The real problem with enterprise AI agents isn’t model intelligence. It’s an architectural void: there is no reliable way to govern what an agent intends to do before it actually does it. When an agent modifies a financial record, triggers a payment, or approves a workflow on behalf of a business, it often operates in a black box — leaving security and compliance teams unable to verify actions until after the fact.
Devenex, which launched at Google Cloud Next 2026 in Las Vegas on April 22, aims to fill this gap with what it calls an Execution Control Plane. The company, built by the team behind the global professional services firm Abacus, positions this as a new category of infrastructure. It sits directly between an agent’s intent and its real-world execution, acting as a gatekeeper that produces audit-grade evidence for every move.
The stakes for getting this right are rising quickly. Gartner projects that by 2028, one-third of enterprise software will incorporate agentic AI, a massive jump from less than 1% in 2024. Yet scaling remains difficult. McKinsey has identified governance and risk as the primary barriers to scaling enterprise AI, ranking them ahead of model quality or talent. A 2026 survey from Deloitte underscores this, finding that 80% of leaders piloting AI agents cite security and compliance as their leading obstacle, up from 68% just a year earlier.
Shoaib A. Khan, CEO of Devenex, argues that the industry is misdiagnosing the issue.
“This isn’t an AI problem. It’s an architectural gap — and it’s the one our experience has prepared us to solve.”
He points out that most organizations currently lack the visibility required for accountability.
“Enterprises cannot answer four questions about any agentic action: who authorized it, what policy governed it, why it executed as it did, and whether they can prove it after the fact.”
To answer those questions, Devenex forces every agentic action through a four-artifact model: an Intent Record, an Execution Plan, the Governed Execution itself, and finally, the Execution Evidence. By requiring these artifacts, the platform enables pre-execution policy enforcement and dynamic human-in-the-loop governance. It also provides an immutable, append-only ledger for audit purposes, designed to help firms meet requirements for the EU AI Act, SOC 2, and ISO 42001.
The team behind Devenex brings nearly 40 years of experience from Abacus, a firm with over 5,000 professionals and 1,500 enterprise clients. Chairman Aly Kuly Khan emphasizes the necessity of this approach:
“Governance at this layer is not optional. It is a precondition for enterprise AI at scale.”
The competitive landscape is forming fast. The year 2026 has been framed by Futurum Group analyst Mitch Ashley as the start of the “Agent Control Plane Wars.” Microsoft introduced its own Agent Control Plane at Build 2026, while Google Cloud Next featured announcements around Agent Gateway, Agent Identity, and Agent Registry. Devenex differentiates itself by being framework-agnostic and cloud-native, offering deployment options that include SaaS, hybrid, and self-deployed models.
One practical caveat for organizations considering this: while the platform promises unified observability, integrating it into existing, fragmented agent workflows will require significant operational discipline. It is not a set-it-and-forget-it solution — it requires mapping existing policies into the Devenex framework.
For those ready to move past the pilot phase, Devenex is available immediately, with enterprise pilot programs open to qualified organizations. The operational challenge, as always, will be mapping existing policies into a new framework without creating the very bottlenecks the platform promises to eliminate.