Marketo AI in 2026: What Adobe Actually Ships to Marketo Engage Customers
Search “marketo ai” and you land in a mess that Adobe created. There are at least four things wearing that name — Marketo AI (the skills panel in your instance), the AI Assistant in Email Designer, the generative AI inside Dynamic Chat, and the AI Assistant in Adobe Experience Platform that Adobe markets heavily across the Experience Cloud. Only three of those are things a Marketo Engage customer can actually turn on. The fourth is for other Adobe products entirely.
If you run demand gen on Marketo, here’s what’s real, what’s gated, and what still lands on a human.
The four things called “Marketo AI,” sorted
1. Marketo AI (the skills panel) — open beta, read-only. This is the newest and the one most people mean. You reach it from the Marketo AI tile on your My Marketo home screen. As of the July 2026 release (rollout started July 10), it carries two skills in open beta:
- Product Knowledge — natural-language answers about Marketo itself, grounded in Adobe’s documentation. Read the fine print: it has no access to your subscription data. It knows how smart campaigns work in general; it does not know how yours is configured.
- Investigate Leads — this one does read your instance. Ask why a specific person never hit MQL, or why they got excluded from a campaign, and it walks the lead’s activity history and explains in plain language. Adobe is explicit that it “tells you what happened, but does not make changes to the lead record or program membership.”
Both require account manager approval for beta access plus acceptance of Adobe’s Core Gen-AI terms. Neither is self-serve, and neither takes an action.
2. AI Assistant in Email Designer — GA, and the one you’ll use daily. This one is generally available, not beta. It generates subject lines, preheaders, individual content sections, whole emails, and images via Firefly, with controls for persona, journey stage, tone, and brand assets. Two gates people trip over: your org has to have agreed to Adobe’s Gen-AI terms, and a Marketo admin must grant the “Access AI Assistant” permission on the user role — it is off by default. If your copywriter says the button isn’t there, that’s usually why, not entitlement.
3. Dynamic Chat’s generative AI — conversational, front-of-site. Dynamic Chat targets people and accounts on your website, and its generative layer processes intent signals and past behavior to personalize the conversation in real time. It books meetings against Outlook and Gmail calendars and can trigger Marketo activities natively. This is a genuinely different product from the skills panel, and it’s the one that touches prospects rather than marketers.
4. AI Assistant in Adobe Experience Platform — not yours. This is the distinction worth internalizing, because Adobe’s marketing blurs it constantly. The AI Assistant in Experience Platform supports Experience Platform, Real-Time CDP, Journey Optimizer, and Customer Journey Analytics. Marketo Engage is not on that list. When you read an Adobe keynote recap about AI Assistant answering operational questions about audiences and datasets, or about Agent Orchestrator running multi-step tasks, that is describing a different stack than the one your Marketo instance sits in. Unless you’re also an AEP customer, none of it applies to you.
The same caution goes for “Adobe Sensei.” Sensei is Adobe’s umbrella brand for machine learning across the whole company, and it appears in a lot of Marketo-adjacent marketing. It is not a feature you enable in Marketo Engage. Judge what you’re buying by the four items above, not by the brand name on the slide.
What Marketo already does that you should not pay an AI vendor to redo
This is where a lot of “AI for Marketo” pitches fall apart, and it’s worth being blunt about it: Marketo’s core competency is event-driven orchestration, and it’s good at it. Trigger campaigns fire on form fills, score thresholds, email clicks, data value changes, and web activity. Smart lists segment on behavior. Programs handle nurture logic and lifecycle progression. If someone tells you Marketo can’t react to things happening, they’re selling you something.
So the honest gap is not “Marketo needs triggers.” The gap is that Marketo triggers Marketo things. It moves a lead to a nurture stream, changes a score, sends an email, syncs a field to Salesforce. What it does not do is the human coordination that hangs off those events — the part where a demand gen manager notices the webinar had 140 registrants and 31 attendees, pulls the no-show list, drafts a different follow-up for the no-shows than the attendees, pings the two AEs whose named accounts showed up, and puts a debrief on the calendar. Marketo can fire the trigger. Somebody still has to run the meeting, write the Slack message, and chase the reply.
That’s the seam. Not automation Marketo lacks, but coordination that lives in email, calendars, and chat, downstream of a Marketo event.
The REST API: self-serve, and the limits are specific
Marketo’s REST API is one of the better-documented ones in B2B marketing software, and — unlike a lot of vertical enterprise tools — a customer can provision their own credentials without a partner agreement. You do need admin rights on the instance.
The setup is two steps that people routinely do in the wrong order:
- Admin → Users & Roles — create a role with the permissions you need, then create an API-Only user assigned to it. The role is what scopes what any integration can touch, so scope it narrowly.
- Admin → Integration → LaunchPoint → New Service → Custom — create a custom service owned by that API-only user. Its View Details link exposes the Client ID and Client Secret. Your Identity URL lives under Admin → Integration → Web Services, and your Munchkin ID identifies the instance.
Authentication is 2-legged OAuth 2.0: exchange client ID and secret at the Identity URL for a bearer token that lives 3,600 seconds. Note the deprecation — passing the token as an access_token query parameter is going away after July 31, 2026. Any integration you stand up now should send it in the Authorization header from day one.
