Organizations have intelligence.Most of it is trapped.

Galucy Niels Enterprises
GIEoS · Google Africa Applied AI Lab
02The world changed
2022: 2026
Foundation models became
Multimodal text · image · audio · video
Agentic tool use
Reasoning-capable extended thinking
Connected APIs · protocols
Autonomous long-horizon tasks
Organizations still run on
Email
CRM
Tickets
Spreadsheets
Chat
Documents

AI got smarter. Organizations did not automatically become smarter.

03The fragmentation problem

Every system knows something. Nothing knows enough.

THE ORGANIZATION
04The insight

The next layer of AI is not another model.
It is the system around the models.

Foundation models provide intelligence. Organizations need a system that can:

Remember memory
Reason context
Coordinate orchestration
Act workflows
Learn feedback
Evolve architecture
05Introducing
Enterprise Intelligence Runtime
GIEoS
Galucyniels Intelligence Engine OS
The Enterprise Intelligence Runtime.

Where AI-powered applications share knowledge, orchestrate capabilities, reason across workflows and evolve through organizational experience.

runtime GIEoS
Built from Africa. Designed for organizations everywhere.
In development
06The architecture

Models below. Organizations above. GIEoS in between.

Applications Organizations GIEoS ENTERPRISE INTELLIGENCE RUNTIME MODEL-AGNOSTIC Foundation models
Swap the model. Keep the organization's intelligence.
● Gemini among other frontier, open-weight and fine-tuned models
07The intelligence loop

Intelligence should compound through experience.

01Understand
02Remember
03Reason
04Act
05Observe
06Learn
07Evolve

Most deployments stop at Act.

organizational EXPERIENCE
08Organizational memory · illustrative
One year
Day 1

A ticket arrives.

Resolved by one agent. The answer stays in one head.

#GC-1042open
Customer cannot receive OTP on a new SIM. Escalated, resolved manually in 41 min.
Day 30

Similar tickets accumulate.

Same root cause, different words. No one sees it as one thing.

#GC-1310closed
"No verification code after number change"
#GC-1287closed
"OTP not arriving, new line"
#GC-1251closed
"Code delay after SIM swap"
Day 90

GIEoS recognizes a pattern.

Clustered by meaning. A failure mode gets a name.

cluster: otp-after-sim-change
Day 180

A workflow is recommended.

A known pattern becomes a proposed path.

RECOMMENDATIONAuto-verify line status → re-issue OTP via alternate channel → confirm with customer.
Day 365

The learning is institutionalized.

New agents inherit it. The organization behaves differently.

#GC-4871auto-resolved
Matched known pattern. Resolved in 3 min without escalation.

Memory is more than history. Memory changes what the organization does next.

09Proof point · GACiL
Built · in use
GACiL · Service Deskqueue · 14 open
#2210Payment reversal not reflected on statementBillingP2
#2209Cannot log in after password resetAccessP1
#2208Request: add branch to delivery coverageRequestP3
#2207App crashes on document upload (Android)TechnicalP2
Assist#2209 matches 6 resolved tickets this month. Likely cause: cached session after reset. Route: Access → Tier 1. First response drafted.
CaptureEmail · web · WhatsApp · phone
ClassifyCategory · priority · intent
RouteRight team, context attached
AssistPast resolutions surfaced
ResolveHuman closes the loop
LearnResolution becomes knowledge

From ticket resolution to organizational learning.

10Proof point · AIleesh
Built · live demo

An interface for intelligence.

Not GIEoS itself: a live demonstration of the intelligence experience being explored.

Conversation multi-turn
Files documents
Vision images
Voice speech
Model routing per task
Reasoning step-wise
aileesh.galucyniels.com
AIleesh
New conversation
Q3 supplier review
Onboarding policy
Site photos: Kano
Model
auto · routed
⌗ supplier_contract_2026.pdf · 14 pages
Which penalty clauses conflict with our standard terms?
reasoning · 4 stepsrouted · long-context
Two of three conflict. §7.4 sets a 2% weekly late-delivery penalty; your template caps at 0.5%. §11.1 allows termination on 7 days' notice; you require 30. §7.2 is consistent.
Does this delivery note match the contract quantity?
visionrouted · multimodal
480 units on the note, 500 in the schedule. The 20-unit shortfall is within the §7.2 tolerance of 5%.
Ask anything, attach a file, or speak⏎ send · ◉ voice
11Proof point · Shona · WhatsApp Business

From conversation to action.

