Getting your tools to talk to each other (an MCP guide for social intelligence practitioners)

Somewhere in your organisation, a perfectly good question is waiting for someone to have a free morning. Why are customers cross about the latest product release? Is the campaign answering what shoppers ask? Is interest in alcohol-free beer here to stay? Does that creator's audience still match the brief? The answers usually exist already. They're spread across 4 platforms and a folder of exports.

MCP (Model Context Protocol) is the standard that lets an AI assistant such as Claude or ChatGPT connect to the tools you already pay for, fetch what it needs from each, and sometimes act inside them: save a draft post, add rows to a sheet, build the first 5 slides of a deck, post a summary in Slack. At least 150 vendors in this year's SITech directory now offer an MCP server, across all 9 pillars, so there's a good chance one is already in your stack, whether or not anyone has switched it on.

So I've written a 4-part guide. It starts from zero, keeps the detail that matters, and ends each part with a short section for those of you already running a few servers.

• Part 1: What MCP means for social intelligence. What it is in plain English, followed through one product-launch question from a listening platform to a Google Slides deck.
• Part 2: Workflows across the 9 pillars. One worked example per pillar, from a hotel group's reviews to a city's recycling schedule, plus the automation tools (Zapier, n8n, IFTTT) that can run them on a schedule.
• Part 3: 6 questions before you connect anything. What to ask a vendor, what to check before an assistant posts anything for you, and the licensing question that's easy to miss.
• Part 4: Where this takes social intelligence. What changes when colleagues, automations and agents can all reach your research, and why the practitioner's judgement matters more as a result.

Start with Part 1, and bring your questions along - read here.

‍