The work, in production
across operator-led businesses.
We anonymize every case study by default. Our customers are operators in regulated industries — hospitality, freight, capital, healthcare, energy. They don't need a logo on our site. They need calibration that holds. The studies below are written with their permission, the numbers verified by their own audit teams.
- Operations
Sports & Entertainment · Operator A
A 47-venue group cut Monday operations meetings from four hours to thirty minutes.
"The agents are right often enough that we now spend the meeting on the cases they escalated."
— COO, Operator A · Engaged month 14- Decisions per week
- 14,200
- Operator override rate
- 2.4%
- Time-to-deploy
- 11 weeks
- Margin
Specialty Logistics · Operator B
A North American freight desk replaced two-thirds of its routing decisions with calibrated agents.
"The system told us the night dispatcher was actually the best router in the building. We promoted her."
— VP Operations, Operator B · Engaged month 9- Routing decisions / day
- 6,800
- Margin lift, sustained
- +11.4%
- Time-to-deploy
- 9 weeks
- Leverage
Multi-strategy Capital · Operator C
A discretionary fund built an agentic research desk that authored 41% of intake memos by month six.
"It does not replace the analyst. It removes the part of the analyst job nobody became an analyst to do."
— Head of Research, Operator C · Engaged month 6- Memos / month
- 480
- Analyst hours returned
- 2,100 / mo
- Time-to-deploy
- 14 weeks
- Throughput
Healthcare Infrastructure · Operator D
A national imaging network compressed prior-authorization from days to hours, system-wide.
"The clinicians said it the most clearly: the agent argues with the insurer so they do not have to."
— CTO, Operator D · Engaged month 11- Prior auths / week
- 38,000
- Median latency
- 2.4h
- Time-to-deploy
- 13 weeks
- Quietude
Outdoor Retail · Operator E
A 320-store retailer used a swarm to rebuild its weekly assortment and markdown decisions.
"Buying was the most political room in the company. Now it is one of the quietest."
— Chief Merchant, Operator E · Engaged month 8- Decisions / week
- 92,000
- Markdown waste
- −19%
- Time-to-deploy
- 10 weeks
- Resilience
Energy & Grid · Operator F
A regional generator handed peak-load dispatch to a calibrated swarm in fourteen weeks.
"The board wanted to know the failure mode. The honest answer was: we no longer have a single point of failure."
— Plant Director, Operator F · Engaged month 12- Dispatch decisions / day
- 11,400
- Cost-to-serve
- −8.6%
- Time-to-deploy
- 14 weeks
"We stopped having
the same fight every Monday."
The brief
Operator A runs a 47-venue group across hospitality, live events, and food & beverage. Monday operations meetings averaged four hours and resolved the same nine recurring categories of issue every week — staffing, inventory, service-recovery, supplier disputes.
The installation
Eleven-week deployment. Three calibration engineers on the account. A swarm of nine specialist agents covering the recurring decision categories, with the operations team on the loop and explicit escalation thresholds.
What changed
The Monday meeting now starts with the agent log, not a pile of issues. The team reviews the 12 to 18 cases the swarm escalated, signs off on a handful of edge calls, and is out the door in thirty minutes. The agents handle the rest before the meeting starts.
Where we are now
Fourteen months in. Operator override rate has stabilized at 2.4%. The calibration team meets monthly. Operator A has begun rolling the same swarm pattern out to two adjacent businesses in the holding company.
Production installations,
across forty-one industries.
- Hospitality & Entertainment38
- Logistics & Freight31
- Capital & Asset Management29
- Healthcare Infrastructure24
- Retail & Commerce22
- Energy & Industrials19
- Insurance & Risk17
- Professional Services14
- Public Sector12
- Other11
A study of your operation
is the first conversation.
We open every engagement with a thirty-minute working session and a calibration estimate within a week. No deck, no pitch — your actual decision data, replayed against the agents we would propose.
Book a deployment