The operations layer for AI/automation pipelines
at engineering-led mid-market IT teams.
Your team built 80% of the stack. We operate the production-grade last 20% — reliability, observability, audit trail, exception handling — on your infrastructure. You keep the software. We run the operations.
For CIOs/CTOs at $50M–$1B engineering-led IT teams in manufacturing, distribution, logistics, professional services, and financial services.
The wedge
You've already done the 80%. The last 20% is exponentially harder.
You've built it in Power Automate. Or n8n. Or Python + Claude. Or Graph API custom code. Whichever stack — the first 80% took weeks of focused engineering and it works.
The last 20% is different work:
- Production reliability across long-tail vendor variability
- Per-tenant credential isolation for clients or business units
- Exception escalation with named approvers
- Audit trail that satisfies SOC 2, state regulator, and customer-driven compliance asks
- Observability for incidents your team needs to debug at 2 AM
- Multi-channel orchestration when the agent action surface expands beyond email
That work is exponentially harder than the 80%. It's also work that doesn't differentiate you — it just has to work. Your senior engineers' next 6 months shouldn't go there.
JieGou operates that last 20% on your infrastructure.
What we operate
The operations layer — six primitives we ship and run
Per-tenant credential vault
AES-256-GCM encrypted credential storage for each client tenant or business unit. Permission scoping per-agent + per-action. You bring your own KMS or use ours.
MCP gateway with ACL + audit
Single endpoint your existing AI agents call. JieGou logs every action with structured metadata, applies scoping, queues high-risk actions for approval. Drop-in replacement for direct LLM API calls.
Approval queue with escalation paths
Configurable per tenant — full autonomous for low-risk actions (drafting), required-human-approval for high-impact actions (sending, modifying, deleting). Named approvers, email + Slack + Teams escalation.
Structured audit logs (CloudEvents v1.0)
Every agent action timestamped, signed, exportable to your SIEM (Splunk, Datadog, custom). Schema-compatible with the audit standard your compliance team already knows.
13-channel multi-channel scope
Email, SMS, Microsoft Teams, Slack, WhatsApp, Discord, voice, and 6 more. Same governance applies across all channels — once you ship the layer, it works wherever your agents act.
Monthly compliance evidence pack
Concrete PDF + signed audit trail. Customer-facing artifact for your SOC 2 audit, state regulator inquiry, customer compliance request, or E&O renewal.
Deployment
Three deployment shapes. Default: customer-VPC.
Default for engineering-led IT
Customer-VPC
JieGou code deploys to your AWS or Azure account via Helm chart or Terraform module. You own the software, the data, the infrastructure. We operate remotely — monitoring, incident response, observability, monthly evidence pack delivery.
Zero infrastructure burden
JieGou-hosted SaaS
For teams that want zero infrastructure burden. We host; you access via web UI + API. Same governance primitives, different ownership.
Data-residency-strict
On-prem
For data-residency-strict deployments. Customer's data center; JieGou container; customer's IT runs it. Quarterly upgrades from JieGou ops.
How it works
The 6-stage engagement
| Stage | Duration | What happens |
|---|---|---|
| 1. Discovery | 30–60 min | We walk through your existing automation stack with you and name the specific last-20% pain. |
| 2. Technical Fit Assessment | 1–2 weeks | Joint architectural review. Your engineering lead + our team. Gap analysis. POC scope agreed. |
| 3. POC Scoping | 2–4 weeks | Signed POC agreement. Access to systems. Named success criteria. |
| 4. POC Execution | 4–6 weeks | We deliver the POC pipeline end-to-end on your VPC. Your team observes and validates. |
| 5. Commercial Close | 2–4 weeks | Pricing, contract, procurement. |
| 6. Implementation | 4–12 weeks | Production deployment. Your team becomes self-sufficient on day-to-day. We operate ongoing. |
Total: 3–6 months from first call to commercial close. Plus 1–3 months to production.
Pricing
Transparent pricing
Engagement fee + annual support + BYOK LLM. No SaaS subscription, no per-seat lock-in. You own the software.
| Component | Price | Notes |
|---|---|---|
| First pipeline engagement fee | $75K | One-time. Covers POC + production implementation + 90-day hyper-care. |
| Annual support (first pipeline) | from $50K/year | Maintenance, SLA, observability, monthly evidence pack, quarterly business review. Includes 60K-invoice/year volume cap; higher tiers available. |
| Additional pipelines | from $35K + from $30K/year | Per-pipeline fee declines as you add more. Integration infrastructure is reusable across pipelines. |
| Platform Tier (3+ pipelines) | $300K – $850K / year | Replaces per-pipeline math at scale. Tier 5 / Tier 10 / Tier 20 available. |
| LLM costs | BYOK pass-through | You bring your own Anthropic / OpenAI keys. JieGou does not mark up LLM costs. Typical: ~$5K/year for a mid-volume pipeline. |
| Multi-year terms | up to 25% off | 3-year ~18% off (sweet spot); 5-year up to 25%. Optional prepay bonuses stack on top. |
Example — single pipeline
$75K engagement + $50K/year support + ~$5K/year BYOK LLM = $130K Y1, $55K/year steady-state.
Example — 3 pipelines, mid-market
$140K engagement (P1 $75K + P2 $35K + P3 $30K) + $105K/year support + BYOK LLM ≈ $250K Y1, $105K/year steady-state.
Multi-year terms unlock up to 25% off (3-year ~18% off is the sweet spot). At 3+ pipelines, Platform Tier replaces per-pipeline math; ask us on the call.
Current pilot
Where the layer is running today
$750M North American manufacturer. Invoice-extraction pipeline (Python + Claude + n8n existing stack). Phase 1 NDA'd May 2026. We operate the production-grade last 20% — vendor-format variability handling, exception escalation, per-tenant audit trail, structured output to their middleware.
Customer name added with permission post-Phase-1 delivery.
The landscape
Where we fit
We're not
The substrate
Cisco DefenseClaw (Apache 2.0 free) + Microsoft Agent Governance Toolkit (MIT free) + NVIDIA NemoClaw are open-source substrate your team can install themselves. If you want the tools, install them — they're great.
We're not
The consultancy
McKinsey AI, BCG GAMMA, Accenture AI, and Slalom run delivery engagements at $150K–$400K per engagement, then hand off. Your team owns the result post-handoff.
We are
The ongoing operations layer between
We operate the substrate (whether you've installed AGT yourself or we ship JieGou's) on your infrastructure, after the consultancy hands off. We sell operations; we don't sell tools you'd run yourself.
Things every engineering-led IT team asks
Honest answers — including where free OSS tools or in-house engineering are a better fit
Talk to a human, not a chatbot
15 minutes. No deck. We look at your existing automation stack with you and tell you whether JieGou is a fit. If it's not, we'll tell you who is — DefenseClaw, AGT, NemoClaw, or any of the consultancies — based on what your stack actually needs.