The Agency Pitch Deck for Agentic Commerce, Annotated

Most agentic commerce pitches fail for the same reason: they open with a slide about ChatGPT and close with a shrug. The client nods, says it's interesting, and nothing gets scoped. The fix isn't a better platform explainer — clients don't need convincing that AI shopping is real, they need to see their own store's gap in it, in language that ends in a proposal. This is the deck structure we hand our own Agency Partners and the one we use ourselves. Six slides, each doing one specific job, none of them generic.

What should be in an AI commerce sales pitch?

An AI commerce pitch needs exactly two ingredients most agencies skip: proof the wave is real and dated, and proof specific to the client sitting across the table. Everything else — your capability slide, your case studies, your pricing — sits on top of those two. Get the order wrong and the meeting turns into a lecture; get it right and the client is asking you what it costs before you reach the proposal slide.

The other thing a good pitch needs is restraint. You are not trying to explain the entire agentic commerce landscape in one meeting. You're trying to move the client from 'I've heard of this' to 'audit my store,' which is a much smaller cognitive lift — and it's the same lift regardless of whether the client ends up buying an install from you directly, through a white-label partner, or eventually in-house. If you want the fuller context on that decision before you build the deck, see pricing agentic services for agencies and the agency agentic commerce playbook.

Slide 1-2: frame the wave without hype

Open with three facts, not opinions. Shopify shipped Brand Agents and Copilot Checkout in its Spring '26 edition on June 17, 2026 — both are live, both are Shopify Plus features today, and both change how a shopper can buy without ever loading a product page. Off-site, the same shift is running in parallel: shoppers are asking ChatGPT, Claude, Gemini, and Perplexity what to buy, and Amazon shoppers are asking Rufus. None of this is speculative — it shipped, it's dated, and it's the reason the client took the meeting.

Agentic commerce

Commerce where an AI agent — on-site (a Brand Agent, on-site AI search) or off-site (ChatGPT, Perplexity, Rufus) — reasons over a store's product data on the shopper's behalf, and in some flows completes the purchase itself. The store's job shifts from persuading a human eye to answering a machine's questions correctly.

Keep this section to two slides and under ninety seconds of talk time. The mistake most agencies make is spending the first fifteen minutes on platform education — by the time you get to the client's own numbers, you've used up the attention you needed for the part that actually sells. State the wave, then move.

Slide 3: show them their own gap

This is the slide that decides the meeting, and it's also the one most agencies skip because it takes ten minutes of prep. Before the call, run the client's domain through the free Agentic Commerce Readiness Score — a 100-point assessment covering catalog completeness, structured data, and PDP machine-readability that takes three minutes — and separately run the AI Citation Check to see whether the client's products are actually showing up when assistants like ChatGPT and Perplexity get asked relevant buying questions. Screenshot both. Put the client's actual score and actual citation results on the slide, not a category average.

This single move — walking in with the client's own number instead of an industry statistic — is what separates a pitch that lands from one that gets a polite 'send us something in writing.' A generic score means nothing to a business owner. Their own score, especially a low one next to a competitor's higher one, is the single most persuasive thing in the deck, and you didn't have to write a word of copy to make it.

1. 1

   Run the Readiness Score on their domain

   Three minutes, no login required. Screenshot the overall score and the three sub-scores (catalog, structured data, PDP readiness) before the call.

2. 2

   Run the Citation Check on two or three of their hero products

   See whether ChatGPT, Claude, Gemini, or Perplexity mention the client's products when asked a natural buying question in their category.

3. 3

   Pull the same two checks on one direct competitor

   A side-by-side — client's score and citations versus a named competitor's — turns an abstract gap into a competitive one, which is what actually motivates budget.

4. 4

   Build one slide, not five

   Score, sub-scores, citation screenshot, competitor comparison. Resist the urge to explain every point deduction — that's what the audit deliverable is for.

Slide 4: the cost of waiting

Once the client has seen their own gap, name what it costs to leave it open — concretely, not abstractly. If a competitor's products are getting cited by AI assistants and the client's aren't, that's demand the client is not competing for at all; the shopper never sees a comparison, because only one store was in the answer. If the client's catalog isn't structured well enough for a Brand Agent to answer product questions accurately, any Brand Agent they configure later will underperform until the underlying data work is done — so waiting doesn't just delay the upside, it compounds the cost of eventually catching up. Keep this slide short: one or two sentences tied directly back to the gap slide, not a market-sizing exercise nobody in the room can verify.

How do agencies sell agentic commerce to clients?

