The 3‑Tier Pricing Playbook for AI Services (Nomad‑Proof, Setup + Retainer)

Clients don’t buy tokens or prompts — they buy measurable outcomes. Pick 1–3 metrics you can move, then design your offer around guaranteeing progress on those metrics, not hours logged.

Choose your core outcome metrics (pick 2–3):

• Time‑to‑first‑touch (e.g., <5 minutes on new leads)
• Booked calls per week / per rep
• Lead qualification accuracy (%)
• First‑contact resolution (%) for support flows
• Cycle time reduction for a target task (minutes → seconds)

Define the minimum viable proof you’ll show in the first 30 days:

• Baseline: last 30 days before your system
• Target: what “good” looks like by Day 30 (range, not a single number)
• Evidence: logs, dashboard screenshots, and a short Loom showing the real workflow

Package the system, not just the agent:

• The agent with a documented persona and response templates
• A real‑time reporting dashboard with the 2–3 outcomes above
• A human handoff SOP with clear escalation triggers

Price bands below are intentionally narrow so clients know where they fit. Swap features for guarantees where possible.

Starter — $500–$750/mo + setup

• Channel scope: single channel (email or form)
• What’s included: 1 agent, templated persona, next‑business‑day support, shared Slack, monthly summary email
• Guardrails: usage cap; no custom integrations; no weekend incident response
• Ideal for: pilots and low‑volume teams

Growth — $1,200–$2,000/mo + setup

• Channel scope: multi‑channel (email + SMS/WhatsApp)
• What’s included: 1–2 agents, custom persona training, conversion dashboard, 4‑hour response SLA, dedicated Slack, monthly optimization pass
• Guardrails: capped monthly incidents; named integration list
• Ideal for: most SMB teams ready to bet on outcomes

Enterprise — $3,000+/mo + setup

• Scope: multiple agents, API access, SSO, sandbox + staged rollouts
• What’s included: 2‑hour response SLA, quarterly regression suite, white‑label options, residency/compliance accommodations, optional on‑prem/self‑host
• Guardrails: change‑control policy, success criteria with exec sign‑off
• Ideal for: regulated or complex environments

Setup fee (typical): $1,500–$5,000

• Covers discovery, data prep, integration wiring, eval harness, and runbook creation
• Make the setup deliverables tangible: “You leave with a working agent, a dashboard, and a human handoff SOP”

Keep your math simple and repeatable. Price the retainer so it funds ops reliability, not just model calls.

Inputs you must track monthly (per client):

• Human time: QA/evals (hrs), incident response (hrs)
• Vendor usage: model/API fees, observability, comms tools
• SLA/coverage: response windows you’re on the hook for

Quick math (sanity checks):

• Delivery cost (monthly) = (QA hrs + incident hrs) × your internal hourly cost + vendor usage
• Target price (monthly) = Delivery cost / (1 − target gross margin)
• SLA premiums to apply on top of base price:
• 4‑hour response window: +20%
• 2‑hour response window: +50%

Reality check on usage:

• Mid‑tier agents commonly consume millions of tokens/month; raw model spend is usually “tens of dollars,” not hundreds, when prompts and retries are controlled. Your retainer funds monitoring, evals, versioning, and incident coverage.

What to include in every retainer line‑item:

• Weekly QA/evals and prompt/version management
• Monitoring + alerting with named thresholds
• Incident process with response window and credit terms
• Monthly optimization summary with before/after metrics

Tip: Update your token/infra cost assumptions monthly and lock them in an appendix for that billing cycle.

Use adders to reflect real risk and effort. Keep a short menu so clients can predict spend.

Recommended multipliers (stack cautiously):

• Industry risk/compliance
• Healthcare, finance, gov contractors: ×1.5–2.2
• Adds BAAs/DPAs, audit logs, stricter access controls
• SLA speed
• 4‑hour: +20% • 2‑hour: +50%
• Legacy/ERP integration: +30–50%
• Multi‑language support: +20%
• Custom model training / RAG over private corp data: +50–100%
• Data residency/compliance hosting: pass through vendor uplift + your compliance premium
• White‑label delivery: +20–30%
• On‑prem/self‑host: custom SOW + change‑control; expect 3–5× total contract value over time via updates

Rules of thumb:

• Never hide pass‑through vendor uplifts; itemize and explain why they reduce risk
• Cap adders at +150% without a custom SOW to avoid “Franken‑scope”
• Tie every adder to an outcome or a risk removed

If it can’t run while you’re on a flaky connection or a flight, it fails. Bake these into every engagement.

