Firstsource · Kairos OS · Intelligent Context Framework
Load Your Sources
The ICF ingests fragmented operational knowledge — troubleshooting guides, SOPs, process maps, and call transcripts — and unifies them into a continuously evolving intelligence layer for ABC Mobile tech-visit operations.
LOADED
Troubleshooting Guides (DOCX)
Step-by-step resolution flows for Broadband, ABC Q, and satellite issues
4
Documents
LOADED
Process Flow PDFs
Visual troubleshooting flowcharts — Slow Speeds, No Connection, No Signal, No Sound, Connectivity, Engineer Visit
7
Flow Diagrams
LOADED
Engineer Visit Transcripts
3,193 real booking conversations between advisors and customers
Ingestion complete · 1,847 signals extracted from 243,000 lines
57
ICF Nodes
84
Edges
42
Gap Flags
70,453
Sessions
ABC Mobile Tech-Visit Intelligent Context Framework — v0.5
The booking funnel has a 76% no-fault problem.
Across 15,967 engineer visits in the no-fault corpus, 12,198 returned with no fault found or a TT-no-fault clear.
The same advisors who dispatch these visits reference the canonical Service Checker tool in only 3% of their booking calls.
This dashboard turns the underlying corpus into a living knowledge graph of issues, decisions, systems, and gaps.
Total visits analysed
70,453
TV + BB visits across the corpus
Knowledge-graph nodes
57
Issues · Resolutions · Systems · Decisions · Exceptions
Edges
84
Issue → System → Resolution paths captured
% No-Fault-Found
76.4%
of NFF-corpus visits • 12,198 of 15,967
Session Volume by Domain
Booked symptom volume rolled up to ICF domain (where engineer cause analysed)
Corpus Breakdown
Source files analysed for v0.1 build
SOURCE
SESSIONS
PERIOD
Engineer Bookings & Transcripts
3,193
May–Jun 2025
No Fault Found Data — ABC
15,967
Jun–Sep 2024
FS Queries — TV Complete
38,908
2024
FS Queries — TV Not-Complete & Cancelled
7,922
2024
FS Queries — Broadband
23,623
2024
FS Queries — Volume Aggregates
85
Jun 3 – Sep 10 2024
FS Queries — Cancellation Summaries
132
2024
Engineer visit Data — FS monthly
22
Jan – Aug 2024
Node Composition & Resolution Outcomes
How the 57 nodes split across the five ICF types
Issue
13
Resolution
14
System
11
Exception
11
Decision
8
Resolution mix across the 15,967-visit NFF corpus
% NFF Close (engineer found no fault)
76.4%
% Genuine engineer fix
21.7%
% Guided self-serve (estimated)
≈ 1%
% Cancelled / Not-Done
11.5%
Gap Register Summary
Counts by type — full register on Page 3
Process Gap
Active failures, no doc'd fix path
22
SOP Gap
Real scenario, no procedure
10
Training Gap
SOP exists, agents skip it
21
Product Gap
Customer ask doesn't exist
4
Top Operational Risks
01Service Checker is referenced in only 3.1% of bookings. The single biggest control point for whether to dispatch an engineer is being skipped on 96.9% of calls.
0297.3% of NSS-Permanent bookings end NFF. 4,162 of 4,276 dispatches with no fault found — the single biggest wasted-cost cluster.
04231 'I'm not trained for this job' notdone codes point to dispatch logic that doesn't validate skill prerequisites.
0529.4% repeat-call rate after a tech booking — friction not progress.
ABC Mobile Tech-Visit Context Graphv0.1
Sessions 70,453
Nodes 57
Edges 84
Critical 14
v0.1 baseline May 2026
Source ABC Mobile tech-visit corpus 2024-2025
Built by Firstsource Kairos
Next: v0.2 +new Q4-2025 data
ABC Mobile Tech-Visit ICF · Knowledge Base · Troubleshooting Intelligence
Structured Knowledge Base
All troubleshooting guides, process flows, and resolution procedures — extracted from 11 source documents (4 DOCX + 7 PDF) — structured into a searchable, agent-ready knowledge base with co-pilot integration.
Knowledge Articles
11
From troubleshooting guides & process flows
Resolution Steps
147
Structured decision steps extracted
Categories
6
Broadband · TV · WiFi · Signal · Sound · Engineer
Co-Pilot Ready
✓
Agent Assist integration enabled
Broadband
4 articles
No Connection, Slow Speed, Unable to Browse — hub diagnostics, Service Checker flows
No Satellite Signal, dish alignment, power inserter checks, communal vs own dish flows
No SignalDishPower Inserter
Sound Issues
1 article
No sound on live TV, volume checks, software version, digital output, speaker system, cables
No SoundLive TV
Engineer Visit Process
1 article
End-to-end engineer visit process map — arrival, checks, ABC Q vs Glass/Stream, MDU handling
Process MapMDU
All Articles
11 total
Browse the complete knowledge base — all troubleshooting guides and process flows
Full Library
Articles (4)
Select an article
Click on an article from the list to view its structured troubleshooting steps, decision branches, and resolution paths.
