From Work Slop to Agentic AI: Making Sense of the Latest Marketing AI Tools

In this episode of Artificially Intelligent Marketing, Martin Broadhurst and Paul Avery reunite to explore how AI has transformed marketing over the past 18 months. They cover reasoning models, agentic automation, Microsoft Copilot’s evolution, open vs closed-source AI, and the rise of AI-powered hardware—sharing real-world insights and examples from their work.Major Evolutions in AI for MarketersReflection on the rapid progress of AI tools and modelsOverview of major shifts since the last episodeHow marketers are adapting to new AI capabilitiesAI Reasoning ModelsDifference between chain-of-thought prompting and modern reasoning modelsImprovements in accuracy and reduced hallucinationsTrade-offs between speed and reasoning depthGroq CEO’s insights on the value of ultra-fast inferenceAI Tools Adoption and Platform MaturityMicrosoft Copilot’s leap from basic to highly capableKey tools: Researcher agent, Analyst tool, and Copilot StudioIntegration across Microsoft 365 (SharePoint, OneDrive, Teams)Comparisons with Google and OpenAI’s platformsOngoing confusion over pricing and valueModel Selection: The “Model Roundabout”Recent advances in GPT, Claude, Gemini, and open-source modelsBalancing reasoning and instant modesCommon use cases: coding, summarisation, planning, and copywritingQuirks such as GPT-5’s writing tone and output styleTips for reducing hallucinations and improving reliabilityOpen vs Closed Source AI DebateRise and stall of open models like DeepSeek and Llama 4Meta’s shift from open development to proprietary AGIOpen source’s future in experimentation rather than frontier innovationMarket consolidation, privacy, and trust concernsAI-Integrated Hardware and the Attention EconomyGrowth of wearable AI, e.g. Meta’s Ray-Ban smart glassesPrivacy and social implications of constant recordingAdoption driven by convenience and content habitsMeta’s competing aims: productivity vs attention monetisationAgentic Progress: AI Agents and Automation“Agentic AI” explained: systems acting autonomously to complete goalsFrom document retrieval to full workflow automationTools like Make.com, Zapier, and N8N enabling marketersClaude Code as an advanced example of self-directed agentsUse cases: automated slide decks, proposals, and scheduled reportingMCP (Model Context Protocol) ConnectorsOverview of MCP for connecting LLMs to CRMs and cloud toolsMartin’s experience linking Claude to HubSpot and Google WorkspaceExamples of AI updating pipelines and deal notes automaticallyBenefits balanced against setup complexityCurrent State of AI for MarketersHonest look at AI-generated content and “work slop”AI as a speed and productivity enhancer, not a replacement for expertsAdvances in visual and video generation:Faster, more consistent imagery (Midjourney, DALL·E 3, Nano Banana)Real-world use in proposals, events, and social mediaEmerging video models (Veo 3, Sora 2, Kling) offering realism and soundReflection on low-quality AI output and the lasting importance of trusted brands