Anyone can build agents, but it takes a platform to run them
> **Source:** [Vercel Blog](https://vercel.com/blog/anyone-can-build-agents-but-it-takes-a-platform-to-run-them)
## Prototyping Is Democratized, but Production Deployment Isn’t
AI models have commoditized code and agent generation, making it possible for anyone to build sophisticated software in minutes. Claude can scaffold a fully‑functional agent before your morning coffee gets cold. But that same AI will happily architect a **$5,000‑per‑month** DevOps setup when the system could run efficiently at **$500‑per‑month**.
In a world where anyone can build internal tools and agents, the *build vs. buy* equation has fundamentally changed. Competitive advantage no longer comes from whether you can build—it comes from rapid iteration on AI that solves real problems for your business and, more importantly, reliably operating those systems at scale.
To do that, companies need an internal AI stack as robust as their external product infrastructure. That’s exactly what **Vercel’s agent orchestration platform** provides.Why the Economics Have Shifted
-
Historically, custom internal tools only made sense for large‑scale companies.
- Up‑front engineering investment was high.
- The real cost was long‑term operation with high SLAs and measurable ROI.
-
For everyone else, buying off‑the‑shelf software was the practical option.
-
AI has fundamentally changed this equation.
- Companies of any size can now create agents quickly.
- Customization delivers immediate ROI for specialized workflows.
Today the question isn’t “build vs. buy.” The answer is “build.”
Instead of separating internal systems and vendors, companies need a single platform that can handle the unique demands of agent workloads—and run them.
The Core Problem: Production Is Still Hard
- Shadow‑IT explosion – “Vibe coding” has created one of the largest shadow‑IT problems in history.
- Production expertise is scarce – Security, observability, reliability, and cost‑optimization remain rare skills, even as building becomes easier.
The ultimate challenge for agents isn’t building them; it’s the platform they run on.
A Real‑World Example: d0 – Our Internal Data Agent
Like OpenAI, we built our own internal data agent named d0 (open‑source template).
At its core, d0 is a text‑to‑SQL engine—nothing new. What made it a successful product was the platform underneath.
- Using Vercel’s built‑in primitives and deployment infrastructure, one person built d0 in a few weeks, spending only 20 % of their time.
- This was possible because Sandboxes, Fluid Compute, and AI Gateway automatically handled the operational complexity that would normally take months of engineering effort to scaffold and secure.
Impact
- d0 has completely democratized data access that was previously limited to professional analysts.
- Engineers, marketers, and executives can ask natural‑language questions and get immediate, accurate answers from our data warehouse.
How Vercel Powers Agent Workloads
| Primitive | What It Does | Why It Matters |
|---|---|---|
| Sandboxes | Gives agents a secure, isolated environment for executing sensitive autonomous actions. | Protects core systems; contains damage from untested code or prompt‑injection attacks within isolated Linux VMs. |
| Fluid Compute | Automatically scales compute up and down based on demand; you only pay for actual compute time. | Keeps costs low and predictable, especially for data‑heavy workloads (files, images, video). |
| AI Gateway | Provides unified access to hundreds of models with built‑in budget control, usage monitoring, and load balancing across providers. | Avoids vendor lock‑in, routes simple requests to cheap models, complex analysis to powerful ones, and fails over automatically. |
| Workflows (Durable Orchestration) | Enables multi‑step operations with retry logic and error handling at every step. | Guarantees reliability for critical business processes; failures don’t require manual intervention. |
| Observability | Shows exactly what agents are doing—prompts, model responses, decision paths, token usage, etc. | Essential for debugging, performance tuning, and cost control. |
Both internal and customer‑facing products run on the same primitives.
This uniformity simplifies operations and security across the organization.
The Future Is Agent‑Centric
Every enterprise will soon build its own version of:
- d0 – internal data‑access agent
- Code‑review agent
- Customer‑support routing agent
- Hundreds of other specialized tools
The success of these agents depends on the platform that runs them. Companies that invest in their internal AI stack now will not only move faster—they’ll experience far higher ROI as their advantages compound over time.
Key Takeaways
- Build vs. buy ROI has fundamentally changed.
- The platform is the product: how our data agent runs on Vercel.
- Vercel is the platform for agents. Build your agents; Vercel will run them.
Vercel’s AI Agents
Overview
- Lead‑qualification agent – enables one SDR to do the work of 10 agents.
- Customer‑facing financial‑impact calculator – built by Stripe on a flight!
“What was our Enterprise ARR last quarter?”
The agent (d0) receives the message, determines the appropriate data‑access level based on the user’s permissions, and starts the workflow. A user can ask a question directly in Slack.
Semantic Layer
The semantic layer is a file‑system of five YAML‑based configuration layers that describe:
- Our data warehouse
- Our metrics
- Our products
- Our operations
The agent explores this semantic layer to understand the context of each request.
Core Capabilities
- Streaming responses, tool use, and structured outputs work out‑of‑the‑box.
- No custom LLM plumbing was built; we use the same abstractions any Vercel developer can use.
AI SDK
- Handles all model calls.
- If a step fails (e.g., Snowflake timeout, model hiccup), Vercel Workflows automatically retries and recovers state.
Durable Orchestration
- File exploration, SQL generation, and query execution all run in a secure Vercel Sandbox.
- Runaway operations cannot escape the sandbox.
- The agent can execute arbitrary Python for advanced analysis.
Isolation & Routing
- AI Gateway routes simple requests to fast models and complex analysis to Claude Opus, all within a single code base.
- Multiple models are used to balance cost and accuracy.
Output Delivery
- Formatted results (often with a chart or Google Sheet link) are sent back to Slack via the AI SDK Chatbot primitive.
- The answer appears directly in the Slack conversation.
Example Agents
| Agent | Primary Use |
|---|---|
| Lead‑qualification agent | Automates SDR outreach |
| d0 (analytics agent) | Answers data‑driven questions |
| Customer‑support agent | Reduces tickets by one‑third |
| Abuse‑detection agent | Flags risky content |
| Content agent | Turns Slack threads into draft blog posts |
| v0 (code‑generation agent) | Generates code snippets |
| Vercel Agent | Reviews PRs, analyzes incidents, recommends actions |
Bottom Line
Every company needs an internal AI stack—and Vercel provides the tools, infrastructure, and agents to make it happen.