By giving them a ladder of options instead of one all-or-nothing pitch. Clients rarely say yes to a big unscoped commitment in a first meeting, but almost all of them will say yes to naming where they currently sit on a spectrum — which is a much easier conversation and a natural bridge into a proposal.

The ladder itself has four rungs, and your job in the meeting is to help the client self-identify where they are, then propose the next rung up — not the top of the ladder in one jump.

1. 1

   Rung 1: do nothing

   Valid for a client with no agentic-ready platform yet and no evidence of lost AI-channel demand. State this honestly if it's true — it builds trust and costs you nothing, because most clients on this rung aren't actually here after they've seen their gap slide.

2. 2

   Rung 2: get scored and audited

   The entry point for almost everyone. A formal readiness audit — deeper than the free three-minute score — covering catalog attribute coverage, structured data, PDP readiness, and off-site citation gaps, with a prioritized fix list.

3. 3

   Rung 3: fix the foundation

   Catalog enrichment and structured data work, informed by the audit. This is the unglamorous layer that makes every agentic feature actually work — see catalog enrichment for AI for what that work looks like in practice.

4. 4

   Rung 4: configure and launch

   Brand Agent setup, Copilot Checkout configuration, on-site AI search, and Commerce GEO for off-site visibility — a fixed-scope install like our own Agentic Commerce Setup, live in two weeks once the foundation is sound.

How do I pitch Brand Agents to my clients?

Narrow the general pitch to the one feature clients ask about by name. Most clients who've read the Shopify announcement don't ask about "agentic commerce" — they ask "what's a Brand Agent, and do we have one." Answer that question directly and specifically, in this order: what it is in one sentence, what it depends on (their own catalog data, which you already scored), what it's currently limited to (Shopify Plus, for now), and what configuring one well actually requires. Bring one screenshot of a Brand Agent conversation if you have a demo store to show — an abstract description of a chat interface undersells how different this is from a search bar.

The trap to avoid is answering the Brand Agents question as a platform feature explainer. The client isn't asking whether Shopify shipped something; they read that already. They're asking whether it will work well for their catalog, and the honest answer for most mid-market merchants is: not yet, because the underlying product data isn't structured enough for an agent to answer confidently. That's not a discouraging answer — it's the one that turns into an audit conversation and eventually a design conversation about how the agent should actually behave once it's configured; see brand agent conversation design for what that involves once you're past the pitch.

Slide 5-6: the proposal and the close

The proposal slide names one thing: the specific next rung for this specific client, with a scope list and a price, not a range of possible futures. If the meeting surfaced a rung-2 client, the proposal is the audit — named deliverables (attribute coverage report, structured data gap list, citation findings, prioritized fix list), a timeline, and a price. If the client's gap slide already showed obvious foundation problems, you can propose straight into rung 3. Whatever the rung, one proposal, one price, no menu of five options presented simultaneously — a menu at this stage produces indecision, not a signed scope.

The close is the smallest yes that moves the engagement forward, with a date attached. "Let's explore AI together" is not a close — it's a way to end a meeting with nothing scheduled. "We'll send the audit scope by Friday and can start the week after" is a close. If the client isn't ready to commit to the audit on the spot, the fallback close is scheduling the follow-up meeting before anyone leaves the room, not promising to "circle back."

• Close 1 — sign the audit. The strongest outcome: a scoped, priced audit agreed to in the room or within 48 hours.
• Close 2 — schedule the audit-scoping call. If budget approval takes longer than one meeting, get a specific date on the calendar before you leave.
• Close 3 — send the gap slide, dated. At minimum, leave the client with their own Readiness Score screenshot and a one-line follow-up date. A dated follow-up outperforms an open-ended "let us know."

Reusing this deck across clients

The efficiency of this structure is that four of the six sections — the wave, the options ladder, your capability framing, and the general close pattern — stay fixed across every pitch. Only slide 3 (the client's own gap) and slide 5 (the specific proposal) change meeting to meeting. That means once you've built the deck shell once, adding a new prospect to your pipeline is a ten-minute prep task, not a rebuild. Keep a running comparison table of every client you've scored — it becomes its own asset, letting you show a new prospect "here's where twelve stores in your category landed" without naming anyone, which is a stronger opener than a cold platform slide on its own.

If you're building this deck as part of deciding how you'll actually deliver the work once a client says yes — white-label, in-house, or referred — that's a separate decision with its own math, covered in white-label vs. in-house agentic delivery. Don't let that decision hold up building the deck; you can pitch and sell the audit before you've locked delivery, and the audit itself is often the reconnaissance that makes the delivery decision easier.