What 30 Years of Broadband Intelligence Looks Like When It Runs as an Agent
Rhyan J. Neble | Founder & CEO, Extended Systems Intelligence | April 2026
Everyone pitching AI to ISPs claims to have telecom domain knowledge. I want to be specific about what that actually means — and why the provenance of domain knowledge matters more than almost anything else in agentic AI deployment.
The XSI LodeStone Telecom Skill Library is being built in partnership with ETI Software Solutions. ETI has been building broadband OSS/BSS software for Tier 2/3 ISPs for 30 years. Our founding team built some of ETI's most critical operational tooling from the inside. Understanding what that means for the quality of the AI we're building requires understanding what carrier-grade domain knowledge actually is — and how it gets encoded into an agent.
Domain knowledge for an ISP AI agent is not documentation. It is the accumulated operational intelligence of engineers who have shipped working software into production ISP environments and debugged it when it failed. That knowledge cannot be bought. It can only be built.
WHAT "DOMAIN KNOWLEDGE" ACTUALLY MEANS FOR FAULT MANAGEMENT
A Nokia 7750 SR router has roughly 800 pages of CLI documentation. The documentation tells you what commands exist. It does not tell you, for a specific fault condition on a specific interface type under specific traffic load conditions, which command sequence to run first, which to run second, and what output to check before proceeding.
That knowledge — the sequence, the decision logic, the edge cases, the "when this happens, don't do that" institutional memory — lives in the heads of engineers who have spent years resolving production faults on specific equipment in specific network environments. It is built from incident reports, from post-mortems, from the painful experience of having done the wrong thing and learning why it was wrong.
Encoding this into an agent Skill Library is not a data entry project. It requires access to those engineers, their incident histories, their operational playbooks, and the OSS/BSS platforms that have been accumulating this knowledge in structured form for years. ETI's Intelegrate platform is exactly that: decades of carrier-grade operational intelligence encoded in production software that serves hundreds of ISPs across North America.
WHAT THE ETI PARTNERSHIP ACTUALLY PROVIDES
The ETI partnership gives XSI LodeStone four things that no amount of general AI capability can substitute for:
1. API access to Intelegrate — the operational data platform that real ISP networks run on. The agent is not working from scraped documentation or synthetic training data. It is working from the same real-time network data that ETI's customers use to manage their operations every day.
2. Co-development on the Skill Library — ETI engineers who built the fault management, provisioning, and compliance workflows that Intelegrate implements are contributing directly to the agent knowledge base. The CLI sequences in XSI LodeStone's Nokia, Calix, and Adtran libraries are validated against real equipment by engineers who have worked with that equipment in production.
3. A channel to operators who trust ETI — the hardest part of selling any new technology to Tier 3 ISPs is the trust deficit. A new vendor with a new product has no track record with an operator who has seen many vendors come and go. ETI introducing XSI LodeStone to its customer base transfers 30 years of ETI's trust equity to our initial deployments. That trust cannot be manufactured.
4. Commercial validation — ETI's pre-purchase of 10 appliances, with a significant deposit on signing, is not courtesy. It is a commercial commitment from a company with full visibility into our technical approach and full knowledge of the market we're serving. When ETI writes a check, it means something to every ISP operator they call on our behalf.
WHY THIS RELATIONSHIP IS STRUCTURAL, NOT INCIDENTAL
Our founding team worked together at ETI. We built ETI's FCC BDC compliance automation, cloud infrastructure, and program delivery capability. We left not because of any conflict with ETI but because we saw a product that ETI, as a 30-year OSS/BSS vendor, was not positioned to build: an AI appliance layer that extends their platform's value into autonomous operations.
The ETI partnership is the formalization of a relationship that predates the company. We are not approaching ETI as a channel partner we found in a vendor directory. We are working with colleagues who know our capabilities in detail, who understand the market problem we're solving, and who have a direct commercial interest in seeing our product succeed.
That relationship is the moat that is hardest to replicate. A competitor with capital can hire engineers. It can license Nokia CLI documentation. It can build an agent framework. What it cannot do in 12 months is establish the institutional trust with ISP operators that ETI has built over three decades — or the working relationship between technical teams that makes co-development actually produce carrier-grade results rather than plausible-sounding demos.
