Skip to main content
UlexIoTy
Conduitby UlexIoTy
Features
OT Engineers
Query data across historians
IT Directors
Security-first data access
Plant Managers
Real-time operational KPIs
Division Directors
Multi-facility visibility
Routing Intelligence
AI-learned decision routing
All Solutions
View all roles
Use Cases
Blog
Insights and tutorials
ROI Calculator
Calculate your savings
Glossary
Industrial data terminology
ContactRequest Demo
Features
Use Cases
ContactRequest Demo

Footer

UlexIoTy

Conduit — Industrial Context Mesh

The Industrial Context Mesh that adds meaning to your OT data without moving it.

Meaning without movement.

Product

  • Features
  • How It Works
  • Integrations

Resources

  • Use Cases

Company

  • About
  • Contact

Legal

  • Privacy
  • Terms

© 2026 UlexIoTy LLC. All rights reserved.

Routing Intelligence

Your Questions Get Lost.AI Learns to Route Them.

Questions die in email chains, meetings, and hallway asks. Middle management becomes a telephone game. Knowledge is trapped in individuals. Seshe's routing intelligence watches how questions flow, learns who actually solves them, and progressively takes over the routing.

Request DemoHow It Works

The Problem

Organizational Knowledge Has a Routing Problem

3.2 hrs
Average time to reach the right expert
Questions bounce through managers who forward without adding value
73%
Of routing is unnecessary hops
Middle layers act as telephone relays, adding delay without insight
1
Single point of failure per topic
Most teams rely on one go-to expert per domain — when they leave, knowledge leaves

How It Learns

Three Phases: Watch, Suggest, Route

The system starts in pure observation mode. No configuration, no domain setup, no expertise tagging. It learns YOUR organization.

01

The AI observes (Phase 2A)

Watch & Record

Every completed question chain teaches the system. Who asked? Who forwarded? Who actually answered? The AI builds a routing graph — a map of how expertise flows through your organization. Question clusters emerge organically from real usage, not from preconfigured categories. (99 tests)

  • Route recording tracks every interaction in the graph
  • Question clustering discovers patterns automatically (not hardcoded)
  • Pattern extraction identifies relay nodes and skip-level opportunities
  • Embedding service measures semantic similarity
Routing Graph Building
// Question clusters discovered
"Line 7 yield issues" — 23 questions
"CNC tooling maintenance" — 18 questions
"Resin supplier quality" — 12 questions
// Routing pattern detected
Director → VP → Team Lead → Bob (resolves)
Skip-level candidate: Director → Bob directly
02

The AI recommends (Phases 2B-2C)

Suggest & Learn

After enough data, the AI starts suggesting who to route questions to. It shows confidence scores and reasons. Every acceptance, override, and redirect teaches the system via the feedback handler. Multi-route detection handles compound questions with fan-out orchestration. (104 + 154 = 258 tests)

  • Confidence-adaptive routing suggests at 40%+ confidence
  • Routing suggestions prioritized with explanations
  • Feedback handler learns from overrides and redirects
  • Multi-route detector and fan-out orchestrator for compound questions
Exploration in Action
Best known route (85%)exploit
Bob Martinez — Process Engineering
14 resolved, 92% satisfaction
Exploration candidate (15%)explore
Carol Chen — Quality Engineering
Resolves adjacent cluster, chats with Bob about Line 7
03

The AI decides (Phases 2D-2E)

Route Autonomously

When confidence exceeds 75%, the AI routes autonomously — no human decision needed. Per-company config controls thresholds and NQE auto-execution. Organic expertise tracking and personnel lifecycle adapters handle role changes and departures automatically. Cross-pollination discovers new expertise paths. (101 + 120 = 221 tests)

  • Auto-routing at 75%+ confidence (configurable per company)
  • NQE auto-execution pulls Conduit data in parallel with routing
  • Organic expertise tracks passive signals from conversations
  • Personnel lifecycle adapts to new hires, departures, and role changes
Parallel Execution
Expert Route
Question sent to Bob with context from routing history
Data Query
NQE fires to Conduit: yield data for last 60 days
AI Synthesized Answer
Expert analysis + Conduit data combined into one response

Confidence-Adaptive

See How the AI Adapts

The same question, handled differently based on system maturity. Switch between confidence levels to see the AI's behavior evolve.

AI Routing Confidence15%
Live routing demo
D
DirectorWhy did yield drop 3% on Line 7?
AI
AI AssistantI want to make sure I route this to the right person. A 3% yield drop could be process-related, material-related, or equipment-related. Do you have a sense of which area changed recently?
D
DirectorWe switched resin suppliers last month.
AI
AI AssistantGot it — so this is likely a material/quality issue affecting the process. Who on your team would be best to look into this?

Every Interaction Teaches

Four Learning Signals

The system learns from formal routing decisions AND organic conversations. No explicit training required.

Acceptance

User accepts the AI's routing suggestion. The graph strengthens that routing path.

Override

User picks someone different. The AI analyzes why and adjusts — overrides teach more than acceptances.

Redirect

Recipient says "not me, try Ben." The strongest learning signal — the AI immediately reroutes and remembers.

Organic Conversation

People chatting in threads. The AI passively extracts expertise signals — the lowest-friction, highest-volume source.

Resilient by Design

Handles Your Org's Lifecycle

People join, leave, change roles, and teams restructure. The routing intelligence adapts automatically.

New Hire

Starts with zero routing weight. System increases exploration ratio for their team's clusters. Within 2-4 weeks, the graph learns what they resolve.

Key Person Departure

AI identifies the closest backup for each question type. Exploration spikes temporarily. New routing paths stabilize within 2-4 weeks.

Role Change

Old expertise decays naturally via temporal decay. New expertise builds as they answer different question types. Transition is gradual.

Org Restructure

Routing edges are person-to-person, not position-to-position. They survive restructures. The system discovers new cross-team paths.

Measurable Impact

578
Total tests
across 5 phases
40%+
Suggest threshold
confidence band
75%+
Auto-route threshold
confidence band
18
Backend services
72k+ LOC, 4,310+ tests

Works for Any Industry

Unlike systems with hardcoded expertise domains, Seshe discovers YOUR organization's topics from real usage. No configuration, no domain setup.

Manufacturing
Energy
Pharma
Food & Bev
Mining
Automotive
Chemicals
Any Industry

Human Expertise + Machine Data

Seshe routes to the right person. Conduit pulls the right data. The AI combines both into one complete answer. This is what enterprise intelligence looks like.

Seshe

Routes to the right expert based on learned patterns. Skips unnecessary layers.

Conduit

Queries production data across all systems in natural language. No data movement.

Together

Expert interpretation + machine data = complete, data-backed answers in minutes.

Ready to stop losing questions in email chains?

See how routing intelligence transforms your organization's knowledge flow. Get a personalized demo.

Request DemoSee All Features