Private AI Data Platforms
Your data. Your AI. Your environment.
Mid-market companies are sitting on a goldmine of operational data scattered across a dozen systems — your LMS, your CRM, your finance tools, your support platform, your communication apps, your homegrown databases. Almost none of it is connected. Most of it can't be analyzed without exporting to spreadsheets. And the AI tools that promise to help all want you to ship that data to their environment to do it.
PKG Systems builds the alternative: a private data platform, hosted in a Kubernetes environment we design and operate exclusively for you. We aggregate your data from every source you have, layer AI tools on top, and give your team a way to actually use it. Nothing leaves your environment. Ever.
What mid-market companies do today
Three patterns, all broken in different ways.
1. Spreadsheet exports + manual reconciliation
Someone on the operations team logs into each system, exports CSVs, opens them in Excel, runs VLOOKUPs, sends a PDF to leadership. The numbers are stale before they arrive. Nobody trusts them. Nobody can replicate them. The person who built the spreadsheet is the only one who knows what it actually means.
2. A separate analytics dashboard for every SaaS
Every vendor sells you "their" analytics module. You end up with eight dashboards, none of which talk to each other. The dashboard inside the LMS can't tell you whether learners who completed Course X are also closing more deals in the CRM — because the CRM is a different vendor, a different dashboard, a different export.
3. Feed it all to a public AI tool and hope
The most dangerous option. Take your customer data, your financials, your training records, paste them into a chat window, and ask for insights. It works — until your data ends up in someone else's training set, your competitor's prompt, or a regulator's enforcement action. Once data leaves your environment, you cannot un-leak it.
What PKG delivers instead
A complete data platform, built for you, owned by you, run by us.
Every component is yours. Every byte of data stays inside your cluster. Every AI prompt runs against models we deploy in your environment, against indexes we build from your data — never against shared models trained on someone else's information.
What's included
A packaged engagement covers the whole platform:
- Source connectors — REST API, database, file-drop, webhook ingestion from every system you run. Built once, scheduled, monitored, retried.
- Canonical data store — Postgres-backed, modeled to your business, designed for both reporting and AI retrieval.
- Reports & dashboards — Live SQL views surfaced through a web UI. CSV export. Role-aware access.
- Admin tooling — Operations console for the platform itself: monitor sync jobs, manage users, run migrations, audit everything.
- AI integration layer — Retrieval against your data, agents that take actions in your systems, conversational analysis tooling. Built on models deployed in your environment.
- Migration tooling — When you need to move data between systems (vendor swap, system retirement, M&A), we have working tools for it.
- Ongoing operation — We run the cluster, patch the software, respond to incidents, ship enhancements. Monthly run-rate, not pay-per-incident.
The privacy posture
This is not "we anonymize your data" or "we promise we won't look." It's architecture.
| Question | Answer |
|---|---|
| Where does my data physically live? | In a Kubernetes cluster in our AWS account, dedicated to you, with cryptographic isolation. Or in your AWS account, if you prefer. |
| Does PKG see my data? | Only when you ask us to debug something, and only with your explicit ticket. There is no "PKG Insights" team mining your platform. |
| Do third-party AI vendors see my data? | No. AI runs in your cluster, against models we deploy there. |
| Is my data used to train any model anywhere? | No. Single-tenant by design. |
| Can my data be subpoenaed from a SaaS provider? | There is no SaaS provider. PKG operates infrastructure for you; we don't aggregate customer data. |
| What happens if I leave PKG? | You take the cluster, the database, the code, and the documentation. We hand it over and walk away. |
What this looks like in practice
Two anonymized walkthroughs from work we run today.
Training franchise — 200+ locations
The customer is a global training and consulting franchise with hundreds of franchisee locations and a parallel direct-enterprise sales channel. Their challenge: they had two completely different LMS platforms running in parallel — an aging incumbent and a new strategic platform — with thousands of learning paths, tens of thousands of users, and overlapping content that nobody could reconcile.
We built them a private platform that:
- Pulls daily from both LMS APIs into a canonical Postgres database (~250 organizations, 1,500+ courses, 5,000+ users)
- Runs an automated migration engine that moves customers from the old platform to the new one organization-by-organization, with full rollback safety
- Provides an admin web UI for the operations team to monitor migrations, fix data issues, and run audits
- Surfaces dynamic reporting views that cross both platforms — something neither vendor could provide on its own
- Feeds a separate communication-coaching dataset (a Yoodli-pattern integration) into the same platform with AI-driven conversational analysis on top
The platform has been running daily in production. The customer's operations team uses the admin UI directly. Their leadership reads the unified reports. Both LMS vendors and the AI coaching vendor remain blissfully unaware of each other.
Communication coaching — daily AI-driven analytics
A second customer, in the same parent organization, runs a communication-coaching SaaS that produces conversation transcripts and behavioral analytics. We built them a daily ingestion pipeline that:
- Pulls every coaching session and transcript into the same private cluster
- Produces a daily analytics dashboard for the leadership team
- Runs AI-driven analysis against the corpus, with prompts the customer controls
This is the same platform pattern as the training franchise — different sources, different reports, different AI use case, same architectural foundation. That's the point of "packaged."
Pricing model
Three phases, all transparent:
- Discovery — A fixed-price scoping engagement (typically 2–4 weeks). We map your data sources, talk to your team, and produce a written architecture and a fixed-price build estimate. You own the document either way.
- Build — A fixed-price implementation against the discovery deliverable. No hourly billing. No scope expansion without your written sign-off. You see working software at the end of every two-week iteration.
- Run — A monthly run-rate that covers operating the cluster, the database, the application, and the integrations. Includes ongoing enhancements within an agreed envelope. Cancel with 60 days' notice; we hand over the system.
We do not bill time and materials. We do not charge per-user SaaS fees on top of the platform. We do not have account managers whose job is to upsell you.
Who this is for
Mid-market companies (roughly $20M–$500M in revenue) who:
- Run more than three operational systems that should be connected and aren't
- Have a strong instinct that their data is theirs and shouldn't be in someone else's cloud
- Don't have a 30-person internal IT department to integrate things themselves
- Have been quoted by a Big 4 firm and walked away
- Have tried to "just use" a SaaS analytics product and hit its ceiling within six months
If that's you, we should talk.