How agencies are using Omnisend MCP to find missed opportunities faster

Omnisend MCP changes that by coming to a place where you already do much of your work. You connect Omnisend to ChatGPT, and instead of clicking around, you just ask questions. “Which campaigns made the most money last month?” “Which automations are dead weight?” “Where am I losing engagement?” You get answers in plain English.

We talked to Omnisend agency partners already using it with their clients to see how it’s changing their day-to-day.

Digital Strategy Hub: Catching what dashboards hide

Gediminas Dilertas is a Chief Marketing Manager at Digital Strategy Hub, an agency working mostly with Lithuanian ecommerce brands. His clients usually have an Omnisend account set up, but haven’t built out everything they could — welcome series, post-purchase flows, abandoned checkout, upsells. His job is figuring out what’s missing and what’s broken before recommending what to build next.

The audit phase used to be the slow part. “Cross-referencing automation IDs with performance data used to take an hour. Now it’s minutes.”

But the speed isn’t even the best part. During one recent client audit, he asked AI to pull every active automation side by side. One welcome series was carrying a huge share of total automation revenue — way more than the others. Scrolling through the dashboard, he probably would have noted that automation was doing well and moved on. Seeing it next to everything else made the imbalance impossible to ignore, and it changed what he recommended to the client.

The same query also caught an abandoned checkout flow doing basically nothing. Almost certainly a broken trigger. “Easy to miss manually, obvious when queried.”

The diagnostic phase that used to take hours now happens in a single session, and Gediminas walks into client conversations with the audit already done. That changes the dynamic of every meeting.

V8 Media: Finding segments you’d never think of

Carmen Theron is an Email Marketing Specialist at V8 Media. She works across a wide mix of B2C brands — some completely new to email marketing, others with more established email programs.

When they set up the account from scratch, the agency provides clients with essential automations and builds out more based on the client’s needs. For brands that already use email marketing, they identify gaps in their current automation setups and segmentation strategies, reworking strategies until they find what works.

One of their accounts wasn’t getting enough out of its subscriber base. Carmen knew there was an opening somewhere; she just couldn’t quite see it yet. She threw a prompt into ChatGPT.

“It came out with a very strong segment recommendation that hadn’t even occurred to me at the time. I probably would have thought of this on my own in due time, but not as quickly as AI identified it.”

She also uses it for weekly campaign decisions. For clients sending more than one campaign a week, AI helps her decide which segments to include and exclude, and how often to send based on what’s worked before. All of these suggestions are made based on real data, grabbed straight from the client’s Omnisend account.

Here’s how she describes the time savings:

“The time you spend clicking through multiple dashboards across Omnisend and Shopify is saved by the time you spend connecting to ChatGPT. It collects small nuggets of data from multiple pages, analyses it, and condenses it for you in moments. Instead of spending an hour reading 50 pages, it gives you a 1-page summary with the most important and relevant information.”

Ecommerce Boost: More time for the work that actually matters

JP Charles founded Ecommerce Boost, an agency that works with Shopify brands in Australia, New Zealand, North America, and Europe. His clients already have traffic and customers — they just know they’re leaving money on the table with email.

Before MCP, a lot of his team’s work was clicking around in Omnisend, pulling reports, checking automations, looking at segments, and trying to turn all of that into useful strategic advice for the client.

Now they just ask. Which campaigns performed best. Which segments drove the most revenue. Where engagement is dropping. What to test next.

“It allows us to spend less time digging for data and more time making strategic decisions. The value we provide is not just sending emails. It’s knowing what to send, who to send it to, when to send it, how to position the offer, how to segment the audience, and how to improve performance over time.”

JP thinks the shift is bigger than it looks:

“MCP has the potential to do for marketers what great AI coding tools have done for engineers. It can make every good marketer a 10x marketer, and every good email strategist a 10x email strategist.”

Where to start

All of this may sound a bit technical. MCP, AI, LLM, API — lot’s of abbreviations, not a whole lot of context. But trust us, the hardest part is connecting Omnisend to ChatGPT, and even that only takes a few minutes (we’ve got a guide for that).

After that, just be curious. Ask questions the way you already ask AI. There’s no wrong question, and you don’t need to phrase anything in a special way.

If you want a nudge to get going, here’s what our three agencies would type first.

Gediminas likes to start with an audit:

“Pull a full automation performance audit — every active automation with attributed revenue, open rate, click rate, and conversion count for the year to date. Most people have automations running that they set up once and never looked at again. That one query shows you exactly where your account is healthy, where it’s broken, and where the money is hiding.”

JP goes broader and looks at the whole program at once:

“Analyze the last 90 days of email performance and tell me the biggest opportunities to increase revenue from campaigns, automations, and segmentation.”

From there you can dig into whatever stands out — welcome flows, winback, VIPs, unengaged subscribers.

Carmen has a different suggestion, and it’s the one most people skip. Before you ask the LLM anything else, ask it what it can’t do:

“Ask if it can read all your data from the API and include what its limitations are. Let it tell you what it cannot analyze or distinguish. This will save you time and allow you to better train your model.”

Pick whichever one sounds most useful for where you are right now and just try it. You’ll know within a few minutes whether this changes how you work.