Co-PilotLIVE QUERY
Co-Pilot
ABC Co-Pilot
Connected to KB
Click a topic chip or type a query to surface relevant SOPs, troubleshooting steps, and suggested responses from the knowledge base.
Agent AssistFEATURES
On-Screen Assist Button
Agent Assist button on-screen during live customer interactions. One click surfaces the relevant troubleshooting flow based on the customer's described issue.
Step-by-Step Guidance
Shows structured resolution steps to agents when they query a specific topic. Decision branches adapt in real time as the agent progresses through troubleshooting.
Suggested Responses
Shows how to respond to the customer in real time — phrasing guidance, empathy cues, and next-best-action suggestions drawn from the knowledge base.
Source Attribution
Every response shows where it was sourced from — SOP document, troubleshooting guide, process flow PDF, or previous call interactions — building agent trust and transparency.
Active failure, no documented fix path. Typically engineer no-show, fault-rectified-pre-visit, or premature booking.
10
SOP Gap
Scenario exists in real life, no written procedure. Generic 'connectivity' codes, communal-flat handling, repeat-NSS auto-flag.
21
Training Gap
Solution is in the SOP, but agents are not implementing it. Service Checker non-use, OneHelp bypass, My Photo skip.
4
Product Gap
Customer asks for something the platform can't deliver — communal-dish handling, building-management workflow, no pre-visit auto-validation.
All (42)
Critical (14)
Process Gap (22)
SOP Gap (10)
Training Gap (21)
Product Gap (4)
Domain × Flag Matrix
ABC Mobile Tech-Visit ICF · v0.4 · From insight to intervention
We see the 12,198
No-Fault-Found visits. Here is how Kairos OS removes them.
The Volume, Graph and Gap Register tabs proved where the value leaks: 76.4% NFF in the engineer corpus,
Service Checker referenced in 3.1% of bookings, a 29.4% repeat-call rate. This tab is the working demonstration of how
Firstsource's Kairos OS — Domain Harness, Intelligent Context Framework (ICF) and Operations Intelligence sensors —
converts those findings into four live interventions on the booking journey.
Powered by Kairos OS · Intelligence that Operates
PILLAR · 00
Channel Strategy Redesign
The four-pillar AI thesis that frames every intervention on the engineer-booking journey.
50–60% AI-handled interactions
PILLAR · 01
Proactive Engagement
Detect friction signals on the line and at the ABC Hub. Notify, nudge and self-serve before a call is ever placed.
10–15% contact reduction
PILLAR · 02
Smart Routing & AI Concierge
A digital front door that classifies intent, complexity and vulnerability — and lands the right interaction in the right channel.
40%+ fewer misroutes
PILLAR · 03
AI-Enabled Agent Efficiency
Co-Pilot, Agent Assist and post-call automation. Service Checker pulled to screen. Notes written automatically.
25–40% AHT reduction
The thesis
Right channel, right time, every time.
The ABC CX approach calls for a transformative model with four strategic pillars. We have mapped each pillar to a specific failure mode from
the Tech-Visit ICF — and to a Kairos OS layer that executes the fix. Every intervention shown in the next tabs is a working demonstration of how it plays
on a live booking journey.
Today's economics
A short summary of what the corpus shows — and what each pillar is targeting.
76.4%
Engineer visits return No-Fault-Found. 12,198 of 15,967 dispatches with no engineer-actionable fault.
3.1%
Bookings that reference the canonical Service Checker tool — the single biggest dispatch control.
29.4%
Repeat-call rate inside 30 days post-booking — friction, not progress.
3,381
"Customer changed their mind" cancellations — booking conversations that don't survive customer reflection.
Four pillars · Four Kairos layers
Each pillar is anchored to a Kairos dimension. Click through the sub-nav to see each one demonstrated on an ABC Mobile booking scenario.
01
Proactive Engagement
Operations Intelligence sensors watch network health, ABC Hub diagnostics and call-history signals — and trigger prevention before contact.
02
Smart Routing & Concierge
The ICF + Domain Harness classifies intent, scores complexity, flags vulnerability and routes to the right lane — with full context preserved.
03
AI Co-Pilot for Agents
Skills from the Domain Harness surface Service Checker, NBA, sentiment and ACW automation in-flight. The 3.1% Service Checker problem becomes a 100% one.
04
Agentic AI NFF Risk Score
An autonomous agent runs the dispatch decision through Harness playbooks under Governed Autonomy — HITL today, HOTL tomorrow.
