Attributed runs
A run was counted when the production measurement attributed recorded execution activity within the stated window.
Bewize research
What did managed AI agent operations look like in one bounded production measurement?
This report records first-party operational evidence collected on September 16, 2026. It measures platform activity and participation within stated boundaries. It is not an industry benchmark and does not claim customer ROI, accuracy, productivity improvement, or business impact.
Measured operational scale
59,124 attributed runs across 115 calendar days
From May 25 through September 16, 2026, the measured system attributed 59,124 runs and 25.853 billion tokens. At collection time, 1,108 tenants were reported ready; public health and the listed platform checks passed at that collection point.
Production measurement
Exact values retain their collection date and boundary.
May 25–September 16, 2026
- 59,124 attributed runs
- 25.853 billion attributed tokens
- 115 calendar days
- 1,108 tenants reported ready
Token activity
The token total describes measured platform usage. It is not a cost, efficiency, quality, or value calculation.
Ready tenants
Ready means reported ready at collection time with public health and listed platform checks passing then. It is not a claim about every possible workflow.
Time boundary
The activity window contains 115 calendar days. Later activity is outside this report and should not be inferred.
Roomcord participation sample
Agents produced 51.5% of messages in the sampled active rooms
The authenticated operator-visible Roomcord workspace contained 22 rooms, all with agents installed, and 54 agent installations. A separate 757-message sample from 16 active agent rooms contained 390 agent messages and 207 structured A2UI results.
Bounded room sample
These percentages describe only the sampled rooms and messages.
757 messages across 16 active agent rooms
- 390 agent messages
- 51.5% agent-message share
- 207 structured A2UI results
- 27.3% of sampled messages used A2UI
Active participation
The sample shows agents contributing directly to room activity rather than existing only as installed integrations.
Structured interaction
A2UI results indicate that part of the sampled work used structured interactive output, not only conversational messages.
Workspace boundary
The 22-room and 54-installation counts belong to the authenticated operator-visible workspace, not every Roomcord environment.
No outcome inference
Message share and structured results do not establish correctness, usefulness, adoption quality, labor savings, or business results.
Observed non-development workflow classes
The collection identified real operational categories without attributing them to a named customer or claiming outcomes.
Team coordination
Team and portfolio coordination around shared operational context.
Meetings
Meeting-related capture, decisions, and follow-up work.
News intelligence
Source collection, verification, briefing, and editorial review.
Marketing operations
Research, drafting, review, delivery support, and outcome learning.
Evidence work
Legal evidence and other work requiring traceable source handling.
Customer feedback
Collection, classification, follow-through, and resolution support.
Creative production
Coordinated creative planning and production workflows.
Methodology and limitations
How to read this evidence
This is a dated first-party operational measurement. Every number should travel with its date, identity boundary, window, and limitation. Missing dimensions are unavailable, not zero.
Collection date
The read-only collection was taken on September 16, 2026.
Different measurement scopes
Platform run and token totals, ready-tenant status, visible workspace counts, and the message sample have distinct scopes and must not be combined into unsupported ratios.
Anonymized reporting
No customer identity, logo, testimonial, or customer-specific outcome is asserted by this report.
What the report supports
The evidence supports claims of measured operational scale, agent participation in the sample, structured interaction, and the listed workflow classes.
What it does not support
It does not prove ROI, accuracy, productivity improvement, cost reduction, user satisfaction, fleet-wide behavior, or causal business impact.
Connect the evidence to the operating model
Governed AI operating rooms
See how people, agents, context, evidence, and approval boundaries fit into recurring team work.
Explore →From personal agents to company scale
Read why individual agent deployment became a broader context and operations problem.
Explore →Operational evidence
Review the run, request, schedule, usage, and attribution controls behind managed agents.
Explore →Define the evidence for your first operating room
Choose one recurring workflow, then define the activity, completion, corrections, human intervention, adoption, operating cost, and approval evidence required for a stop, revise, repeat, or expand decision.
Architecture conversation
Tell us what your team needs to control
Share your deployment boundary, number of agents, work surfaces, and governance requirements. We will reply by email to arrange a focused technical discussion.
Email the Bewize teamThis opens your email application. Read our Privacy Policy.