THE INTELLIGENCE ACCUMULATION EFFECT
As XSI LodeStone deploys across ETI's customer base, the Telecom Skill Library becomes richer. Each new fault type encountered in production, each new equipment configuration handled, each edge case resolved in a real network environment adds to the operational intelligence encoded in the platform. The first deployment is backed by 30 years of ETI domain knowledge. The tenth deployment is backed by 30 years of ETI knowledge plus the operational experience of nine live production deployments.
This is the compound advantage that first-mover position in a vertical market creates. The platform gets better with every deployment. The competition, starting later, starts with less knowledge and falls further behind with every passing month. This is how you build a moat in an AI market — not through model capability, but through the accumulation of domain intelligence that can only be built from production deployment.
Rhyan J. Neble | Founder & CEO, Extended Systems Intelligence | rneble@xtendedsystems.com | xsilodestone.ai
Q&A with Rhyan
Extended questions from discussions — answered in full.
Domain knowledge is not documentation—it is accumulated operational intelligence from engineers who have shipped working software into production ISP environments and debugged it when it failed. A Nokia 7750 has 800 pages of CLI documentation, but it doesn't tell you which command sequence to run first for a specific fault under specific conditions. That knowledge—sequence, decision logic, edge cases—lives in engineers' heads and is built from incident reports and painful experience.
You can hire engineers, license documentation, build an agent framework—but you cannot establish in 12 months the working relationship with ISP operators that ETI has built over 30 years. Competitors starting now face a choice: build integrations from scratch (18-month project) or integrate with TM Forum-aligned platforms that exist (3-month project). That timing moat is the hardest to replicate.
ETI provides four things no amount of general AI capability substitutes: API access to Intelegrate (the operational platform real ISP networks run on), co-development with ETI engineers who built the fault management workflows, a channel to operators who trust ETI (30 years of trust equity), and commercial validation (pre-purchase of appliances with significant deposit, signaling market confidence).
As XSI LodeStone deploys across ETI's customer base, the Telecom Skill Library becomes richer. Each new fault type encountered in production, each equipment configuration handled, each edge case resolved adds to encoded intelligence. The first deployment is backed by 30 years of ETI knowledge; the tenth is backed by that plus production experience of nine live deployments. This is the compound advantage first-movers build.
Common Questions
Search-ready answers to the questions we hear most often.
Domain knowledge is not documentation. It is accumulated operational intelligence from engineers who have shipped and debugged software in production ISP environments. A Nokia router has 800 pages of CLI documentation, but it doesn't tell you which command sequence to run first for a specific fault under specific traffic conditions. That knowledge lives in engineers' heads and is built from incident reports and painful experience.
You can hire engineers and license documentation, but you cannot establish operator relationships or familiarity with real ISP patterns in 12 months. Competitors starting now face a choice: build integrations from scratch (18-month project) or integrate with existing TM Forum-aligned platforms (3-month project). The timing moat created by domain knowledge is harder to replicate than capital or engineering resources.
ETI provides: (1) API access to Intelegrate platform serving hundreds of ISPs, (2) co-development with engineers who built fault management workflows, (3) channel to operators who trust ETI's 30-year track record, (4) commercial validation through pre-purchase commitment. This domain knowledge transfer cannot be manufactured.
As XSI LodeStone deploys across customer base, Telecom Skill Library becomes richer. Each new fault type, equipment configuration, and edge case resolved in production strengthens the platform. The first deployment is backed by 30 years of partner knowledge; the tenth is backed by that plus nine live deployments. This compound advantage is how durable moats form in vertical markets.
The founding team built ETI's FCC BDC compliance automation, cloud infrastructure, and program delivery capability from inside. They left not from conflict but to build a product—an AI appliance layer—that ETI's 30-year OSS/BSS platform positioning couldn't address. The ETI partnership formalized a relationship predating the company, with colleagues who know technical capabilities in detail and have direct commercial interest in success.
A competitor with capital cannot quickly replicate institutional trust with ISP operators built over 30 years, or the working relationships between technical teams that produce carrier-grade results rather than plausible-sounding demos. Domain knowledge captured in production software serving real customers creates a moat that compounds with every deployment—a structural advantage that capital cannot quickly overcome.