ibl.ai Agentic AI Blog

Insights on building and deploying agentic AI systems. Our blog covers AI agent architectures, LLM infrastructure, MCP servers, enterprise deployment strategies, and real-world implementation guides. Whether you are a developer building AI agents, a CTO evaluating agentic platforms, or a technical leader driving AI adoption, you will find practical guidance here.

Topics We Cover

Featured Research and Reports

We analyze key research from leading institutions and labs including Google DeepMind, Anthropic, OpenAI, Meta AI, McKinsey, and the World Economic Forum. Our content includes detailed analysis of reports on AI agents, foundation models, and enterprise AI strategy.

For Technical Leaders

CTOs, engineering leads, and AI architects turn to our blog for guidance on agent orchestration, model evaluation, infrastructure planning, and building production-ready AI systems. We provide frameworks for responsible AI deployment that balance capability with safety and reliability.

Back to Blog

Cloudflare: AI Agent Requests Up 1,700%. Who Routes Them?

Miguel AmigotNovember 19, 2026
Premium

Cloudflare reports daily AI agent requests on its network grew more than 1,700% in a year and that more than half of 2026 Internet traffic was not human. When software sends most requests, the layer that decides which agent answers each one is business logic an organization should own.

The Short Answer

Cloudflare reports that daily requests from AI agents on its network grew by more than 1,700% in a year, and that more than half of Internet traffic in 2026 was not human. When software sends most requests, the layer deciding which agent answers each one is business logic. On ibl.ai you own all the code and the data, including that routing layer, on any model.

The figures come from Cloudflare's September 30, 2026 post "The Internet has a second audience" and its October 20, 2026 press release announcing more than 10 million developers on its platform, up from 7.4 million the prior quarter.

They are Cloudflare's measurements on Cloudflare's network. More than 20% of the web sits behind that network, by Cloudflare's own count, so they are a large sample, not a census of the whole Internet.

What did Cloudflare actually report about AI agent traffic?

Cloudflare's September 30 post makes four claims that matter for anyone building or buying AI agents. Each is a number Cloudflare published, and each measures something different, so they should not be added together.

Cloudflare figure Value What it measures
Daily AI agent requests +1,700% Growth over the past year in requests from AI agents
Non-human share of traffic >50% First year more than half of Internet traffic was not human
Average HTTP requests per second 63M β†’ 115M End of 2024 to September 2026, all traffic, peaks above 150M
Crawler requests for AI training 22% β†’ 52% Spring 2025 to June 2026, by the crawlers' stated purpose
Human traffic in heavily crawled categories down up to 40% Retail, software, IT services and financial services, in under a year

The 1,700% is growth in daily AI agent requests. The 63 million to 115 million is total HTTP traffic. The first is a rate of change in one slice; the second is the whole network, so one does not explain the other.

Why does AI agent traffic matter to organizations that are not websites?

Cloudflare's post is written for site owners: a second audience that reads pages, rarely clicks, and may or may not pay. That reading is correct, and it is only half of the shift.

The same kind of agent that reads websites can also file support tickets, query internal systems, book appointments and ask for refunds on a person's behalf.

Any organization that runs agents of its own now receives requests from people and from software, at volumes no help desk was sized for.

Cloudflare's own framing points this way. Its post argues that agents can be customers, and that a site turning an agent away may be turning away the person behind it.

The question for an organization is not whether to answer agents, but which of its own agents answers each request.

What is an AI agent routing layer, and why is it separate from the model?

A routing layer is the step that reads a request, decides what it is about, and hands it to the agent built for that job. It sits in front of the agents, and its decision is usually a small, fast model call rather than a long generation.

It became its own layer in 2026. Between September 15 and October 6, six vendors shipped dedicated decision models for exactly this job, three of them with downloadable Apache 2.0 weights, which we covered in Six Decision Models in Three Weeks. Three You Can Own.

