Anthropic

Claude's consumption-based token pricing means an agency that quotes a fixed retainer without usage caps can lose money on every heavy client, while an agency that meters usage and resells Claude Code and Cowork access at the $20 Pro or $200 Max tier builds a defensible margin layer. The real strategic stake is not model quality but who owns the client's prompt library and workflow configuration, because that asset is what makes the retainer sticky.

Claude bills by consumption, not seats: Free at $0, Pro at $20 monthly ($17 annual), and the Max plan at $200/mo for 200+ requests per minute. That structure means an agency's gross margin on any Claude-powered retainer is set by token burn, not by headcount. Picture a content agency selling the Claude Starter Content Kit at $1,800 for 16 hours of setup: the build fee is fixed, but every client chat widget reply and every regenerated social post draws on the same API meter. Before quoting a monthly retainer, log a week of real client usage, multiply by the model tier in play, and set the retainer floor above that number. Agencies that skip this step discover the leak in month three, when a chatty client's FAQ widget has quietly eaten the setup fee. Cap usage per client in the contract, then reprice annually against measured burn.

Frontier assistants change month to month, so any agency workflow hard-wired to one model version carries a hidden rework tax: prompts, evals, and client-facing claims that must be retested every release cycle. The framework asks a simple question before you standardise: how much of this deliverable survives a model swap? Workflows that survive are the ones worth productising into retainers; workflows that break are experiments, not delivery lines. OpenAI's October 2, 2026 guide splits the GPT-6 family into three tiers for prototyping, feature development, and multi-step orchestration, which means a single client workflow can now span multiple models with different cost and quality profiles. Anthropic's Claude 5.5 release drew independent scepticism about announcement hype, and a September 2026 technical analysis found top benchmark scores do not reliably predict production performance. Agencies that map churn exposure per deliverable can quote model-agnostic retainers instead of absorbing every release as unpaid rework.

Frontier assistants change month to month, so any agency capability claim has a shelf life measured in weeks, not quarters. The Capability Drift Window is the gap between when a model ships and when your client-facing promises stop matching what the tool actually does. Agencies that treat model releases as procurement events (buy once, standardise, move on) accumulate silent drift: the retainer scope describes work the assistant could do six months ago, while competitors quote the current ceiling. The fix is a scheduled re-scoping cadence tied to release events, not annual reviews. When Anthropic shipped Claude 5.5 in September 2026, the practical question for agencies was not whether to switch, but which existing deliverables could now be quoted at a higher margin or lower hour count. Forrester's Q3 2026 research puts workflow integration, not model capability, as the bottleneck, which means drift shows up in delivery process before it shows up in output quality.

Price Claude work on metered token consumption or a hard usage cap, never on an uncapped flat retainer.

Route each deliverable type to the model tier that matches its complexity, then standardise only the interface your team uses daily.

IF your delivery model bills clients for AI-assisted content, code, or document analysis and you can meter token consumption per engagement, THEN start on the Free tier to prototype, move to Pro at $20/month for Claude Code and Research access, and only commit to the Max plan at $200/mo for 200+ requests per minute once concurrent client workflows demand it. IF you sell flat-rate retainers with no usage cap or pass-through clause, THEN defer Anthropic until you can price the consumption risk, because input and output token costs vary by model and cannot be white-labeled.

The choice is not which assistant is strongest; it is whether the agency needs seats for staff or an API for client products, and most shops end up running both. Seat tools carry the delivery work, while API access carries anything a client pays for directly, and the split should be decided per retainer rather than per preference. Whichever route an agency takes, the binding constraint is evaluation: model releases arrive faster than most teams can test them, so budget review time alongside licence cost.

A five to seven day sprint that takes a client from raw Anthropic account to a live Claude-powered workflow, using the API, Slack and Microsoft 365 connectors, and a documented handoff so the agency can move straight into a retainer.