The quotas are worth writing on a whiteboard before anyone builds anything:
- 50,000 API calls per day per subscription, resetting at 12:00 AM CST. Increasable through your account manager.
- 100 calls per 20 seconds per instance.
- Ten concurrent calls, maximum.
That daily ceiling sounds generous until you remember it’s shared. Your Salesforce sync, your data warehouse ETL, your enrichment vendor, and your webinar platform are all drawing from the same 50,000. Adobe exposes a usage endpoint returning daily and 7-day statistics — check it before you assume you have headroom for a new integration, and design anything new around bulk endpoints and batched reads rather than per-lead polling.
Whether ChatGPT and Claude can reach Marketo
Neither ChatGPT nor Claude ships a native, one-click Marketo connector in its consumer connectors directory. What exists is Adobe’s own Marketo MCP server at marketo-mcp.adobe.io/mcp, exposing 100+ operations across forms, programs, smart campaigns, leads, emails, snippets, lists, and folders. Three caveats, all of which matter:
- It’s a limited beta — you request access via a form with your Munchkin ID.
- Authentication passes your client ID, secret, and Munchkin ID in HTTP headers, not OAuth. That’s why it works in Claude Desktop, Claude Code, Cursor, VS Code Copilot, and Codex, but not the claude.ai web app, which requires OAuth for custom connectors.
- It executes REST calls on your behalf — meaning every operation draws down the same 50,000/day quota above.
We covered the beta in more depth in Claude + Marketo: what the integration can and can’t do, and Adobe’s broader agent posture in Adobe MCP.
There’s a structural limit underneath all of this that no beta access fixes:
Both ChatGPT and Claude will now do work on a schedule, unattended. Neither reacts to events. A task that runs hourly still finds out an hour late, and one that runs each morning finds out tomorrow — the trigger is the clock, never the thing you actually care about.
Which is exactly backwards from how Marketo works. Your instance is event-driven; the assistant sitting next to it is not. Ask Claude to pull a smart list and draft follow-up copy and it does that well. Ask it to notice that an MQL sat untouched for three days and chase the AE, and you’re outside what an MCP client does.
Where Carly fits around a Marketo instance
Carly is the layer for the coordination Marketo hands off. It watches email and calendar, and it runs workflows that fire on events rather than on a chat prompt — so the follow-up after the webinar, the meeting that has to get booked with the two AEs, the reminder that a demo request from a named account has gone unanswered for 36 hours, all happen without a person babysitting them.
For Marketo Engage itself, Carly is BYO-key: you create the API-only user and LaunchPoint custom service exactly as above, then paste the client ID, secret, and Munchkin ID at carlyassistant.com/integrations. Adobe is also a native Carly connector, which is a reasonable adjacent path if the work you’re automating touches Adobe’s creative and document side — asset handoffs, campaign creative review — rather than the Marketo database. The two are separate connections; the native Adobe connector does not reach your Marketo instance.
Pricing: free, unlimited Zapier-style workflows; AI agents from $35/month. If your demand gen motion is already paying enterprise money for Marketo, the useful comparison isn’t cost — it’s whether the coordination work currently sitting in your marketing ops person’s head gets to run on its own.
If you’re evaluating more broadly, the best AI workflow automation tools and the best AI CRM tools cover adjacent ground.
FAQ
Is Marketo AI included in my Marketo Engage subscription?
The Email Designer AI Assistant is generally available, but requires your organization to accept Adobe’s Core Gen-AI terms and a Marketo admin to grant the “Access AI Assistant” permission on the user role — it’s off by default. The Marketo AI skills panel (Product Knowledge and Investigate Leads) is in open beta and requires account manager approval. Check your contract rather than assuming.
Does Marketo use Adobe Sensei?
Sensei is Adobe’s company-wide brand for machine learning, not a toggle inside Marketo Engage. When evaluating what you actually get, look at the named features — Marketo AI skills, Email Designer AI Assistant, Dynamic Chat’s generative layer — rather than the Sensei branding on Adobe marketing material.
Can the AI Assistant in Adobe Experience Platform answer questions about my Marketo data?
No. Adobe’s documentation lists Experience Platform, Real-Time CDP, Journey Optimizer, and Customer Journey Analytics as the supported applications. Marketo Engage is not among them. Marketo’s own Investigate Leads skill is the equivalent capability for Marketo data, and it’s a separate, beta-gated product.
Can Marketo AI change something in my instance?
Not currently. Product Knowledge has no access to your data at all, and Investigate Leads is explicitly read-only — Adobe states it does not modify lead records or program membership. Anything it diagnoses, you fix by hand.
What is Marketo’s API rate limit?
50,000 calls per day per subscription, resetting at 12:00 AM CST, with a rate limit of 100 calls per 20 seconds and a maximum of ten concurrent calls. The daily quota can be raised through your account manager. Every integration on the instance shares that pool, so check the usage endpoint before adding another.
Do I need a developer to connect Marketo to an AI tool?
You need a Marketo admin, not necessarily a developer. Creating the API-only user and the LaunchPoint custom service is a UI task that takes a few minutes. What comes after depends on the tool — Adobe’s MCP server needs a beta approval and a desktop MCP client, while a BYO-key platform takes the same three credentials pasted into a settings page.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
