Agents for the whole company, an architecture lesson from the checkout of one of the biggest e-commerce platforms and strategies to keep the AI bill under control.
Cloudflare open-sources Cloudflare OS, which gives every employee an agent with company context
Cloudflare open-sourced a new version of Cloudflare OS, which gives every person an agent and a workspace connected to the company's context and systems. Internally, thousands of employees across every function, many outside engineering, already use it to create documents and slides, automate tasks and build small data apps. A shared library of skills records terminology, procedures and the best way to do recurring work, so an improvement made by one person serves everyone.
Taking agents beyond engineering requires connecting AI to internal systems and recording company knowledge in a reusable way. Any organization can deploy and adapt the tool, connecting it to its own systems.
How Dexpert helps: Digital TransformationSources
- The Cloudflare Blog: Cloudflare OS: an open platform for agents, apps, and work
Shopify swaps Redis for MySQL in checkout inventory reservations and scales at peak
Shopify reserves inventory during payment so two buyers cannot take the last unit, and for years this ran on Redis. While unifying its database strategy, the team rebuilt reservations on MySQL with one row per inventory unit and MySQL 8's SKIP LOCKED feature, after earlier attempts failed due to contention on a single quantity row. The system hit its targets at the 2025 peak, when the platform recorded $5.1 million in sales per minute on Black Friday.
Good data design lets the relational database a company already has handle high-concurrency scenarios, without another technology to operate. E-commerce and booking systems can apply the same pattern.
How Dexpert helps: Backend & APIsSources
- Shopify Engineering: We replaced Redis with MySQL for inventory reservations—and it scaled
Databricks gathers techniques from companies such as Stripe and Uber to control AI coding costs
Databricks says agentic coding has improved every velocity metric it tracks, but nearly every company adopting these tools at scale hits costs that grow too fast. Based on its own experience and conversations with Stripe, Coinbase, Uber and Ramp, the post gathers the main levers: moving to open and cheaper models, routing each request to the right model and giving developers visibility, alerts and budgets. The company released part of the infrastructure it uses for this.
You can give broad access to AI tools while keeping the cost per person within a predictable limit. Before scaling, set up consumption tracking by team and routing rules across models.
How Dexpert helps: Artificial IntelligenceSources
- Databricks: Managing AI Coding Costs at Scale
Want to put one of these to work in your company?
Dexpert builds custom systems, apps, integrations and AI solutions. Tell us what you need and get an estimate, or take the free digital maturity assessment.