Pillar 01 · Proactive Engagement
Stop the call before it starts.
Operations Intelligence sensors watch four signal streams — network health, ABC Hub telemetry, app usage and call history. When a deviation pattern matches a known prevention recipe in the
Domain Harness, the ICF holds the customer context and the orchestrator fires a personalised nudge. Press Run scenario and watch an ABC broadband-instability moment play through.
Customer experience · ABC app + SMSLive demo
09:41● ● ● ABC 5G
Watch the customer's phone — no call placed.
Kairos OS · Operations Intelligence sensorsPrevention mode
Sensor · Network
T+00:00
Local CMTS sees 23% packet loss on the customer's port for >6 minutes. Pattern matches "intermittent line drop" recipe in the ABC-Connectivity Harness.
Sensor · ABC Hub telemetry
T+00:08
Hub log shows 3 micro-reboots in the last hour. Wi-Fi RSSI on connected devices stable — fault is upstream.
ICF · Entity context
T+00:12
Account flagged: 1 NSS-Permanent visit closed NFF in the last 60 days. Customer is in the 29.4% repeat-call cohort — high priority to intercept.
Self-serve diagnostic completes successfully. Line clears after auto-reseat. No engineer dispatched. ICF logs the decision trace for compounding learning.
Why this lands in the ABC corpus
The booking funnel today does not have a prevention step — every reported symptom defaults toward a visit. Proactive engagement reframes the funnel: detect → notify → self-serve → only book if all three fail. Even at the ABC CX deck's lower bound of 10% contact reduction, that's an estimated ~7,000 deflectable visits against the 70,453 corpus.
Pillar 02 · Smart Routing & AI Concierge
The digital front door.
An incoming customer intent — voice, chat or app — lands at an AI Concierge. The Concierge calls Domain Harness skills to classify intent, score complexity, check ICF for entity context (open faults, vulnerability flags, prior NFF) and pick the lane that maximises first-time resolution. Vulnerable customers are always a human-first lane.
Low complexity→ Self-service
Medium complexity→ AI chat + handoff
High / Priority→ Specialist agent
Each intent is classified by Harness skills · routed to the lane that maximises first-time resolution.
Why this lands in the ABC corpus
Today, the lack of a structured triage front door is what produces the 231 "I'm not trained for this job" not-done codes and a large slice of the 3,381 "changed-mind" cancellations. The Concierge applies the canonical Service Checker logic as a default — converting today's 3.1% Service Checker usage into a 100% gate.
Pillar 03 · AI Co-Pilot for Agents
Service Checker, on every call.
When a call does land with a human advisor, Co-Pilot listens to the conversation, runs Domain Harness skills in the background, and surfaces the next-best-action in-flight — Service Checker auto-launched, knowledge surfaced, sentiment scored, ACW written. The single biggest control in the booking funnel is no longer optional.
Click play to listen to a 12-line booking call as Co-Pilot intercepts in real time.
Why this lands in the ABC corpus
Today's gap register shows Service Checker referenced in 3.1% of booking calls. With Co-Pilot the tool launches automatically from the customer-intent signal — Service Checker becomes a system behaviour, not a habit. The ABC CX deck projects 25–40% AHT reduction and 30% FCR improvement from this layer alone. ACW notes write themselves to Sirius (MHR notation), removing a documented post-call cost.
Pillar 04 · Agentic AI
A reasoning agent decides the dispatch.
Agentic AI moves past suggestion to autonomous decision. The ABC-NFF-Risk agent is a Kairos pattern: brain (Harness playbook), hands (Service Checker, network probe, ICF read), memory (Decision Traces, Entity Context). It produces a Dispatch / Self-Serve / Diagnostic recommendation with a confidence score. Under Governed Autonomy, today it runs Human-in-the-Loop; on proven performance it earns HOTL/HATL.
ABC-NFF-Risk · v0.3 agenttrace-id 0xA47C·F19
Intelligent Context Framework (ICF) · live read
Booking intent
Customer reports TV signal loss · evening peak · 3 affected channels
Account · Entity
ABC Glass · UK · 4yr tenure · 1× NFF in last 90 days NFF repeat risk
Service Checker
Healthy No regional outage · transponder ok · signal-quality nominal
Network probe
UDP loss 0.4% (well below threshold) · last 30 min
Domain rule fired
Harness rule DH-TV-014 · "If Service Checker clean + repeat NFF history + symptom < 24h → run guided diagnostic before dispatch"
Probability the engineer visit returns NFF: 81%. Probability self-serve clears the symptom: 63%. Recommend a 6-minute guided diagnostic over ABC app and call back only if unresolved.
The most acute waste cluster — 97.3% of NSS-Permanent bookings return NFF (4,162 of 4,276) — is precisely the pattern the agent is designed to catch. Every dispatch is now scored, every decision is traceable, and every avoided visit feeds Compounding Intelligence back into the Domain Harness.