The reason is arithmetic. If request volume rises an order of magnitude, sending every routing call to a frontier model multiplies the bill by the same factor. A routing decision does not need prose; it needs a label and a confidence score.

How does ibl.ai route each request to the right agent?

ibl.ai agent workflows include a Classifier step, shipped in the November 6 to 13, 2026 releases. It sends each message to a decision model, reads back an answer, and follows the path wired to that answer.

There are three answer types. Pick a Category takes 2 to 20 categories, each with its own path to its own agent. Yes or No covers binary decisions. Rate on a Scale takes 2 to 10 levels for triage-style decisions.

Every type has a Fallback path. Below a confidence threshold, 0.6 by default, the request goes to a fallback agent that asks a clarifying question instead of guessing. In the release demo, "I was charged twice for my plan this month" was routed to Billing at 0.94 confidence.

A team can test the decision on a sample message before publishing, and the same test is exposed over the platform's REST API, so real messages can be replayed against a routing rule to see which would fall back before it goes live.

Who should own the routing rules when most requests come from software?

Routing rules decide which customer reaches which agent, with which data, at what confidence. That is business logic in the same sense as a pricing table or an escalation policy, and it changes every time a product or a team changes.

On ibl.ai the workflow editor ships as source in the SDK, the workflow definitions can be read back over the REST API, and the runtime comes under ibl.ai's Full Code License to run on our hosted cloud or on your own servers. You can build it into your own app with Agentic Base.

The decision model comes from the organization's model registry and each agent runs the model chosen for it, so neither the router nor the agents behind it is tied to one vendor's models. When a cheaper or better decision model ships, it is a setting, not a migration.

Why does per-seat AI pricing break when agents send the requests?

Per-seat pricing counts people. Cloudflare's numbers describe a world where requests from AI agents grew more than 1,700% in a year while human traffic fell in some categories.

An agent acting for a customer has no seat, and a router answering every inbound request cannot be licensed to a named few.

A seat-priced platform scales its bill with headcount whether the requests come from those people or not. A usage-based or self-hosted platform scales with the work actually done, which is the only shape that tracks a 1,700% rise in agent requests.

ibl.ai has no per-seat pricing: you pay for the tokens consumed or own the stack and pay for the compute. More than 1.6 million users across 400+ organizations run it this way, including NVIDIA, MIT and Syracuse University, through Agentic OS.

What should a team do about rising AI agent traffic this quarter?

Start with three questions. Which inbound requests already come from agents rather than people? Which of your agents should answer each kind? And what happens when the routing decision is unsure?

The first is a measurement question your edge provider can help with. The second and third are design questions, and they are cheaper to answer with a routing step you can test on real messages than with one large agent that tries to cover everything.

ibl.ai is family-owned and operated from New York, NY. If you want to see a Classifier wired to two or three of your own agents and tried on messages your team actually receives, talk to the ibl.ai team.

Why does owning the AI stack matter?

ibl.ai is the agentic AI platform where you own all the code and the data. You self-host the entire stack inside your own perimeter, run it model-agnostic across any LLM and switch anytime, and pay by usage with no per-seat pricing β€” so you can deploy anywhere: your cloud, on-premise, GovCloud, or fully air-gapped.

  • You own all the code and the data

    Full source code under a perpetual license, running on your infrastructure. Not API access to someone else's platform β€” the stack itself is yours.

  • Model-agnostic

    Run any LLM β€” Claude, GPT, Gemini, Llama, Command, or your own fine-tune β€” and switch providers without rewriting the platform.

  • No per-seat pricing

    Usage-based billing against a budget cap you set. Cost tracks what your organization actually uses, not how many people you employ.

  • Deploy anywhere

    Your cloud, your VPC, on-premise, GovCloud, or a fully air-gapped network with no outbound connectivity.

1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.

ibl.ai is family-owned and operated from New York, NY β€” a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on.

Related Articles

Six Decision Models in Three Weeks. Three You Can Own.

