Agents are going mainstream. So who will manage them?
Welcome to the first issue of Butterfly Radar. Agents are spreading. So who will manage them? In this issue we look at managing AI agents in the enterprise and at Butterfly’s latest projects.
Agents are scaling in large companies
McKinsey’s 2026 global AI survey shows that AI agents in large companies are moving from pilots to use at scale.
Share of companies with annual revenue above $1 billion that are scaling AI agents. It was 27% in 2025.
Source: McKinsey, The State of AI in 2026As the number of agents grows, organizations face a new need: designing, authorizing and monitoring agents from a single place.
Butterfly brings agent design, management and connection to workflows together on one platform.
The shift from human-in-the-loop to "human-on-the-loop"
Among the enterprise AI trends at The AI Summit New York, governance, transparency and human oversight stand out alongside multi-agent systems. This is also changing the human role in the process.
A person approves every step; work waits during approval.
The agent handles routine work; a person steps in for exceptions and critical decisions.
In Butterfly, control points are defined within the process through workflows and rule engines. Routine steps move forward automatically; cases that need approval are routed to the right person.
RG Labs moved its project management from Jira to TaskFlow
89 projects, 32,130 issues, 11,190 comments and 34,112 hours of time logs were migrated to TaskFlow, which runs on Butterfly; sprints, boards, workflows and sprint closing reports were also part of the migration.
The migration was compared with live Jira on eight separate measurements and the results matched exactly; all 69 migration tests passed. Years of work history were carried over with no loss.
Self-service reporting at a rail transit operator
Butterfly made reporting self-service at a large urban rail transit operator. HR and ridership reports are now prepared by the teams themselves.
Duration of a measured operation after the data infrastructure was renewed.
AI agents in IT operations
In the IT operations of a large telecom operator, Butterfly built a setup where AI agents support the analysis of incident records. For each record, the agents prepare likely root causes and suggested fixes; the decision to apply them stays with the expert. Agents, workflows and integrations are managed from one platform.
Retail expense process, from approval to accounting entry in one workflow
Butterfly digitized the domestic and international expense process of a large retail company. Expenses are entered and approved on the same platform; approvals are routed to the relevant manager according to the organizational hierarchy. Tax breakdowns are automatic, exchange rates are kept current and accounting entries are transferred to SAP with account codes.
Task assignment for field teams is now automated
Butterfly automated task assignment in an internet service provider’s field CRM. Tasks are assigned according to defined rules and notifications are sent instantly. Partners manage their own regions on the same platform.

Where were we this month?
In October, we met with SOCAR Downstream, Baku Metro and the Turkish Embassy in Baku and attended the ADEX fair in Baku; in Tashkent, we met with Ziraat Bank Uzbekistan.
- McKinsey, The State of AI in 2026: On the road to ROI (2026)
- The AI Summit New York, AI in Action: From Experimentation to Enterprise Value (20.10.2026)
Let’s decide where to start, together.
In a discovery call, let’s talk about your existing systems and your priority process, and define the right starting area for you together.