Kairos OS · Architectural map
Where each intervention lives.
Kairos OS is a five-layer stack. The middle three layers — Domain Harness, Intelligent Context Framework (ICF), and Operations Intelligence — are core to Kairos and travel across clients. The top and bottom layers are illustrative — varying by ABC Mobile's specific agent estate and systems of record. The four pillars above map cleanly into the stack.
Five-layer stack with client-context overlays
Illustrative for ABC Mobile engineer-booking journey · v0.4
Layer 5
AI-Native Operations · Agents
Illustrative — varies by client
Voice + chat AI agents, the AI Concierge, advisor Co-Pilot, the ABC-NFF-Risk reasoning agent. Stateless execution, durable context.
25 years of telco service operations encoded as skills, playbooks (PB-BB-022, DH-TV-014…), guardrails and eval suites. The dispatch logic the corpus exposed is now executable.
ABC overlay
Service Checker · NSS-Permanent rules · "not trained for this job" prevention · Vulnerability handling
Layer 3
Intelligent Context Framework (ICF)
Core Kairos OS
The memory behind every decision. Session context, entity context, decision traces, execution signals, domain benchmarks — the context layer this dashboard already exposes.
Core Kairos OS · Prevention · Detection · Learning
Process Mining · Task Mining · Conversation Intelligence · Knowledge Mining · Doc Intelligence. The "Service Checker referenced in 3.1% of calls" insight came from this layer.
⟲ Continuous Learning loop · every decision trace feeds back into the Harness and ICF
The Five Tenets of Intelligence that Operates
How each tenet shows up in the ABC Mobile engagement.
01
Domain Intelligence
25 years of telco service ops encoded as Harness skills — Service Checker is the day-one playbook, not a 6-month training problem.
02
Full-Stack Delivery
Design → Build → Operate as one motion. We don't hand the model to ABC Mobile and leave; we run the booking funnel with you.
03
Compounding Intelligence
Every NFF call, every Concierge route, every agent decision feeds ICF — the second 1,000 visits run smarter than the first.
04
Outcome Accountability
We underwrite the outcome — commercials linked to NFF reduction, FCR, AHT and contact-deflection, not effort.
05
Governed Autonomy
Trust is architected. The NFF-Risk agent starts HITL, earns HOTL on proven performance — per-skill, per-domain.
Outcome forecast
From today's economics to underwritten outcomes.
Each tile pairs a metric we observed in the ABC corpus with a target the four-pillar intervention is designed to land. The targets are anchored to ranges in the ABC CX Approach — Firstsource is willing to underwrite them under a Level-2 / Level-3 Kairos engagement, in line with Outcome Accountability.
No-Fault-Found visit rate
76.4% today
↓ Pillars 01 + 04
< 35% Yr 1 target
NFF-Risk agent intercepts every booking. Proactive engagement deflects symptoms upstream.
Service Checker reference rate
3.1% today
↓ Pillar 03
100% Day 1
Co-Pilot launches Service Checker automatically on every booking-intent call. System behaviour, not a habit.
Repeat-call rate (30d)
29.4% today
↓ Pillars 02 + 03
< 12% Yr 1 target
Smart Routing + Co-Pilot lift FCR by ~30%. ICF preserves context across channels — no "explain it again."
Unnecessary contact volume
— baseline
↓ Pillar 01
−10–15% Yr 1
Per ABC CX Approach. Sensors + nudges remove the conversation that wasn't ever needed.
Right channel, right time + Domain Harness skills land the resolution on first contact.
Misrouted inquiries
— baseline
↓ Pillar 02
−40%+ Yr 1
AI Concierge intent classification + vulnerability flagging. Specialist routing for high-priority and vulnerable customers.
"Changed mind" cancellations
3,381 annualised
↓ Pillars 01 + 02
−50% Yr 1 target
Booking conversations that don't survive customer reflection are pre-empted with proactive nudges and clearer self-serve paths.
The fastest path to outcomes · Level 1 Discovery
You've seen the architecture, the live demos and the engagement model. A single conversation turns this dashboard into a 90-day pilot.
Firstsource will run a fixed-fee, four-week Level-1 Discovery on the ABC Mobile booking funnel — and show you exactly where Kairos OS can change the economics before you commit to anything further.
Week 1–2
Diagnostic
Map AI maturity across the ABC Mobile engineer-booking journey. Validate this dashboard's gaps against your live operations data and CRM.
Week 3
Benchmark
Position ABC Mobile vs. comparable EU telcos on NFF, FCR, AHT and contact-deflection. Identify the highest-value pillar to lead with.
Week 4
Roadmap
Board-ready business case. 90-day pilot scope — one pillar, one process — with outcomes-linked commercials we are willing to underwrite.