Agentic AI in the enterprise: three talks, one expert panel
Understand what agentic AI actually changes, how to choose a model and an infrastructure, and what audit will require. No jargon, concrete cases, sourced figures.
The event in 30 seconds
What you will see on 8 December: three talks, a demonstration and an expert panel.
Who this session is for
CIOs and IT management
Architecture, security, data governance and technical choices.
CEOs, CFOs and finance
Real cost, return on investment, budget arbitration.
Business owners
Operations, finance, HR, IT: where an agent moves a metric.
Auditors and compliance
Traceability, segregation of duties, evidence to produce.
What you will understand
What an agent really is
The difference with a chatbot, what it does, what it must never do alone.
Where it moves a metric
Five functions, five indicators: handling time, closing time, service level, time to hire, MTTR.
What it costs and requires
Three-year total cost, security controls, evidence for auditors.
The three themes
Agentic AI: what it is, and what it changes in five functions
We start from the basics, then cover five cases: CRM, finance, production, HR and IT. For each: what the agent does, the data it needs, the metric it moves, and a worked ROI example — assumptions shown, to be replaced by your own volumes.
- CRM — qualification and follow-up: handling time per lead
- Finance — reconciliation and reminders: closing time, error rate
- Production — forecasting and alerts: service level, inventory held
- HR — screening and onboarding: time to hire
- IT — tickets and runbooks: MTTR, backlog
On-premise or cloud, open source or proprietary: choose a model, not a fashion
A model is not a magic product: it is an engine with a size, a cost, a licence and constraints. We explain the terms, then compare.
- On-premise: sensitive data, high API cost, real IT capacity
- Cloud: fast to start, variable cost, dependency on connectivity
- Hybrid: the choice is made source by source
- Open source versus proprietary: licence, control, quality, support, lock-in
AI and agents: cybersecurity, audit and internal control
An agent that can write into a system is an attack surface. We show the real risks, then the controls that answer them.
- Four risks: prompt injection, tool abuse, exfiltration, over-privileged accounts
- What audit will ask: traceability, segregation of duties, model inventory
- Ten controls to require before letting an agent write
- The evidence to produce, including in a CEMAC/CIMA banking context
The agenda
The speakers
Destin TAMPOLLA
Engineer and IS architect, more than twenty years in architecture, cybersecurity and AI.
Guest expert — CIO
A chief information officer at a CEMAC bank. Confirmation in progress.
Guest expert — Audit and compliance
An auditor or compliance officer. Confirmation in progress.
Moderator
Runs the exchange, shares speaking time, cuts sales pitches.
Our method
Start from the basics
No technical prerequisite. Every term is defined before it is used.
One deliverable per talk
A five-practice grid, a decision grid, a ten-control checklist. Usable on Monday morning.
Demonstrate, do not promise
No invented figures. Quoted numbers are sourced and dated; ROI calculations show their assumptions.
Rules of engagement
No sales pitch, no client name without consent. The moderator may interrupt.
The forum: your expectations before the event
Each theme has its own discussion thread, open now. Your questions are read by the speakers before 8 December: they shape the examples, the panel answers and the demonstrations. Three threads, three themes — write in the one that concerns you.
Theme 1 — Use cases and blockers
Which process would you entrust to an agent? What blocks it today?
Discuss this theme on the forum → or write by e-mail (no account needed)Theme 2 — Models and infrastructure
Sovereignty, connectivity, GPU, operating budget: what are your constraints?
Discuss this theme on the forum → or write by e-mail (no account needed)Theme 3 — Security, audit, internal control
What evidence would your auditors require? Which controls already exist?
Discuss this theme on the forum → or write by e-mail (no account needed)What you take away
Three grids
Practices, decision, controls — downloadable right now.
1 · Use cases: value, data, risk (PDF)
Two white papers
In French and English, sent after the session.
The replay and the academy
The three talks as chapters, with a quiz, available after registration.
Frequently asked questions
Free registration, 90 minutes
Tuesday 8 December 2026, 3:00 pm to 4:30 pm, Douala time. Online, from your browser.