Lisbon Test checklist:

• Async‑by‑default ops: no step requires a live meeting to resolve
• Monitoring: health checks + failure alerts before clients notice
• Caching/fallbacks: automatic model fallback and sensible retry limits
• Observability: per‑client dashboards with success/error rates and latency
• Change control: versioned prompts, test suite, and rollback plan
• Runbook: a human‑readable SOP a VA can follow during incidents
• Offline buffer: queueing so messages aren’t lost if a provider is down
• Access hygiene: scoped keys, rotated secrets, least‑privilege roles

De‑risk the decision and make the value felt in 30 days.

Pilot structure (30 days):

• Scope: 1 agent, 1–2 channels, capped volume
• Success metrics: pick 2 outcomes (e.g., <5‑min first touch; +5 booked calls/week)
• Deliverables: working agent, dashboard, handoff SOP, and cost‑transparency appendix (token logs, cache hit rates, incidents)
• Rollback clause: “If we miss the targets, you keep the artifacts and walk—no questions asked.”
• Exit paths: convert to Growth tier or extend pilot once with a defined change list

Pilot pricing patterns:

• Setup payable up‑front; pro‑rated first month retainer (creditable to Month 1 if converting)
• Discount only on setup, not on SLAs or incident coverage

Handle the “we’ll build in‑house” objection:

• Acknowledge: compute is cheap; reliability isn’t
• Show: your eval hours, incident log, version diffs that protected KPIs
• Offer: to train their team on the runbook as part of Enterprise tier

Turn a low‑MRR subscription into a system with guarantees in two days.

Day 0 (prep)

• Pull last 30 days of results; define two success metrics
• Draft a one‑page offer sheet with Starter/Growth/Enterprise grids and SLAs

Day 1 (productize)

• Add a branded dashboard (booked calls, first‑touch time, qual accuracy)
• Write a 1‑page human handoff SOP with escalation triggers and tagged owners
• Lock usage caps and response windows for each tier

Day 2 (pitch)

• Send the offer with before/after framing and pilot terms

Email snippet you can steal:

Subject: Reframing your lead bot as a system (so it scales)

Hey [CLIENT],

We’ve been running your follow‑up bot at [$199/mo]. It works, but it’s priced like a tool, not the system it’s become. I rebuilt the offer into Setup + Retainer so we can guarantee the two outcomes you actually care about: [<5‑min first touch] and [+5 booked calls/week].

Starter ($[500–750]/mo): single channel, next‑day support
Growth ($[1,200–2,000]/mo): SMS+email, 4‑hr SLA, dashboard
Enterprise ($3,000+/mo): multi‑agent, 2‑hr SLA, compliance options

Let’s run a 30‑day pilot with a rollback clause. If we miss the targets, you keep the artifacts and walk. If we hit, we roll into Growth.

Want me to send the pilot SOW?

– [YOU]

Don’t price yourself into burnout. Set capacity before you sell.

Quick capacity model:

• Ops time per client (monthly): 8–12 hrs (QA/evals + incidents) is common
• Personal ops budget (monthly): [YOUR AVAILABLE HOURS]
• Max active clients = floor(available hours ÷ hours per client)

Examples (solo operator):

• If you target 70% gross margin and charge $1,500/mo, expect to handle ~8–10 clients before operations eat your week
• At $750/mo Starter with the same ops load, you’ll need 16–20 clients for the same revenue — painful on the road

Set your 10K MRR path:

• Option A: 6× Growth at $1,700 = ~$10.2k MRR (sane)
• Option B: 3× Enterprise at $3,500 = ~$10.5k MRR (fewer accounts, higher expectations)
• Option C: 2× Enterprise + 4× Starter = ~$9–11k MRR (watch SLA load)

Rule: raise before you overfill. If you’re at 80% capacity for 2 consecutive months, lift base prices by 10–15% on new deals and add an implementation queue.