Customers

Teams that run agents like infrastructure.

Support desks, document pipelines, and revenue teams run production fleets on Ballast — with the audit trail to prove what every run did.

Illustrative examples of the workflows teams build on Ballast — not current customers. Real, named references will appear here as we onboard design partners.

Meridian

Fintech · 400 employees · Support automation

Meridian's support team was drowning in tier-1 volume: password resets, statement questions, and dispute intakes consumed 70% of agent hours. Their first LLM pilot drafted good answers but couldn't be trusted to send them — every reply still needed a human, and the pilot stalled.

On Ballast they rebuilt the flow as a gated workflow: classify, draft, then a human gate for anything touching money movement. Low-risk categories auto-send; regulated ones pause for approval with full context. Compliance signed off because every decision lands in the audit log with the exact state the approver saw.

The human gate is what got this past our compliance team. Approvals are in the trace, next to the model output they approved. That is the whole audit conversation.

Dana Whitfield, VP Customer Operations

63%

fewer tier-1 escalations

11,400

runs per day

$0.014

cost per resolution

Northbeam Logistics

Supply chain · 1,200 employees · Document pipeline

Northbeam processes bills of lading, customs forms, and rate confirmations from 300+ carriers, each in a different format. Their extraction scripts broke weekly, and a single crash mid-batch meant re-running thousands of documents from zero.

They moved the pipeline to Ballast: an extraction agent fans out per document, a condition node routes low-confidence extractions to a review gate, and checkpoints mean a worker restart resumes mid-batch instead of restarting it. The pipeline has survived 14 deploys and 2 node failures without losing a document.

We used to schedule extraction runs around deploy windows. Now we deploy whenever we want — the runs just resume from their last checkpoint.

Marcus Oyelaran, Head of Platform Engineering

38,000

documents per week

99.2%

straight-through rate

0

documents lost across restarts

Structa

B2B SaaS · 90 employees · RevOps routing

Structa's inbound leads sat for a median of 9 hours before routing. Reps did their own research, wrote their own outreach, and skipped the CRM hygiene. Every attempt to automate it stalled on one question: what did the system actually do with this lead?

Their RevOps team built the routing workflow in the visual builder without an engineering ticket: score, qualify, parallel research and outreach drafting, merge, then hand off to Slack. Every run is versioned against the workflow that produced it, so when scoring logic changes, they can answer exactly which leads were scored by which version.

I am not an engineer. I built the workflow our AE team runs 2,000 times a day, and I can defend every routing decision it has ever made.

Priya Ramanathan, RevOps Lead

9h → 4min

median lead response

2,100

runs per day

31%

more qualified meetings

2.4M

runs orchestrated*

99.97%

checkpoint recovery

48

production fleets

$120K

LLM spend attributed / mo

*Illustrative launch targets — Ballast is in public beta. Real usage numbers will replace these as customers go live.

Powering agent fleets at

Vantage

Meridian

Coreline

Northbeam

Helios Systems

Structa

Datum Labs

Ridgeline