MoEngage Raises $100M to Scale Merlin AI Agents: A Two-Week Pilot for Marketers

A funding jolt for AI marketing, and a clear plan you can run next Monday\n\nMoEngage just put fresh fuel behind its Merlin AI agenda. The company raised $100 million to accelerate AI agents that decide who to target, where to reach them, and what to say, across mobile, web, email, and messaging. That matters because the hard part in modern lifecycle marketing is not writing copy, it is deciding which customer should see which offer, on which channel, at what time.\n\nAccording to the official announcement, the round is led by Goldman Sachs Alternatives with A91 Partners, and the capital will expand Merlin AI, including its offer and campaign decisioning agents. See the details in the MoEngage gets $100M funding announcement. This article breaks down what the raise signals for AI marketing, what Merlin AI likely does under the hood, and how to run a two-week pilot that shows lift in conversion rate, incremental revenue, and customer acquisition cost payback. You will leave with a day-by-day plan that a lean team can execute without stalling other campaigns.\n\nFor related two-week templates, see our guides on the WPP and Google AI pilot playbook and the TikTok Smart+ split test playbook.\n\n## What Merlin AI is solving, in plain language\n\nMost teams use a stack of point tools for analytics, orchestration, and content. The friction is not a lack of options. It is the pace of decisioning. Marketers still stitch together segments, channel picks, send times, and message variants by hand. Merlin AI promises to convert that human decision tree into a set of agents:\n\n- Audience and eligibility agents that decide which customers are in play based on recency, frequency, and monetary behavior, as well as product views and churn risk\n- Offer agents that match customers to discount, free trial, or product bundles aligned to predicted margin and propensity\n- Channel decisioning agents that pick email vs push vs SMS vs in-app, plus send time and frequency\n- Creative generation and ranking that drafts, scores, and chooses message variants based on prior performance and guardrails\n\nIf these agents are effective, they compress the time from idea to live test from weeks to days, and they make lift more repeatable. The new capital suggests MoEngage will harden these agents for enterprise scale and compliance.\n\n## The competitive context and why the timing is smart\n\nBraze, Salesforce Marketing Cloud, Adobe Journey Optimizer, Iterable, and CleverTap have each shipped AI helpers, but most teams still rely on manual rules and scheduled sends. Two forces hold back adoption: measurement anxiety and data plumbing. Leaders do not want to flip a switch without a credible way to prove incrementality, and many teams worry about the data they will need to feed the model.\n\nMerlin AI is positioned to attack both. First, the decisioning agents allow smaller controlled experiments rather than big bang migrations. Second, the platform already sits on first party event and profile data inside MoEngage, which trims the integration tax. For a complementary approach inside Adobe’s stack, review our breakdown of the Adobe Journey Agent pilot.\n\n## Your two-week pilot, designed for executive credibility\n\nThe goal is simple. Automate channel selection and message variants on one lifecycle moment, prove lift with a holdout, and calculate CAC payback. Keep scope tight so results are unambiguous.\n\n### Pilot scope\n\n- One journey, one KPI: pick either onboarding activation, cart recovery, or at risk churn save\n- One audience of 100K to 1M users, split into treatment and control\n- Channels allowed: email, push, in-app, and SMS if you already have consent\n\n### Primary metrics\n\n- Conversion rate: percent of users who complete the target action within the window\n- Incremental revenue: revenue in treatment minus revenue in statistically matched control\n- CAC payback: paid media and incentive costs divided by gross profit from pilot conversions, measured until payback hits 1.0\n\n### Guardrails\n\n- Frequency cap by person per week\n- Deliverability and consent: do not degrade deliverability, respect opt outs, and log model choices for audit\n\n## Day-by-day plan, with responsibilities\n\n### Day 1, align the brief\n\n- Executive sponsor signs off on one KPI, one audience, and a clear baseline\n- Define the measurement window, usually 7 days for onboarding and cart, 14 to 28 for churn\n- Assign roles: marketer as pilot lead, data analyst for measurement, lifecycle manager for creative guardrails\n\n### Day 2, wire up the audience and events\n\n- Confirm event names for key actions, for example