cloud (VPS or local), hosting, observability · scales as the agent stack grows, then plateaus. |
| AI / API | **$250** | $200 – $300 | LLM inference, third-party API fees · scales with agent activity. |

### What's included, every month of operate

- The Repo — owned codebase, agents, memory, audit log (SteinLaw's property)
- Carcin daily digest — end-of-day rollup of all Phase 1 signals to role-scoped views
- 3 Phase 1 workflow agents running daily against Actionstep
- Compliance scaffolding (BAA-tier models, redaction, UPL guardrail) on every prompt
- Eval harness with labeled sets and regression gates
- Audit trail with per-decision lineage, defensible for FL Bar / counsel review
- Quarterly leadership review with Brandon · weekly working sessions
- Cancel or scale back any workflow with 30 days' notice — hours saved come off the retainer

### Three streams, one bill

Retainer *ramps* in M1 ($3.5K) as foundation kicks off, *plateaus* at peak through M2–M3 ($4.5K) while the three Phase 1 workflows are built, then *declines* M4 ($3K) → M5 ($2.5K) → M6 ($2K) as the system stabilizes into operate-and-tune mode. **Optional ~$2K hardware in M3** covers a local Mac mini for privileged-data residency — recommended as part of the hybrid architecture but a one-time line item the firm can defer or replace with full cloud. Infrastructure scales modestly; AI/API floats with workflow activity. *Value of the work compounds past M6 as evals tighten and supervisors trust the system.*

### 6-month sprint, month by month

| Month | Focus & deployments | Retainer | Infra | AI/Tokens | Total |
| :--- | :--- | ---: | ---: | ---: | ---: |
| **M1** | *Foundation kickoff · hybrid infra stand-up* — Discovery + Actionstep field audit · Cloud VPS provisioning · Compliance scaffolding start. | $3,500 | $200 | $200 | **$3,900** |
| **M2** | *Foundation finish + Carcin live + W1 build* — Carcin daily digest · Audit trail + eval harness · W1 Missed Task Enforcement build. | $4,500 | $250 | $250 | **$5,000** |
| **M3** | *W1 ships + W2 build · optional local hardware* — W1 Missed Task Enforcement · 3x daily sweep + escalation ladder · W2 Inactivity scoring build · + Mac mini (optional, as needed). | $4,500 | $250 (+$2,000 opt) | $300 | **$5,050** ($7,050 w/ hw) |
| **M4** | *W2 ships + W3 build* — W2 Case Inactivity Detection · Yellow/Orange/Red severity tiers · W3 Settled Banking build. | $3,000 | $250 | $300 | **$3,550** |
| **M5** | *W3 ships + UAT + supervisor onboarding* — W3 Settled Case Banking · Per-workflow acceptance gates · Supervisor onboarding + training. | $2,500 | $200 | $250 | **$2,950** |
| **M6** | *Operate + tune + Phase 2 scoping* — Eval drift monitoring · Prompt tuning + friction review · Phase 2 / Case Vision / LawScore re-scope. | $2,000 | $200 | $250 | **$2,450** |
| **Total** | **Foundation + 3 Phase 1 workflow agents**, deployed against Actionstep with hybrid (cloud + local) infrastructure. *Case Vision and LawScore sprints tabled at client request.* | **$20,000** | **$1,350** (+$2,000 opt) | **$1,550** | **~$22,900** (~$24,900 w/ hw) |

---

## 10. The Flexibility

*Built to bend. Not to lock.*

The whole point of The Repo is that it's SteinLaw's. The whole point of the retainer is that it adjusts. Three guarantees keep both true: at any point in the engagement, the firm can pause work, scale it back, or walk away with everything we've built. Re-scoping Case Vision or LawScore in a future engagement uses the same flexibility.

### 01 · Pause or cancel any workflow, monthly.

Each Phase 1 workflow runs on its own line item. Want to pause W2 (Inactivity) for a quarter while you focus on W1 (Enforcement)? Done. Want to dial back operate hours after M6? Done. Changes apply at the start of the next billing month — the retainer adjusts down with the scope.

### 02 · Cancel the whole sprint with 30 days' notice.

No multi-year contract. No claw-back clauses. 6-month sprint commitment, but if the work stops earning its keep mid-sprint, give us 30 days and we'll wrap up cleanly — final retainer covers the wind-down period including a handoff document, a final dashboard snapshot, and confirmation that all credentials, the hardware (if purchased), and The Repo have transferred.

### 03 · All code & assets are yours, always.

Source code, prompts, skills library, memory, configurations, training data, derived data, dashboards, runbooks — everything inside The Repo is SteinLaw's property from the moment of creation. **Work-for-hire.** Hardware (if local path) is SteinLaw's. Cloud accounts (Hetzner, AWS, Anthropic, OpenAI) registered to and paid by SteinLaw directly, never proxied. **The latest version of The Repo is yours to keep** — push it to your own GitHub, fork it, give it to another vendor, archive it, do whatever you want. No lock-in clause, no proprietary runtime, no "but our software won't work without our subscription." It's just code, on your infrastructure, doing your work.

---

© 2026 Ad-Apt