Six decision models shipped between 15 September and 6 October 2026: TypeSafe's Jev, Convai's Laya, Fastino's GLiNER2.5-Decide, Nace AI's Drex 1.5, vLLM Semantic Router's Decision 2.0 family, and OpenAI's Decisions API. Three carry Apache 2.0 weights, one is commercially fenced, two are hosted. The licence is the real difference β€” and the widely-quoted 33 ms belongs to Laya's 322M multilingual checkpoint on a Tesla T4, not to a flagship.

ibl.ai EngineeringSeptember 28, 2026

SaaS Fragmentation Is the Hidden Cost of Enterprise AI

Enterprises run six or seven per-seat tools that each hold a partial copy of the same customer. That fragmentation, not model capability, is what stalls AI deployment β€” and it carries a per-seat bill that grows with headcount. This post itemizes the fragmentation tax and shows the MCP-based orchestration layer that reads across every system instead of adding another one.

Miguel AmigotJuly 31, 2026

AI Agents Outrun Static Credentials: Gartner and the FBI

In four weeks Gartner named AI agent identity a 2027 trend and predicted 25% of enterprise breaches will trace to AI agent abuse by 2028, HashiCorp shipped agent identity tooling, and the FBI described an agentic attack framework reaching privileged access in under three hours. The common thread is the static credential.

Miguel AmigotNovember 19, 2026

Anne Boden's Lockwood: A Bank for AI Agents, Not Yet Licensed

Starling founder Anne Boden is building Lockwood, a planned bank whose current account lets customers' AI agents pay and shop within limits they set. It is not licensed yet: it targets an e-money licence in 2027 and launch in 2028. Here is what was announced and what banks should take from it.

Miguel AmigotNovember 19, 2026

See the ibl.ai AI Operating System in Action

Discover how leading universities and organizations are transforming education with the ibl.ai AI Operating System. Explore real-world implementations from Harvard, MIT, Stanford, and users from 400+ institutions worldwide.

View Case Studies
Work with our team

Pilots, deployment, and full ownership

Every engagement is scoped to the work: you integrate ibl.ai with your own data, deploy it on your own infrastructure, and the engineering hours scale with the scope β€” so the price tracks the scope, not your headcount. Most customers keep us on afterwards for platform updates, new agents and integrations, and ongoing services.

Start here

Pilot

from $15K

fixed scope Β· fixed timeline

A time-boxed proof of value on your real data β€” not a slide deck.

Best for: Teams that want to see ibl.ai working before committing.

  • Deployed on your infrastructure or our cloud
  • 1–2 production agents wired to a slice of your data
  • One integration (LMS / SIS / SSO / data source)
  • Weekly working sessions with our engineers
  • Pilot fee credits toward a full engagement
Scope a pilot
Most common

Integration & Deployment

$25K – $80K

one-time fee Β· scoped to the work

Full deployment integrated with your data and systems. Engineering hours scale with scope.

Best for: Organizations rolling ibl.ai out across a department, campus, or business unit.

  • Platform deployed in your VPC, on-prem, or air-gapped
  • Integrated with your data + identity (SSO / SAML)
  • Multiple custom agents built to your workflows
  • Engineering hours proportional to scope
  • You own the data Β· run any LLM you choose
Plan a deployment
Full ownership

Codebase Transfer + Custom AI Engineering

Custom quote

perpetual license Β· you own the stack

We transfer the full source code. You own and self-host the entire platform β€” outright.

Best for: Organizations and enterprises that benefit from perpetual ownership and sovereignty.

  • Complete source-code transfer + perpetual license
  • Dedicated AI engineering team on your roadmap
  • Custom agents, models, and integrations to spec
  • Air-gapped capable Β· zero vendor lock-in
  • Family-owned, New York–based long-term partner
Talk about ownership
You own the code and data Run any LLM β€” Claude, GPT, Gemini, Llama Family-owned & operated from New York, NY