app_opened, item_added_to_cart, purchase_completed\n- Finalize identity resolution, email and device mapping, and opt in status\n- Draft the control logic: 20 percent no AI decisioning, run your current best practice instead\n\n### Day 3, connect channels and constraints\n\n- Activate email, push, and in-app in MoEngage, ensure suppression lists and quiet hours are respected\n- Set frequency caps: for example 2 emails, 2 push, 1 in-app per user per week in treatment\n- Confirm UTM parameters and campaign naming for downstream analytics\n\n### Day 4, load creative building blocks\n\n- Provide brand voice guardrails and approved offer templates\n- Upload 3 to 5 subject line patterns and 2 to 3 body templates per channel\n- Add prohibited claims and compliance snippets, for example price disclaimers\n\n### Day 5, enable Merlin AI agents for decisioning\n\n- Turn on channel decisioning with channel availability, costs, and ceilings\n- Turn on offer agent with known constraints, for example maximum discount 15 percent in any 30 day window\n- Set the experimentation rate, for example 80 percent exploitation and 20 percent exploration for creative\n\n### Day 6, preflight and shadow test\n\n- Run shadow tests on a sample cohort and log the agent’s suggested channel and message, do not send\n- Review a 100 sample of recommendations for brand safety and relevance\n- Adjust guardrails where needed\n\n### Day 7, go live with treatment and control\n\n- Treatment uses Merlin AI for channel selection and message variant ranking\n- Control gets your current single channel or fixed sequence, for example email at 10 a.m. next day\n- Freeze other overlapping experiments for the same audience\n\n### Day 8 to 10, monitor and adapt within guardrails\n\n- Watch deliverability, bounce, and complaint rates\n- Check early proxy signals, for example click through and add to cart\n- If complaint rate or unsubscribes breach thresholds, tighten frequency caps immediately\n\n### Day 11, interim read and creative refresh\n\n- Analyst runs interim lift, not final, using matched control or CUPED adjustment to reduce variance\n- If a creative cluster underperforms, replace the worst 20 percent with fresh variants, keep guardrails intact\n\n### Day 12 to 13, economics pass\n\n- Pull incentive costs and any paid retargeting spend tied to the pilot\n- Compute gross margin on incremental orders, not on all orders, to avoid double counting\n- Calculate CAC payback: pilot spend divided by gross profit from incremental conversions\n\n### Day 14, final read and go or no go\n\n- Share a one page executive readout: KPI lift with confidence interval, incremental revenue, CAC payback, and next step\n- If payback is less than or equal to 1.0 within 60 days and there is no deliverability damage, expand to the next journey\n\n## Measurement design that stands up in a boardroom\n\nThe fastest way to discredit an AI pilot is fuzzy math. Use a clean A or B split or propensity matched control. For small audiences, CUPED or pre period adjustment reduces noise. Report both absolute lift and incremental revenue. When you discuss cost, separate distribution costs, incentives, and internal labor.\n\nDefinitions you can reuse\n\n- Incremental revenue equals revenue in treatment minus revenue in control, adjusted for pre period behavior\n- Gross profit equals incremental revenue times gross margin percentage\n- CAC payback equals pilot spend divided by gross profit, lower is better\n\nIf you do not have a steady baseline, include a 10 to 20 percent ghost holdout that is never marketed to during the pilot window. That ghost holdout becomes your sanity check if external promotions or seasonality hit during the test.\n\n## Data and integration checklist, before you turn any knobs\n\n### Data\n\n- A unified customer profile with email, device tokens, and recent behavioral events\n- Product catalog with price, margin, inventory flags, and offer eligibility rules\n- Consent status per channel, with last updated time and audit log\n\n### Compliance and governance\n\n- Role based access controls, separate who can edit templates from who can change experiments\n- PII handling standards, redaction at rest in test environments, and audit trails for model decisions\n- Revocation process for any creative that breaks legal or brand rules\n\n### Documentation\n\n- A runbook that defines what happens if the agent suggests a channel you cannot use, for example SMS not allowed in a region\n- A fall back message and default send time if the decisioning service is degraded\n\n## What good looks like in week two, with benchmarks you can defend\n\nEvery company is different, but most consumer brands see the following when they move