|
AI Builders Digest
Friday, July 17, 2026
|
|
Two days in a row, the same theme keeps surfacing: enterprise AI adoption is moving slower than the pitch decks suggest, and the bottleneck isn't the models. It's everything around them. Box CEO Aaron Levie's dinner with IT leaders last night puts real words to what the Vercel analytics story is quietly demonstrating: the companies actually building useful AI products are the ones treating data plumbing as a first-class problem, not an afterthought.
|
|
---
|
|
01
|
What enterprise IT leaders actually worry about with AI agents
|
|
|
Box CEO Aaron Levie hosted a dinner with IT leaders from large enterprises and shared his notes. The top concern wasn't cost or model quality. It was change management: most existing workflows weren't designed to work with agents, so companies face simultaneous upgrades across technology, data infrastructure, and human processes. Getting data into a format agents can actually use, both structured (databases, spreadsheets) and unstructured (documents, emails), dominated the conversation.
|
Why it matters: If your company is six months into an "agent strategy" and still debating which model to use, you're working on the wrong problem. The IT leaders in that room have already figured out the model question. They're stuck on whether their data is even in shape for an agent to touch. That's a 12-to-24-month project, not a pilot.
|
|
Source →
|
|
---
|
|
02
|
Vercel's analytics API is quietly building an agent-readable data layer
|
|
|
Vercel CEO Guillermo Rauch highlighted some emerging uses of Vercel's Web Analytics API: asking agents to correlate site visitors and custom events like purchases and checkouts with deployment timelines and performance changes, then plotting that alongside Stripe and Resend data in custom dashboards.
|
Why it matters: This connects directly to what Levie's dinner guests were wrestling with. The companies that will get agents working in production first are the ones whose operational data is already API-accessible and agent-readable. Vercel is quietly positioning its platform as exactly that kind of data layer, starting with web traffic and transaction data that product teams actually care about.
|
|
Source →
|
|
---
|
|
03
|
Google's Gemini Spark gets faster and learns to edit your Docs
|
|
|
Google Labs VP Josh Woodward announced that Gemini Spark is expanding to more Ultra subscribers globally, with four updates: it can now open and edit Google Docs, read comments in Sheets and Slides, run more than 50% faster, and process multiple sources in parallel. Pro subscribers are next.
|
Why it matters: The Docs editing capability matters most here. An agent that can only read your documents is a research tool. One that can edit them based on feedback is starting to look like a junior colleague. Google is clearly pushing Spark toward the kind of agentic work that competes with the standalone AI assistants your team is already paying for separately.
|
|
Source →
|
|
---
|
|
|
|
**Coding agents as creative tools, not just productivity hacks** — Zara Zhang, who never learned traditional programming, described using coding agents as an act of creativity and self-expression, calling GitHub "basically my Substack." It's a useful reminder that the people getting the most from these tools aren't always the ones with CS degrees.
|
|
Source →
|
|
**Swyx's post doesn't have enough context to cover** — The tweet in question is a single line with a linked image and no explanation of what was tweeted into existence. Nothing to report here.
|
|
Source →
|
|
Follow builders, not influencers. A daily digest of what matters in AI.
Read online ·
Archive
|