Agentic AI Overlay
Purpose-built agents.
Grounded by AKE.
The Agentic Overlay is a fully fledged AI agent with the power of AKE beneath it. Tailored intelligence per client, stable knowledge foundation underneath. Every fact cited. Every decision traceable.
How It Works
Tailored intelligence over a stable system.
AKE provides cited, environment-aware facts through three deterministic query types. The Agentic Overlay applies domain reasoning, physics, and UX on top. Hover any component to see the connection.
Equipment specification libraries
Technical standards and datasheets
Regulatory membership and compliance lists
spec_query
cited specifications
rag_query
cited document pages
config_query
cited eligibility verdict
Tailored Intelligence
user agent
per-client UX and interaction
orchestrator
propose, verify, rank
domain solver
physics, constraints, domain logic
Stable System
AKE is shared, version-controlled, and domain-agnostic. It never changes per client.
Tailored Intelligence
The Agentic Overlay is purpose-built per domain. Its reasoning, UX, and logic are yours.
Two Layers
One ground of truth. Your intelligence on top.
Agentic Overlay: per client
Orchestrator · Domain reasoning · Physics solver · User-facing agent
Purpose-built for your domain. Calls AKE for all factual lookups. Never decides a fact: it receives them, with citation, and reasons on top.
AKE: Ground Layer
Bi-temporal knowledge graph · Deterministic solver · Citation store
Stable, shared, version-controlled. Every fact delivered with its exact source page. Deterministic and reproducible.
What You Get
Designed for accountability, not convenience.
Behaviour-driven development
Rules written in plain English, locked in as tests. The agent knows what it can do and refuses to exceed that boundary.
Architecture from first principles
State, memory, and tool boundaries explicitly modelled, not left to emergent behaviour.
Typed, bounded tool calls
All tool calls carry budgets and circuit breakers. The agent cannot silently exceed a cost limit.
Eval harness built in
Not bolted on after deployment. Evaluation is part of the system design.
Full decision traceability
Every decision is logged and replayable. All facts sourced from AKE, with citation.
Commitment
Bring the problem. We tailor the solution.
Proof of concept first
We demonstrate before you commit.
Every engagement starts with a working demonstration on your own documents. We agree on a specific cited-answer rate for your query types before any contract is signed. Straightforward lookup or multi-layer compliance agent, the architecture holds across the full range of complexity.
If we miss the bar
You owe nothing.
Not a marketing device. We state specific quality criteria before the pilot begins. If the system does not meet them in your domain, the pilot engagement is refunded in full. The confidence to make that offer is the signal.
Example
Solar inverter configuration.
The agent checks five physical and regulatory constraints. Each number is retrieved from the cited source page. It stops when it finds a binding violation.
“Verify DC string wiring for SMA STP15000TL-30 inverter with 20 × LONGi LR5-72HPH-555M modules proposed as 2 strings × 10 connected to MPPT-A only. Site: Riyadh, SA. Is this configuration safe to deploy?”
System Configuration
Status
NON-COMPLIANT
Binding failure: MPPT-A over-current (27.44 A > 18.0 A limit)
Remediation
Every number above was retrieved from the cited source page by AKE. The agent handled the physics. Neither step involved a model deciding a fact.