Connect Your Database to Google Sheets

Welcome back. AI is shaking up operational efficiency. Here's how.

Developer and open-source toolmaker Simon Willison published a practical walkthrough showing three distinct methods for pulling live SQL query results from Datasette — a tool that wraps SQLite databases in a browsable, queryable web interface — directly into Google Sheets. The approaches range from a simple built-in Sheets formula using IMPORTDATA(), to a custom "named function" that packages the call cleanly, to a Google Apps Script workaround for situations where you need to pass an authentication token in an HTTP header (something the native formula can't do). A working example sheet demonstrates all three patterns side by side.

This fits into a quiet but meaningful trend in operational tooling: the gap between where business data lives and where business decisions get made is shrinking — not because everyone moved to a unified data warehouse, but because lightweight connectors keep appearing that let analysts and ops teams skip the formal integration entirely. Datasette in particular has built a following among teams that want to publish internal data sets in a queryable format without standing up heavyweight infrastructure. Connecting it to Sheets means your live database can feed dashboards, reports, and decision trackers that non-technical teammates already know how to use.

for a 10-to-200-person company, the practical signal here is about reducing the copy-paste layer. Analysis suggests that if your ops data sits in any SQL-queryable source, there's likely a path to pull it into Sheets on demand rather than exporting CSVs manually. The authentication pattern documented by Simon Willison matters specifically for internal tools where your data shouldn't be publicly accessible. So what: the friction between "data exists somewhere" and "data is visible in the tool my team actually opens" is now mostly a configuration problem, not an engineering one.

Simon Willison documented three methods — IMPORTDATA() , a named function wrapper, and Google Apps Script — for querying a Datasette instance directly from Google Sheets, including handling authenticated requests.

Lightweight connectors like this compress the distance between operational databases and the spreadsheet layer where most non-technical decision-making still happens — no formal integration project required.

Teams with data in any SQL-accessible source can likely eliminate manual export workflows; the authenticated Apps Script pattern extends this to internal, access-controlled data sets.

OpenAI has rolled out a prompting fundamentals course through its academy, focusing on crafting clear, effective prompts to get better responses from ChatGPT. On the surface, it's basic training. But from an ops perspective, it signals where the real friction lies. After years of model improvements, the bottleneck isn't the AI. It's the instructions going into it. Most automation teams know this: a workflow's success often hinges on a clear prompt, not a model limitation. OpenAI's structured guidance isn't just about education. It's an admission that prompt quality is the variable most operators haven't nailed down yet.

OpenAI's Academy material on custom GPTs frames them as purpose-built assistants — configured once, deployed repeatedly, with consistent outputs baked in by design. The practical read for operators: a custom GPT is closer to a trained role than a chatbot. You define the scope, the tone, the constraints, and the workflow context up front. What comes back isn't a general-purpose response — it's output shaped for a specific job. That distinction matters when you're thinking about where AI actually fits in your team structure versus where you're just prompting ad hoc and hoping for consistency.

OpenAI has launched a dedicated healthcare education track through OpenAI Academy, aimed at clinicians using ChatGPT for diagnosis support, clinical documentation, and patient care workflows. The positioning is deliberate: HIPAA-compliant tooling, not a general-purpose chatbot pointed at a regulated industry. That distinction matters. Healthcare has been one of the sectors most cautious about AI adoption precisely because the liability exposure on a generic tool is real. By building a compliance-first offering and pairing it with structured training, OpenAI is trying to reduce the adoption friction that has kept most health systems at arm's length. Whether the clinical use cases hold up under real-world pressure is still an open question.

Source: Simon Willison · simonwillison.net · Apr 20, 2026

Datasette — an open-source tool for publishing and querying SQLite databases via a web interface — can pipe query results directly into Google Sheets using nothing more exotic than a built-in spreadsheet formula. Three methods exist: a native IMPORTDATA() call, a named function wrapping it, or a Google Apps Script for authenticated requests where an API token needs to travel in an HTTP header. What nobody's telling you: the bottleneck in most ops reporting isn't the database — it's the ten manual export steps between the database and the spreadsheet someone actually reads. Simon Willison claims that this pattern cuts the manual export steps to zero. Simon Willison suggests that your data stays live, your sheet stays current, and nobody has to remember to run the export on Monday morning. <div style="padding: 14px 18px; border-left: 3px solid #1A5C5A; background: rgba(26, 92, 90, 0.06); font-size: 14.5px; color: #1A1A1A; line-height: 1

AI finds the signal. Human judgment sharpens it. Same workflow we'd build for your team.

🚀

AI Workflow Platform

Build, deploy, and scale AI agents fast and secure. Cloudflare's integration with OpenAI's LLMs, including GPT-5.4, makes it happen. That's it.

💰

Financial AI Tools

OpenAI's new tools for financial services. They're designed to help you deploy AI securely and effectively. No fluff, just results.

🎤

Transportation Investment

Uber's going big. $10 billion big. They're buying autonomous vehicles and pouring cash into tech development. Good luck keeping up.

Visual Automation

Make.com is a no-code automation platform that lets you connect different apps and automate workflows using a visual drag-and-drop interface. You can build “scenarios” that handle tasks automatically—like moving data, sending emails, or updating tools—without writing any code. It’s especially powerful for creating complex, multi-step automations across multiple apps.

Your support queue doesn't need a bigger team — it needs a first pass that never sleeps. Most tickets fall into three buckets: urgent, routine, and noise. Right now, a human is reading every single one before that sort happens. That's the problem.

Result: Your team reads only the tickets that need human judgment. First-response time drops and queue depth becomes manageable without adding a single seat.

AI finds the signal. Human judgment sharpens it. Same workflow we'd build for your team.