Who we work with
We work with operators, ISPs, MVNOs, wholesale and infrastructure providers, and the full range of fixed, mobile, fibre, and WISP-style connectivity businesses. Strategy through to production, from the inside of the business, not from a case-study deck.
How we read telco operations
When Paul built and ran a wireless operator from scratch, he designed a simplified operating model based on the TM Forum Business Process Framework. Seven domains that map the full lifecycle of a connectivity business, from lead through to cash. We use that structure to read where AI reaches production across connectivity businesses.
The Diagnostic runs on this map. One executive conversation establishes which domains carry the most value and constraint. The interviews that follow map opportunities inside those domains, scored on feasibility and commercial outcome. The result is a ranked shortlist of use cases that can reach production, sequenced on dependencies and data readiness.
The seven domains
Marketing, sales, quoting, order capture
Where prospects enter, get qualified, and convert into orders. In a connectivity business this includes campaign management, lead scoring, quote generation, and order validation. AI fails here when the business does not have clean conversion data or when quote logic is too variable to codify. It lands when lead volumes are high, qualification is repeatable, and the sales motion follows a process.
Example constraint: Can you pull the last 12 months of leads with outcome, source, and time-to-close without a data project?
Provisioning, fulfillment, service activation
From order acceptance through to service live on the network. Covers order orchestration, inventory assignment, network provisioning, and activation testing. AI fails when provisioning workflows are manual and inconsistent, or when the systems behind them cannot be queried. It lands when exceptions are predictable, fallout reasons are logged, and order data is structured.
Example constraint: Do you know where orders fall out of automation and why, or does each fallout get triaged from scratch?
Network monitoring, fault management, service assurance
Detecting faults, diagnosing them, and restoring service. Spans network monitoring, alarm correlation, ticket routing, and outage management. AI fails when fault data is noisy or when resolution depends on institutional knowledge that has never been documented. It lands when alarms are structured, resolution patterns are repeatable, and the mean time to detect or repair matters to the P&L.
Example constraint: Can your network operations team describe the top five fault types and their resolution steps, or is every ticket different?
Usage mediation, rating, billing, revenue assurance
From usage records through to invoice and payment. Includes mediation, rating, billing runs, invoice generation, payment processing, and collections. AI fails when usage data is inconsistent or when billing errors are discovered too late to correct. It lands when mediation errors can be caught before the billing run, when disputes follow patterns, and when revenue leakage is measurable.
Example constraint: Do you bill correctly the first time, or do you chase credit notes and adjustments after customers complain?
Customer service, ticketing, issue management
Handling inbound customer enquiries, service requests, and complaints through to resolution. Spans call routing, ticket management, self-service portals, and case escalation. AI fails when contact reasons are too varied to classify or when resolution depends on context that lives only in the agent's head. It lands when contact volumes are high, reasons are repeatable, and resolution steps can be scripted.
Example constraint: Can you classify inbound contacts by reason and outcome, or is your reporting limited to channel and duration?
Vendor management, interconnect, wholesale settlement
Managing supplier relationships, interconnect agreements, wholesale service delivery, and partner settlements. Covers procurement, vendor performance tracking, wholesale billing, and interconnect reconciliation. AI fails when supplier data is inconsistent across contracts or when settlement disputes require manual investigation. It lands when interconnect volumes are high, settlement cycles are predictable, and discrepancies can be detected before they compound.
Network planning, capacity management, rollout
Strategic and tactical network planning, capacity forecasting, capex planning, and infrastructure deployment. Spans demand forecasting, site planning, network dimensioning, and rollout execution. AI fails when planning depends on informal judgement or when historical demand data is incomplete. It lands when capacity planning follows a repeatable model, when rollout dependencies are documented, and when capex decisions can be tied to measurable demand.
How we use this in engagements
One executive conversation with the CEO or GM establishes which domains matter for the business and where value concentrates. The structured interviews that follow map opportunities inside those domains, scored on feasibility (people, process, data, systems) and value (revenue, cost, customer experience, risk). The output is a ranked shortlist of use cases that can reach production, with the production constraints stated explicitly. We are blunt about what will not work.
Strategy work extends the Diagnostic shortlist into a full roadmap. Each opportunity gets a build-versus-buy call, an order-of-magnitude cost, and a sequence that accounts for data readiness and system dependencies. The roadmap maps back to the process domains so the leadership team can see where the work concentrates and where it does not. The board summary names the three initiatives that move first and why.
Beyond the seven operational domains, the Diagnostic also scores data readiness (accessibility, quality, structure), governance posture (risk appetite, compliance constraints, decision rights), and delivery capacity (team skills, vendor dependencies, integration capability). These gates determine which opportunities can be resourced and which need to wait. We do not publish the full instrument. It exists to make the read defensible.
Paul Salvage built and ran technology from the operator seat: CIO of a major telco group across 24 countries, Head of R&D at a telco, and Founder and COO of an open-source wireless operator built with no vendor software or licence fees. Outside telco, CTO of a Silicon Valley scale-up, CEO of a leading no-code agency, and founder of a wealth management AI startup. Building with AI in production since 2017.
Paul has also operated in mobile financial services and in US healthcare, including medical-record retrieval, summarisation, and retrieval pipelines. Those years inform judgment. We run Assess, Build, Lead for telco and connectivity businesses.