Adaeze O.customer · +234 ··· 4471
Good morning. I paid for order 5582 yesterday but my app still says pending. Can you check?
Good morning Adaeze. Your transfer for ₦42,500 was received at 16:48 yesterday. Order 5582 is confirmed and scheduled for dispatch today. You'll get a tracking link once it leaves the warehouse.
Thank you. Can I change the delivery address to my office?
Yes. Send the office address and I'll update it before dispatch.
01
MessageWhatsApp Business API
02
UnderstandingIntent: payment status · order 5582
03
KnowledgeOrder and payment records
04
ResponseDrafted in the business's voice
05
ActionAddress change raised for dispatch
Shona · Conversationassisted
IntentPayment status enquiry
Order5582 · confirmed · ₦42,500
PaymentBank transfer · received 16:48
ConfidenceHigh: two records matched
"Good morning Adaeze. Your transfer for ₦42,500 was received at 16:48 yesterday. Order 5582 is confirmed and scheduled for dispatch today."
SendEditCreate task · address change
12Built for African realities

A uniquely African problem, solved where it lives.

African organizations do not run on seat-licensed enterprise suites. They run on WhatsApp, USSD, phone calls and spreadsheets. An intelligence layer for this market has to meet organizations on the channels they already trust.

A system designed under these constraints travels well. The reverse is rarely true.

WhatsApp is the storefrontShona · built
USSD reaches every phoneHubFlow · built
Service arrives on every channelGACiL · built
Bandwidth and cost shape designModel routing
Data protection is lawNDPA-aligned
13From applications to platform

Six products. The same eight problems.

KnowledgeMemoryReasoningIdentityAutomationOrchestrationEvaluationIntegration

GIEoS is emerging as the shared layer. Not every product runs on it today.

14The evolution engine
Research direction · early implementation

The system should not only execute work. It should learn how the work itself can improve.

Today: pattern detection and recommendation, with humans approving change.

15Why the Applied AI Lab
An application, not an announcement
Google AI ORGANIZATIONAL INTELLIGENCE LAYER GIEoS memory · reasoning · orchestration · learning Organizational applications

Build the intelligence layer around the models.

The ask: early access to Gemini, pairing with Google experts, and the AICC cohort, to compress this research into product.

16What we want to explore

Nine questions the Applied AI Lab could help us answer faster.

A
Persistent organizational memoryUseful across years, not sessions.
B
Multi-agent orchestrationSpecialized agents, shared context, clear accountability.
C
Continuous evaluationIs the organization improving, not just the answer?
D
Organizational reasoningTickets, documents, records and people, together.
E
Capability discoveryThe runtime notices what it is being asked for.
F
Autonomous workflow improvementFrom recommendation to safe adoption.
G
Multimodal organizational intelligenceVoice, photos, scans and forms as knowledge.
H
Model routing & specializationThe right model per task.
I
Human-AI collaborationPeople correct, approve and teach as they work.
17The vision · one organization, five years
Illustrative

The goal is not an AI-powered organization. The goal is an organization that becomes more intelligent over time.

18The company

Galucy Niels Enterprises

AI-first technology company building enterprise intelligence systems.
Founded in Africa. Building for organizations everywhere.

Mission

Help organizations learn, reason, automate and continuously evolve.

Founder

Daniela Udunwa · solo founder and engineer, Nigeria. Designed, built and shipped every product in this deck; several are already offered commercially to Nigerian organizations. GIEoS grows out of live deployments, not a whiteboard.

GIEoSEnterprise Intelligence RuntimeIn development
AIleeshAI assistant and intelligence interfaceBuilt
GACiLIntelligent ticketing and service managementBuilt
ShonaWhatsApp Business automation and communicationBuilt
CiliaAI-powered job application and career assistantBuilt
HubFlowUSSD workflow and menu orchestrationBuilt
TheGoMulti-channel Meta webhook and integration middlewareBuilt
InsideSheeshKnowledge and organizational intelligence environmentEcosystem

We are not building another AI assistant.We are building the system that helps organizations become intelligent.

GIEoS
Galucyniels Intelligence Engine OS · Google Africa Applied AI Lab application
Website
galucyniels.com
Demo
aileesh.galucyniels.com
Galucy Niels · GIEoS