MoEngage serves more than 1,350 consumer brands, and its customer Glance achieved a 38% increase in push-notification click-through rate through Merlin AI. This shows that agents—AI assistants capable of invoking software tools to complete tasks—are beginning to deliver measurable results rather than merely generate copy.

What this is

MoEngage has converted its existing marketing interfaces into MCP tools—a protocol that lets AI connect to external data and tools in a standardized way—so compatible assistants such as Claude and ChatGPT can access a brand's user segments, historical campaigns and content libraries, while keeping data isolated by tenant and permissions constrained.

Previously, marketers had to move between analytics dashboards, content libraries and campaign editors. Now, they can state their objectives, and Merlin AI will analyze the data, draft and refine campaigns, after which the system selects what to send based on customer characteristics. The infrastructure runs on Amazon Bedrock and AgentCore, while reasoning and content generation use the Claude Sonnet model family.

Industry view

MoEngage's case shows that competition in enterprise software is expanding from whether a product has a complete feature set to whether it can directly complete tasks. With natural language becoming the primary interface, menus, APIs and agents form a three-layer structure. Platforms that already control substantial business data and execution capabilities are better positioned to turn agents into their primary entry point.

The risks are equally clear. Once an MCP tool can read data and execute campaigns, faulty segmentation, unauthorized calls or prompt injection—using content to trick an agent into taking inappropriate action—can turn a draft into a live marketing action. The officially disclosed 38% improvement comes from a single customer case and cannot be treated as a universal result. Differences in data quality, sending strategy and measurement methodology will still affect outcomes across brands.

Impact on regular people

For enterprise IT: Existing systems do not need to be rebuilt from the ground up before APIs can be exposed incrementally as agent tools. But tenant isolation, permission controls, audit trails and human approval remain prerequisites for deployment.

For professionals: Roles in marketing, product management and related fields will increasingly shift from moving information between tools to defining objectives and validating results. Process design, understanding metrics and the ability to review output will become more important.

For the consumer market: Brands may be able to deliver more personalized outreach faster, but consumers will also receive more automatically generated offers and reminders. Opt-out mechanisms, frequency controls and transparency about the source of messages need to strengthen in parallel.