Pricing Agentic Services: Fixed-Scope Logic for Agencies

Every agency figuring out agentic commerce delivery eventually asks the same question: what do we charge for this? It is a fair question and a hard one, because agentic work does not map cleanly onto either of the two pricing models agencies already know. It is not a website build with a fixed spec and a launch date. It is not ongoing retainer work with a stable monthly scope either. It sits in between, and pricing it like either extreme leaves money on the table or, more commonly, leaves the agency underwater on its first few engagements.

This is not a numbers article — GigaCommerce does not publish rates, and neither should you without knowing your own cost structure and market. What we can give you is the pricing logic: which parts of the work are fixed-scope and belong on a fixed fee, which parts are production work that scales with volume, and which parts are ongoing and belong on a retainer. Get the structure right and the number becomes a market question you can answer with confidence. Get the structure wrong and no number saves the engagement.

How should an agency price AI commerce work

Start by separating the three kinds of work you are actually selling, because they behave differently and deserve different pricing logic. Bundling them into one hourly number is the single most common mistake we see agencies make when they first quote this work.

An install — catalog audit, structured data, Brand Agent configuration, Copilot Checkout configuration — is a bounded deliverable. It has a start, an end, and a defined set of things that must be true when you hand it back. That shape belongs on a fixed fee. Catalog enrichment at scale is production work with a countable unit — SKUs brought to a schema — and belongs on a throughput-based rate. Commerce GEO is neither: it is a maintained state that decays the moment you stop paying attention to it, and it belongs on a retainer. Treating all three the same way is how agencies end up either overcharging for the easy parts or, far more often, undercharging for the parts that eat real hours.

The other reason to separate them: clients budget differently for each. A one-time install fits a project budget a client already knows how to approve. A retainer fits an ongoing marketing line item. A throughput-based catalog engagement fits a scope-driven statement of work. Forcing all three into one number makes the proposal harder to approve, not easier.

Should agentic commerce installs be fixed-fee or hourly

Fixed-fee, with a specific and named exception below. The case for fixed-fee is the same case that applies to any well-defined technical deliverable: the scope is knowable in advance, the client wants budget certainty before they approve spend, and hourly billing actively punishes your own competence. The tenth time your team configures a Brand Agent, they will do it in a fraction of the hours the first install took — and under an hourly model, that efficiency gain becomes a pay cut. Fixed-fee is the only structure where getting better at the work makes you more profitable instead of less.

Fixed-scope logic

A pricing approach where the deliverable list is defined up front (catalog attributes enriched to a named schema, structured data live on every PDP, agent trained and tested against a defined conversation set, checkout configured for the client's tax and fulfillment rules) and the fee is set against that list, not against hours worked.

The named exception is discovery. Before you can scope an install fixed-fee, you need to know what you are walking into — catalog size, current attribute coverage, platform version, how many SKUs have compatibility data, whether the client is even on Shopify Plus (a prerequisite, since Brand Agents and Copilot Checkout are Plus-only today). Price discovery as a small, separate fixed-fee or included-but-capped phase. Never fixed-bid the install itself before you have seen the catalog — that is how agencies eat six-figure scope creep on a five-figure deal.

1. 1

   Paid discovery, fixed and capped

   A short, separately priced phase: catalog audit, attribute coverage check, platform prerequisites confirmed. This is where you find out what install #1 actually requires before you commit to a number.

2. 2

   Fixed-fee install, scoped from discovery findings

   Once you know the catalog's real state, quote the install against a named deliverable list — not an hours estimate. The client buys certainty; you price in the ramp you found in discovery.

3. 3

   Change orders for scope that discovery missed

   Fixed-fee does not mean unlimited. If the client adds a second storefront or a new product line mid-engagement, that is a change order, not free hours. Write this into the contract before you sign it, not after the scope creeps.

How do agencies price catalog enrichment work

Catalog enrichment is the part of the install most likely to blow your estimate, because it is production work masquerading as consulting. The fix is to price it like production work: by throughput against a defined schema, not by the hour and not bundled invisibly into the install fee.

Define the unit first. A "SKU enriched" is not enriched until it hits every field in the category's attribute schema — not partially filled, not enriched-except-compatibility. Ambiguity here is where margin disappears, because your team will quietly do more work per SKU than the price assumed, one exception at a time. See category attribute templates for how to standardize the schema before you price against it.

