Boomi announced in Sydney that the AI agents running through its platform now have a watchdog, no matter who built the agent or which model powers it. Companies have spent the past two years connecting AI agents to real business systems, letting them update customer records, approve refunds, and move money between accounts. Most of those agents ran with no one checking what they did once they started acting. Boomi's new AI Gateway sits between every agent and the systems it touches, and it can stop an agent mid-action if that action looks wrong.
The AI Gateway comes from Lunar.dev, the company Boomi acquired earlier this year, now built into what Boomi calls the Boomi Agent Control Plane. Boomi already tracked which agents customers had built and what those agents were supposed to do, through Agentstudio and Connect. The AI Gateway is the part that watches what those agents do in the moment, and can step in.
An Agent Can Now Be Stopped Mid-Action
Every AI agent connects to a company's software through the Model Context Protocol, a standard that lets an agent use a business tool the same way a person would click a button in it. Boomi has already built more than 1,000 of these connections through Boomi Connect, tied to a company's existing login system rather than loose API keys anyone could copy.
What is new is what happens once an agent tries to act. The AI Gateway can cap how many actions an agent takes, cap how much it spends on AI processing, and hold an action for a person to approve before it goes through. Every one of those events gets logged in real time, as it happens.
Most companies still cannot answer a basic question: what an agent did to a customer account last week, and who approved it. Cloud providers and AI model companies each keep their own partial record. Nobody has owned the full picture of an agent's action from start to finish, until this gateway sits in the middle of it.
The Trust Gap Is Already Costing Money
Nearly every finance team is now watching AI spending specifically. In the FinOps Foundation's 2026 survey of 1,192 practitioners representing more than $83 billion in annual cloud spend, 98% said they actively manage AI costs, up from 31% two years earlier (FinOps Foundation, 2026). That happened faster than most finance teams were ready for, and it means AI budgets are getting real scrutiny now, not later.
Trust is the harder problem to solve. A survey Boomi commissioned found 86% of companies had moved their AI agents past the testing phase into real use. Only 34% said they trust what those agents are doing. Companies that pushed agents into production before building that trust paid for it: an average of $2.1 million in extra cost from mistakes, rework, and cleanup (Forrester Consulting, 2026).
Boomi built the AI Gateway to close that specific gap. Clear spending caps and an audit trail are supposed to be the reason a company trusts its agents with more work, faster than a company still guessing what its agents are doing.
Bring Your Own Model Starts With Anthropic, OpenAI, and Gemini
Boomi's gateway is not tied to one AI provider. Companies can run agents built on models from Anthropic, OpenAI, or Google's Gemini, and route that traffic through the same gateway from launch. Companies can also run smaller, custom-built models directly inside Boomi's own infrastructure, whether that's public cloud, a company's own private cloud, or on-premises.
That is a real bet on Boomi's part. Boomi doesn't build AI models, so it has no reason to steer customers toward one. The AI Gateway routes to whichever model a company already uses, and Boomi is betting companies will pay for the control point itself, independent of which model runs behind it.
There is a financial reason behind that bet too. Running models inside a company's own private cloud or data center turns a variable, pay-per-use AI bill into a fixed cost a budget team can plan around. That is the case Boomi CEO Steve Lucas has been making publicly since his "Economics Always Wins" essay earlier this year.
A company willing to route Anthropic and OpenAI traffic through its own gateway is betting the gateway matters more than any model that passes through it.
A compliance officer wants data to stay inside company-controlled infrastructure. A finance leader wants a predictable AI bill instead of a variable one. Running models on a company's own infrastructure answers both, which is why finance now shows up early in Boomi's sales conversations.
The Enforcement Is What a Buyer Should Verify First
Boomi built a long list of features around the AI Gateway. Agent Context grounds an agent's answers in a company's own approved business definitions, so it is not guessing. Boomi Orchestrate lets non-technical staff describe a workflow in plain language and turns it into something agents can carry out. The Agent Trust Scorecard is a free, open-source tool that scores how ready a company's own setup is.
Boomi says a person has to approve any high-risk action before an agent takes it, and every action gets recorded in one place, in real time. Whether that holds up under a real incident is the claim to test, not take on faith. A Boomi-commissioned survey found most companies have not tested it yet.
Before you connect Anthropic, OpenAI, or Gemini traffic through Boomi's gateway, ask for redacted audit-trail examples from real production incidents, not a demo walkthrough, and check them against your own systems: does the record show which agent acted, what it touched, what it cost, and who approved it. A gateway that touches every model and every agent your company runs is worth trusting only if that record would hold up in your own audit, not just Boomi's.
Boomi, LP. "Boomi Delivers the Critical Infrastructure That Brings Control To Enterprise AI." Boomi Newsroom, 1 Sept. 2026, https://boomi.com/resources/resources-library/boomi-infrastructure-enterprise-ai-control/.
FinOps Foundation. "State of FinOps Survey: AI Value and Skills Top Priorities as FinOps Matures Across Technology." Linux Foundation, 19 Feb. 2026, https://www.linuxfoundation.org/press/state-of-finops-survey-ai-value-and-skills-top-priorities-as-finops-matures-across-technology-value-98-manage-ai-90-saas-64-licensing-48-data-center-1.
Forrester Consulting. "The Agentic AI Readiness Gap." Commissioned by Boomi, July 2026, https://boomi.com/content/ebook/agentic-ai-readiness-gap/.
