Alex Barády

How to explain agentic AI to your leadership.

Agentic AI isn’t one single tool.

Think of AI as layers of capability.

Each layer builds on the last.

Each requires different infrastructure, skills, and governance.

Here's the breakdown:

1️⃣ AI & Machine Learning

The foundational systems that learn from data, make predictions, optimise processes.

2️⃣ Deep Neural Networks

Pattern recognition at scale: vision, speech, complex data relationships.

3️⃣ Generative AI

AI that creates text, code, images, and audio. This is where most companies are today.

4️⃣ AI Agents Systems with memory and planning

They break down tasks, use tools, maintain context.

5️⃣ Agentic AI

Networks of agents that collaborate autonomously.

They plan long-term, coordinate with each other, self-evaluate, and improve over time.

The difference between Generative AI and Agentic AI?

Generative AI → Responds to prompts.
AI Agents → Execute multi-step tasks with tools.
Agentic AI → Operates with minimal supervision.

Where is your company on this stack?

♻️ Repost to help your network.
Follow Alex Barády for practical AI insights.

How to Use 6 AI Frameworks

to Transform One Critical Process.

Credit to Timothy Timur for collecting the frameworks.
Link to the original post (with better picture quality)
https://lnkd.in/d3J6p653

AI transformation works best when you start small.

Pick one critical process. Use these 6 frameworks to analyze it from different angles.

Each framework asks different questions.

Here's what they focus on and when to use them:

𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁'𝘀 𝗠𝗮𝘁𝘂𝗿𝗶𝘁𝘆 𝗠𝗼𝗱𝗲𝗹
Focus: Evolution over time.
Use when: Planning 6-18 month roadmap.
Question: How does human involvement decrease over time?

Example:
- Today AI flags issues.
- In 6 months it suggests fixes.
- In 12 months it applies fixes automatically.

𝗣𝘄𝗖'𝘀 𝗖𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻 𝗦𝗽𝗲𝗰𝘁𝗿𝘂𝗺
Focus: Role definition at each step.
Use when: Different steps need different AI involvement.
Question: Should AI advise, assist, execute, or decide?

Example:
- Step 1 - AI advises.
- Step 2 - AI executes.
- Step 3 - Human decides.

𝗗𝗲𝗹𝗼𝗶𝘁𝘁𝗲'𝘀 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸
Focus: Type of AI contribution.
Use when: Categorizing what AI should do.
Question: Automate, augment, or amplify?

Example:
- Automate data entry.
- Augment risk assessment.
- Amplify team capacity.

𝗚𝗮𝗿𝘁𝗻𝗲𝗿'𝘀 𝗔𝘂𝘁𝗼𝗻𝗼𝗺𝘆 𝗟𝗲𝘃𝗲𝗹𝘀
Focus: Degree of AI independence.
Use when: Setting supervision requirements.
Question: What can AI do without human intervention?

Example:
- Manual approval for high-value items.
- AI handles routine items autonomously.

𝗠𝗜𝗧'𝘀 𝗛𝘂𝗺𝗮𝗻-𝗶𝗻-𝘁𝗵𝗲-𝗟𝗼𝗼𝗽
Focus: Control and override mechanisms.
Use when: Legal, safety, or quality requirements exist.
Question: Where must humans review or intervene?

Example:
- AI processes 95% of cases.
- Humans review edge cases and can override any decision.

𝗛𝗕𝗥'𝘀 𝗧𝗲𝗮𝗺𝗶𝗻𝗴 𝗠𝗼𝗱𝗲𝗹
Focus: The human-AI relationship.
Use when: Redesigning daily workflows.
Question: Is AI a tool, teammate, or manager?

Example:
- AI provides data (tool).
- AI drafts responses (teammate).
- AI routes tasks (manager).

𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻: Pick one process tomorrow.
Apply each framework in sequence:
→ Map maturity trajectory (Microsoft)
→ Define roles at each step (PwC)
→ Categorize the work type (Deloitte)
→ Set autonomy boundaries (Gartner)
→ Design oversight points (MIT)
→ Clarify relationships (HBR)

This takes 60-90 minutes. You'll find gaps and conflicts in your current approach.

Different frameworks reveal different insights.
Use all six to stress-test your thinking.

Which framework do you find most useful?

♻️ Repost to help your network.
Follow Alex Barády for practical AI insights.

How to explain AI coding to your leadership.

The wrong AI approach leads to failed projects.
(yet happens all the time)

For example, you can't ship an AI-generated demo.
Even if the CEO wants it in production immediately.

There are three distinct approaches.
Knowing which to use separates success from failure.

Here's the breakdown:

1/ 𝗩𝗶𝗯𝗲 𝗖𝗼𝗱𝗶𝗻𝗴
Think of it as "code without coding."⁣⁣
You describe what you want in plain English.⁣⁣
AI writes the code.⁣⁣

When to use:
→ Testing product ideas before major investment
→ Demos for stakeholders
→ Business teams prototyping without developers

Don't use for:
Production systems or mission-critical applications.

ROI: Speed over precision.
Tools: Lovable, Bolt, Replit

2/ 𝗔𝗜-𝗔𝘀𝘀𝗶𝘀𝘁𝗲𝗱 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁
Developer productivity with complete control.
AI translates requirements into working code.

When to use:
→ Daily development work
→ Writing repetitive code
→ Improving code quality

Don't use for:
Replacing skilled developers.

ROI: 20-25% productivity while maintaining quality.
Tools: Windsurf, Cursor

3/ 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁
Autonomous execution of complex tasks.
You define the desired outcome.
AI plans, codes, and tests.

When to use:
→ Migrating legacy systems
→ Large-scale code updates
→ Multi-step development work

Don't use for:
Vague requirements or projects needing oversight.

ROI: Complete in days what takes weeks of manual development.
Tools: Claude Code

The best product teams don't use one approach.

They use all three, but for different situations.

Here is how to explain which approach to choose:

Vibe Coding → Validate ideas before investing.
AI-Assisted → Accelerate existing talent and teams.
Agentic → Delegate well-defined code migration.

Which approach haven't you tried yet?

♻️ Repost to help your network.
Follow Alex Barády for practical AI insights.