LIA FieldOps: Autonomous AI agents for field technician support
The problem that field operations have been waiting to solve
Field operations are one of telecom’s largest OPEX items and one of
its most human-dependent. To understand the scale, look at TurkNet
alone: 1.46 million subscribers, around 800 technicians, and
approximately 912,500 field visits every year. That is more than 2,500
customer visits every day, over 1,100 visits per technician per year,
and €25.2 million in annual field-operations cost before even
considering repeat visits, customer frustration, or lost commercial
opportunities.
Every one of those visits depends on decisions made under pressure.
Every time a technician needs guidance, an authorisation, or a second
opinion, the call goes to the NOC or CO — a scarce pool of experts who
handle everything sequentially, across 7 to 10 systems. At this scale,
the bottleneck is not the technician. It is the operational substrate
underneath.
Previous attempts at AI augmentation stayed on slides. The tools
weren’t fast enough, the integrations weren’t deep enough, and the
governance wasn’t there. LIA FieldOps is the answer to all three.
A three-body solution
LIA is an autonomous, composable, ODA-aligned AI agent that acts as
the first operational decision layer for telecom field operations. She
pairs with the field engineer through every step of a service visit —
from dispatch to fix and beyond.
The architecture is built from 3 purpose-fit components:
Tech Solutio’s Pantheon-based orchestration is LIA’s mind — governing
decisions, enforcing confidence thresholds, and managing risk
policies.
Volt Active Data’s VoltDB in-memory operational store is her nervous
system — returning every operational read in sub-10ms, fast enough
that the full detect-decide-act-audit cycle runs in real time during
the visit.
Binom’s embeddable SDK is her hands and senses — a white-labeled
mobile interface that plugs into the operator’s existing field
workflow without ripping it out.
Every business capability is exposed via 10 TM Forum Open APIs mapped
to ODA-aligned MCP servers. Any backend, any operator, same blueprint.
What LIA does on a visit ?
LIA reads the customer state, device telemetry, port topology and
equipment manuals from the operator’s systems in real time, and
queries all of them concurrently in a single MCP session. A human NOC
escalation runs those same lookups sequentially, across multiple
systems. LIA collapses them into one pass.
She guides the engineer step by step. When she needs data from the
field, she emits a structured survey that renders as native UX
elements in the mobile app, instead of a text message read in a wiring
closet. Every step is captured, every decision is logged with a full
audit trace, and every claim she makes is one click away from the
source that produced it.
LIA does not improvise. When uncertainty is high, she performs a
structured handover to NOC/CO with full context, rationale, and
actions taken — so the expert picks up exactly where she left off,
with nothing to reconstruct.
After the fix, at the moment customer goodwill peaks, she detects the
trust window and surfaces a context-aware offer for the engineer to
present in person.
In production. Measurable. Scalable.
LIA is live at Vivo, Telefónica Brazil, where she already resolves 37%
of CO-escalation conversations end-to-end with no human handoff —
using only 6 playbooks and limited integration. Average Handling Time
dropped 40% in the first 90 days.
The same architecture at full integration projects 60% autonomous
resolution. New capabilities ship as playbooks, not software releases.
Vivo has 27 more in the pipeline.
And this is where the scale becomes decisive. For an operator like
TurkNet, even a 1% reduction in unnecessary truck rolls means nearly
10,000 fewer field trips per year. A 10–15% reduction means tens of
thousands of avoided visits, millions in annual savings, fewer
kilometres driven, lower emissions, and more technician capacity
returned to the business.
Champion operators on this Catalyst include Telefónica/Vivo, TurkNet
and Omantel. LIA FieldOps is a repeatable blueprint any CSP can adopt
— and a concrete, governed path to TM Forum Autonomous Networks Level
4.
Field operations have been waiting for years. LIA is finally here.
