Build scope, pricing and inclusions
Schedule ID: PS-BUILD-2026-11-15-v1 · Version 1 · 15 November 2026
This schedule applies only when incorporated into an accepted order.
Fixed build quotes are valid for 30 days; the order records the validity start and expiry. The order records counts, tier, adders, price, milestone events, acceptance and go-live. Typical durations are estimates; the fixed quote states the actual timeline. The support window starts at go-live. Calendar, business-day holidays and time zone must be recorded in the order.
The tiers at a glance
| S: Single flow | M: Connected system | L: Operations platform | |
|---|---|---|---|
| Price | $3,500 | $8,000 | from $18,000 (base + scale-ups) |
| Typical time | 2–3 weeks | 4–6 weeks | 7–12 weeks |
| Example | Invoice email → extracted → accounting app | Support triage agent + CRM + helpdesk + review queue | AP intake, approval, ERP sync and reporting across teams |
| Milestones | 50% start · 50% go-live | 40 / 40 / 20 | 30 / 30 / 30 / 10 |
| Support after go-live | 14 days | 30 days | 30 days |
1. Placement criteria (bands don't overlap)
| Criterion | S | M | L |
|---|---|---|---|
| Workflows (one trigger → one outcome) | 1 | 2–3 | 4+ |
| System units (app, API, DB, inbox = 1; a system without a usable API = 2) | 1–2 | 3–4 | 5+ |
| AI steps (LLM calls that extract, classify, decide or draft) | 0–1 (0 = deterministic build, no AI); if 1, output checked before it's written | 2–3, or 1 agent that uses tools | 4+, multiple agents, or an agent that writes to core systems without approval |
| Edge cases handled (each exception type with its own handling; each variant by client, region or document type counts separately) | 0–3 (others go to a human) | 4–10 | 11+ |
Zero-AI builds (plain integrations and rule-based automations) count 0 AI steps, which sits in the S band, so the other three criteria place them. They get the same price as an AI build of the same size, because the time goes into systems and edge cases instead.
"Top of M" means a criterion sits at M's maximum: 3 workflows, 4 system units, 3 AI steps (or 1 tool-using agent) or 10 edge cases.
2. How a quote is built (always in this order)
- Count the 4 criteria. A system without an API is counted as 2 units here, and only here.
- Place: the tier is the highest band any one criterion reaches.
- Bump: if the result is M and 2 or more criteria are at the top of M, the project becomes L. The client can choose phases instead, and each phase is then placed again from step 1.
- Price: S and M are flat. L is $18,000 plus scale-ups (below).
- Adders: add only the items in section 3. They never change the tier, and nothing that step 1 counts is charged here.
L base covers up to 5 workflows, 6 system units, 5 AI steps and 20 edge cases. L scale-ups: +$3,000 per extra workflow, +$1,500 per extra system unit, +$1,500 per extra AI step, +$1,000 per started block of 5 extra edge cases (round up: 1–5 extra = $1,000, 6–10 extra = $2,000; no prorating).
3. Adders (things the criteria do not count; quoted)
| Item | Adds |
|---|---|
| Compliance work (SOC 2 evidence, DPA review, data residency) | quoted per item |
| Data cleanup or migration over 1 day | $800 per day |
| Rush (timeline shorter than the tier's typical minimum) | +25% of the build price |
Not included: AI and infra running costs (the client pays these on their own accounts, and the audit estimates them), and licences.
4. Worked examples (each one gets one tier and one price)
| # | Project | Counts (workflows · system units · AI steps · edge cases) | Steps 2–3 | Tier and price |
|---|---|---|---|---|
| 1 | Two workflows: invoice intake + payment reminder, Gmail + Xero | 2 · 2 · 1 · 3 | Workflows → M, nothing at top of M | M, $8,000 |
| 2 | Three systems: form → CRM + Slack alert | 1 · 3 · 1 · 2 | Systems → M | M, $8,000 |
| 3 | Ten exceptions: one doc-extraction flow, 2 systems | 1 · 2 · 1 · 10 | Edge cases → M (top of M), only 1 criterion at top | M, $8,000 |
| 4 | No API: inbox → desktop accounting app with no API | 1 · 1 + 2 = 3 · 1 · 3 | Systems → M; no surcharge, because the double count already prices it | M, $8,000 |
| 5 | No API, bigger: 2 workflows, 3 API systems + 1 without an API | 2 · 3 + 2 = 5 · 2 · 6 | Systems → L | L, $18,000 |
| 6 | Mixed M/L: 3 workflows, 5 systems, 2 AI steps | 3 · 5 · 2 · 8 | Systems → L, which outranks M | L, $18,000 |
| 7 | Two at top of M: 3 workflows, 4 systems | 3 · 4 · 1 · 5 | M, then bumped (2 at top) | L, $18,000 (or phases, each placed again) |
| 8 | L beyond base: 7 workflows, 8 system units | 7 · 8 · 4 · 15 | L | $18,000 + 2×$3,000 + 2×$1,500 = $27,000 |
| 9 | S with rush: 1 workflow, 2 systems, 1 AI step, delivered in 1 week | 1 · 2 · 1 · 2 | S | $3,500 + 25% = $4,375 |
| 10 | Zero-AI, small: Shopify orders → Google Sheet sync, rule-based | 1 · 2 · 0 · 0 | All criteria in S | S, $3,500 |
| 11 | Zero-AI, connected: CRM ↔ accounting ↔ Slack, 2 syncs, rule-based | 2 · 3 · 0 · 0 | Workflows and systems → M, nothing at top of M | M, $8,000 |
| 12 | L at edge-case base: otherwise-base L (5 · 6 · 5) | 5 · 6 · 5 · 20 | L, 0 extra | $18,000 |
| 13 | Same, 21 edge cases | 5 · 6 · 5 · 21 | L, 1 extra → 1 started block | $18,000 + $1,000 = $19,000 |
| 14 | Same, 25 edge cases | 5 · 6 · 5 · 25 | L, 5 extra → 1 block | $18,000 + $1,000 = $19,000 |
| 15 | Same, 26 edge cases | 5 · 6 · 5 · 26 | L, 6 extra → 2 started blocks | $18,000 + $2,000 = $20,000 |
5. What every tier includes
| Deliverable | S | M | L |
|---|---|---|---|
| Production code in the client's accounts (not a no-code template) | ✓ | ✓ | ✓ |
| Error alerts (email or Slack) and a run log | ✓ | ✓ | ✓ |
| Human review step for anything outside the handled edge cases | ✓ | ✓ | ✓ |
| Test set for each AI step, only if the build has AI steps (S: 20 real examples, run before go-live; M/L: 50+, re-runnable). Zero-AI builds get a test run on 20 real records instead | ✓ | ✓ | ✓ |
| Review queue for flagged items | – | ✓ | ✓ |
| Monitoring dashboard (volume, failures; plus AI cost per run if the build has AI steps) | – | ✓ | ✓ |
| Short docs and a handover call | ✓ | ✓ | ✓ |