I spent the first five months of 2026 analysing where to put our marketing automation budget. The mental model that helped most was realising the tools have splintered into two distinct categories.
Category 1 is the established marketing automation platforms that slapped AI on top - HubSpot, Klaviyo, Customer.io, ActiveCampaign, Marketo. They own your email, forms, contact database, lead scoring. The AI shows up as subject line generation, send-time optimisation, copy suggestions, and sometimes an LLM step inside a journey (Customer.io is furthest along there). The AI is genuinely useful, but it's wrapped around a 2018-era data model. You're not getting AI agents that act across your stack - you're getting smarter email send mechanics.
Category 2 is the AI-native workflow tools that marketing teams run alongside their primary platform. This category barely existed five years ago. Typically it's a visual workflow builder with AI as first-class primitives (extraction, summarisation, drafting, classification) plus connections to your CRM. Designed for tasks like: read incoming demo requests, score them with AI, route the qualified ones to a SDR, draft a personalised first-touch email for human review.
The category 2 tools we evaluated: Lindy for sales/SDR-heavy positioning (decent if your bottleneck is outbound and BDR capacity, less obvious as a general marketing tool), and Relay - a plain-English workflow builder with human approval steps built in. AI is native, not bolted on, which gives cleaner UX for AI-heavy flows. Smaller catalogue than Zapier but covers the marketing stack I actually use.
Our internal deployment for 2026: for email, forms, contact database, journeys, lead scoring - a category 1 platform. For AI-heavy work that spans the stack (demo request triage, AI-drafted outreach with human approval, content workflows, automated lead enrichment) - one of the category 2 tools. We're resisting the urge to make one thing do both, these marketing automation platforms aren't catching up on agent-style work for at least another 18 months.
Answering the question 'where does the AI work belong in my stack?' helped us evaluate faster with less cost. Hope that's useful. What else is working for other marketing ops folks here?