From there, price per SKU-to-schema, tiered by category complexity — a simple accessory category enriches faster than a category with heavy compatibility data ("works with," "fits"). Quote a weekly throughput your team can actually sustain once trained, and set client expectations against that number, not against the whole catalog finishing at once. This mirrors the priority-order approach — hero SKUs first, long tail to a baseline — which gives you a natural way to phase the invoice alongside the work.

Structuring a Commerce GEO retainer

GEO does not fit a project price because there is no finish line. Getting cited by ChatGPT, Perplexity, or Google AI Overviews this month says nothing about whether you are cited next month — the assistants re-crawl, re-rank, and change how they source answers on their own schedule, not yours. Structure GEO as a retainer with a visibility target, the same way an agency would price ongoing SEO, not the way it prices a website launch.

A workable retainer structure has three layers: a baseline audit (fixed-fee, one-time — see the AI citation audit guide), an implementation phase (structured data, llms.txt, content built for citation), and then an ongoing monitoring-and-iteration retainer that tracks citation rate over time and adjusts as assistants change how they source. For the budget conversation with clients who are used to funding SEO but not GEO, the GEO vs. SEO budget split framing is the fastest way to make the ask land as a reallocation instead of new spend.

• Baseline audit — fixed-fee, establishes current citation rate across the assistants that matter for the client's category.
• Implementation — fixed-fee or phased, the structured-data and content work that makes citation possible in the first place.
• Monitoring retainer — the ongoing piece, priced monthly, tied to a citation-rate benchmark so the client can see what they are paying to maintain.

Anchor the retainer conversation in a number the client can watch move, using something like the citation rate benchmark as the reference point. A retainer without a visible metric reads as a subscription with no proof of value, and that is the fastest way to get cancelled at renewal.

The ramp-pricing trap

This is the mistake that costs agencies the most money, and it happens quietly: pricing install #1 at the same rate — or worse, the same hours estimate — as install #5. Your team's first agentic commerce install will take meaningfully longer than the fifth one. They are learning the platform's actual behavior, not the documentation's version of it, discovering the client's catalog exceptions, and building the internal checklists and QA process your fifth client will benefit from for free. If you price the first install as if that learning cost does not exist, you are not winning the deal — you are subsidizing your own training program with agency margin, which is the same conversation covered from the capacity side in white-label vs. in-house delivery.

There are three honest ways to handle this, and "pretend it isn't happening" is not one of them.

1. 1Price the ramp into the fee. Quote early installs higher than your steady-state rate will eventually be, and say so internally — this is not the number you will charge once your team is fast.
2. 2Price the ramp into the timeline. Keep the fee flat but be honest with the client about a longer delivery window on the first engagement. Certainty on price, less certainty on date.
3. 3Frame it as a founding-client rate, in writing, with an end date. If you want to win early logos at a lower number to build a portfolio, that is a legitimate strategy — but name it, cap how many clients get it, and do not let "founding rate" quietly become your permanent rate because renegotiating upward later is much harder than starting there.

The version to avoid is the default one: quote a market rate you have not tested, absorb the ramp silently because the deal felt too good to complicate, and discover three installs in that the engagement has been running at a loss the whole time. If you are still deciding whether to build this capability in-house at all versus partnering for delivery while you find your rate, that decision — and how it interacts with your margin during the ramp — is the whole subject of white-label vs. in-house agentic delivery.

Putting it into a proposal

In practice, most agencies end up presenting agentic commerce work as three or four line items rather than one bundled number, because that maps to how the work is actually priced and gives the client a place to phase spend if budget is tight.

• Discovery — fixed-fee, capped, short. The catalog audit that makes everything downstream priceable.
• Install — fixed-fee, scoped from discovery, named deliverable list, change-order clause for scope not covered.
• Catalog enrichment — throughput-priced per SKU-to-schema, phased hero-first, with the schema itself as a named deliverable.
• GEO retainer — ongoing, tied to a citation-rate metric the client can watch move.

This structure also gives you a natural upsell path: a client who buys discovery and the install can be brought back for the GEO retainer once the catalog and structured data are already in place — the agency pitch deck is worth reviewing for how to sequence that conversation. Whatever numbers you land on, keep them off the page in marketing materials and case studies; publish the structure and the logic, not the rate card, and let each proposal reflect your actual cost basis and the market you are selling into.