Kenny Taber already had a strong Zappy Awards submission before he ever tried Zapier MCP.
In his probate support operations work, Kenny had built an email management system with the traditional Zapier approach: Zaps moved data between tools, ChatGPT categorized and summarized incoming email, Google Sheets held the backend, and Apps Script posted Slack digests so his team could see what still needed a response.
The workflow was useful enough to submit for Operations AI Builder of the Year. It also looked like many real automations after they start working: important, used often, and a little overgrown. Fourteen Google Sheets tabs. Formula chains. Slack permalink matching. Apps Script files. Manual context Kenny carried because he had built the system himself.
Then I asked him if he had tried MCP.
He had not. His workplace was cautious about AI, even as client requests for AI-enabled features had started to shift the conversation. Giving an AI tool broad access to internal systems would need a careful argument.
But Kenny was curious. His first reply was that his mind had gone to "thirteen different rabbit holes" of possible integrations. Then he wrote the sentence that made me want to meet with him: "Sometimes benefit outweighs fear."
The call that changed the shape of the workflow
Kenny and I met for 30 minutes.
I did not try to explain MCP from first principles. That would have missed the point. Kenny already had the thing most people need before MCP becomes useful: a specific workflow he understood deeply and wanted to improve.
So I showed him my own setup. I use Cursor as the place where I think, draft, and manage work. Zapier MCP is what lets that workspace act inside the apps I use every day. It can work with Zapier Tables. It can read and update Google Sheets. It can draft in Gmail. It can help with Slack-facing work. The useful part was simple: I was no longer opening every tab myself.
That clicked because Kenny had the same problem in a different workflow.
His email management system already used Zapier. MCP gave him a different way to work on it. Instead of opening Sheets, copying context into ChatGPT, switching back to scripts, checking Slack, then documenting what changed, he could work from one agent harness and let MCP reach the systems the workflow already depended on.
That is the simplest way to start with MCP: pick a workflow you already care about, connect the apps that workflow touches, and improve one part of it.
Kenny went and did exactly that.
The weekend rebuild
Over the weekend, Kenny moved the backend of his Zappy-submitted workflow from a 14-tab Google Sheets setup into one Zapier Table.
He imported 739 historical rows. He mapped the live Zaps around the new source of truth: one ingest path for incoming email, plus four Slack reaction workflows that update Workflow Status when the team reacts with emoji.
The old system had been built around sheets that acted like a database. The new system had an actual table. One row per email. One status field. One place to understand what was unresolved, handled by phone, marked as null, resolved, or sent.
The visible team workflow barely changed. Emails still showed up in Slack. The team still used reactions. Digests still told people what needed attention.
The backend changed completely.
The cache bridge
The Slack digest became the real test of the migration.
Kenny's team depends on a pinned Slack message and a daily blast that list unresolved emails. Apps Script had been reading from Google Sheets for free. If Kenny replaced that with scheduled Zaps that queried the table every few minutes, he would burn tasks just to keep the same list current.
Before he had fully explained that constraint, his agent pointed him toward a sheet cache bridge.
Zapier Tables would stay the source of truth. A cache sync Zap would write only the active unresolved records to a ZAPIER DIGEST CACHE tab when records changed. Apps Script would keep reading that cache tab for the pinned list and daily blast.
Kenny described it better than I could:
This is the MCP story I wish more people saw first: a specific constraint, inside a real workflow, solved with the systems already in use.
MCP helped Kenny reason across the actual system: Tables as source of truth, Zaps as the sync layer, Sheets as the cache, Apps Script as the free read path, Slack as the team surface.
The result was cleaner architecture without forcing the team to change how they worked.
What changed after the weekend
Kenny did more than move data.
He built a repo his agent can read. Architecture notes. Schema docs. Zap step mappings. Migration notes. A kenny-voice-style.md file so Cursor can draft email and Slack messages in his voice before he approves and sends them through Gmail MCP.
That approval step matters. Kenny is not handing over the business to an agent. He is giving it scoped actions inside specific systems, then approving the work before anything goes out.
After the migration, Kenny saw MCP as a bridge between two things already in motion: client requests for AI integration in specific software products and his own advocacy for broader AI use internally. Now he had a working example to point to.
Why this is the right way to start with MCP
Kenny's story is useful because it is specific. He started with a workflow he already knew how to defend: the same one he had submitted for a Zappy Award.
That is the path I would recommend to almost anyone trying MCP for the first time:
Pick one workflow. Choose the thing you already care about. The process that breaks quietly. The update you dread writing. The sheet you keep nursing along. The Slack thread that should become a task list.
Connect only the apps that workflow touches. Kenny needed Zapier Tables, Google Sheets, Gmail, and Slack. You might need Slack, Google Drive, Notion, Asana, Jira, or your CRM. Start with the minimum set.
Ship one useful improvement. A digest. A status update. A cleaner backend. A triage list. One thing that saves you from opening five tabs.
If you want to start today, use a template and keep the scope small: Catch up on Slack, extract action items from meeting notes, draft a status update from project notes, analyze spreadsheet data with AI, or turn Slack messages into tracked action items. See the template links in the markdown source.
For builders: copy the cache bridge
We generalized Kenny's digest pattern into an open-source Sheet Cache Bridge skill in Zapier's gtm-superpowers repo. Copy SKILL.md into your agent harness, follow SETUP.md, and adapt the cache columns to your table.
The pattern is simple:
Source of truth: Zapier Tables
Sync: Event-driven Zap when records change
Consumers: Apps Script or legacy tools read the sheet cache without per-read Zap tasks
Kenny proved the larger point by solving the smaller one. Starting with MCP really can be that simple when you already know the workflow you want to improve.
Install Zapier MCP · Browse MCP templates
