← VuduVationsThe Intelligence Bureau
VuduIntelConsulting-AI Intelligence
← Edition No. 10
Architecture

Wipro's governance chief says agents now move faster than anyone can watch them.

If agents act before review can catch up, governance has to be built into the system, not bolted on after.

VuduVations Intelligence Bureau · May 27, 2026 · 2 min read
Governance & AuthorityOwn vs Rent
Share

Ivana Bartoletti runs privacy and AI governance globally for Wipro, and her warning is structural, not nervous. Modern AI agents do not simply analyze and wait for sign-off. They plan, decide, and act, and they do it in sequences that move faster than any human review cycle can follow. The moment an agent is running, the traditional oversight loop, the committee that meets, the checkpoint that reviews, the auditor who signs off later, is already behind.

Her answer is a phrase Wipro calls Trust by Design, and the load-bearing word is design. Governance, in her framing, is not a compliance label you attach after the system produces an output. It is an architectural commitment made before the agent runs. Bartoletti also names why after-the-fact review fails on its own terms: agentic systems reach outcomes through long chains of tiny decisions that are not individually explainable in any useful way. You cannot reconstruct each micro-step, so the thing worth governing is the system's behavior over time, assessed against clear boundaries set in advance.

In plain terms, the review has to move to the front. Instead of watching what an agent did and reporting on it afterward, the limits, who it can act as, what it is allowed to touch, when it must escalate, are written into the system before it starts. That is a different kind of deliverable from a governance deck. A deck describes the rules. An architecture enforces them while the work is happening.

The distinction is not academic, because it decides who owns the result. A compliance review is a service with an end date, the firm performs it and moves on. A governance architecture is a specification that either belongs to the client or leaves the client permanently dependent on whoever built it. Bartoletti has named the problem more precisely than most of the market. The open question, for Wipro and everyone reading her, is who turns the naming into something a client can actually hold.

When the agent moves faster than the review, the review is not oversight. It is a record of what already happened.
VuduVations Intelligence Bureau
The VuduVations Read
Bartoletti has described the problem with unusual precision. Consulting-as-Code, delivered by MCOS, is the shipped answer: governance written into the procedure before the agent runs, source-cited and structured for review rather than reviewed after the fact. The client owns that procedure and can keep running it when the model is swapped or the consultants leave. Incumbents write the architecture up in essays and rent out the delivery. The move is to hand the client a governed procedure they own outright, not a slide that describes one.
Share
Sources
Governing the Uncharted: Building Trust and Accountability in the Age of Agentic AI, Wipro
The AI Interview: Ivana Bartoletti, AI Magazine
Firms in this story: Wipro← Back to Edition No. 10