from fixed cadence to AI decisioning on a single journey:\n\n- 3 to 7 percent absolute lift in conversion rate on onboarding activation within 14 days\n- 5 to 12 percent incremental revenue lift on cart recovery in 7 to 10 days\n- 10 to 20 percent reduction in send volume for the same or better revenue, which helps deliverability\n\nA simple model for the readout\n\n- Baseline: 500,000 users enter the cart recovery journey per week, baseline conversion is 6 percent, average order value is 60 dollars, gross margin is 40 percent\n- Treatment uses Merlin AI for channel and message, control uses email only\n- Observed results after 7 days: treatment converts at 7.2 percent, control at 6.1 percent\n- Incremental orders equal 5,000, incremental revenue equals 300,000 dollars, gross profit equals 120,000 dollars\n- Pilot spend equals 40,000 dollars in SMS and incentives, CAC payback equals 0.33, a green light\n\nThese are not promises, they are reference points to help you size your opportunity and get executive sponsorship.\n\n## Where the new capital likely goes, and what to watch next\n\nPublic reporting indicates the round blended primary and secondary shares, roughly 60 percent primary and 40 percent secondary, with a focus on strengthening North America go to market and accelerating Merlin AI. See the TechCrunch on the funding round for context.\n\nExpect to see deeper integrations with transaction and product catalog data, more robust offer optimization that includes margin constraints, and better transparency in how agents pick channels. Also watch for prebuilt playbooks for retail, media, and financial services. Those playbooks will determine how fast a new team can go from zero to a credible pilot.\n\n## How to brief your team tomorrow morning\n\nGive your team a one page brief and a sprint name. Clarity speeds execution.\n\n- Business goal: increase onboarding activation from 36 percent to 40 percent in 14 days\n- Scope: automate channel selection and message variants for users on days 1 to 7 after signup\n- Metrics: conversion rate, incremental revenue, CAC payback, plus deliverability and complaint rate as guardrails\n- Budget: up to 25,000 dollars in SMS and incentives\n- Decision right: lifecycle lead owns day to day, head of growth approves guardrail changes\n\nIf you use Upcite.ai internally to capture experiments, you can clone this pilot as a template with the audience, guardrails, and readout math, then hand it to each product line. Documenting the pilot like this removes ambiguity when you expand.\n\n## Pitfalls and how to avoid them\n\n- Channel crowding: AI decisioning can bias toward the lowest cost channel and overuse it, so enforce channel ceilings and frequency caps\n- Offer cannibalization: do not hand discounts to users who would buy anyway, use eligibility rules and margin caps\n- Creative drift: require human approval for any new copy pattern, and throttle exploration if complaint rates rise\n- Bad baselines: if seasonality is strong, include a ghost holdout or use calendar matched controls from the prior period\n- Data entitlement mismatches: ensure only opted in users receive messages, and verify vendor responsibilities for consent logs\n\n## Executive checklist for scaling after the pilot\n\n- Evidence of lift: conversion rate and incremental revenue are both positive, with confidence intervals and a clean methodology\n- Cost discipline: CAC payback at or below 1.0 inside 60 days, and no drop in deliverability\n- Operative repeatability: the team can run the same flow on the next journey with less setup time\n- Governance: model decisions are logged, and there is a roll back plan that a non engineer can execute\n\nIf these conditions hold, expand scope to a second journey and add a second objective tier, for example lifetime value lift over 90 days.\n\n## Bottom line\n\nMoEngage’s new capital is a signal that AI decisioning is moving from slideware to standard operating procedure. You do not need to bet the quarter to prove impact. Pick one journey, turn on channel and messaging automation, and instrument the pilot so finance is comfortable with the math. In two weeks, you can see whether Merlin AI’s agents earn more than they cost, and if they do, you will have the confidence to scale.\n\nNext steps for your calendar\n\n- Today: choose your journey and confirm metrics and guardrails\n- Tomorrow: wire audiences and events, then preflight recommendations\n- Day 7: go live with treatment and matched control\n- Day 14: read out lift, incremental revenue, and CAC payback, then decide to scale or stop\n\nFor more hands-on templates, explore our guide to the WPP and Google AI pilot playbook and the TikTok Smart+ split test playbook.