Get your assignments checked using BluePen
All presentations
AWS Student Builder Group UNISA — Agentic AI in Action90 min20 May 2026 · Online via Microsoft Teams

Agentic AI in Action — From Chatbots That Talk to Agents That Do

A 90-minute walkthrough of the AI landscape — AI vs ML vs Gen AI, how chatbots work, AI in everyday life, ethics, and the main event: agentic AI on AWS, with a live demo and 5 ways to use Claude in daily life.

Open deck in new tab← / → to navigate · S for speaker notes · F for fullscreen

Click into the deck above, then use arrow keys to step through slides.

agentic AIgenerative AIAI vs MLAWS AI servicesBedrockClaudecareerSouth Africa

TL;DR

A 90-minute tour of the AI world, designed for the AWS Student Builder Group at UNISA. We start at the foundations — what AI, ML, and Generative AI actually are, how chatbots work, and where AI already lives in your day — then spend the back half on the main event: agentic AI. The shift from AI that talks to AI that does, demoed live with an open-source agent that wields the AWS CLI.

We end with five ways to use Claude in daily life — practical habits anyone can adopt this week, whether you ever write code or not.

What you'll walk away with

  • A clean mental map of the AI landscape — no more confusion between AI, ML, and Generative AI.
  • A live demo of a real OSS agent (46K stars on GitHub) running a job-hunt loop on stage, and writing its output to AWS S3.
  • Five Claude habits to start tomorrow morning — no coding required.
  • An understanding of where AWS fits — Bedrock, AgentCore, Comprehend, Rekognition, Lex, Polly, Lambda + EventBridge — and which one is right for your project.
  • A career frame for SA devs: in a tight market, the composer outships the engineer.

The deck spine (20 slides, ~90 min)

# Slide Topic ~Time
1 Title "Agentic AI in Action" 1 min
2 Who I am Akhil intro · Zaio + ATB 2 min
Foundations (≈40 min)
3 AI vs ML vs Gen AI 3 nested circles · transportation / toddler / novelist 8 min
4 How AI learns from data Puppy training analogy · 4 steps 6 min
5 How chatbots work "Autocomplete on steroids" with a live prediction example 6 min
6 AI in everyday life Day-in-your-life grid: Face ID → Maps → Gmail → Spotify → fraud detection 4 min
7 Intro to Gen AI Librarian → novelist · the 2023 leap 4 min
8 Ethics & risks Hallucination · bias · displacement + mitigations 6 min
Main event — Agentic AI (≈30 min)
9 Section divider "From talks to does" 1 min
10 Chatbot vs Agent Intern vs employee analogy 4 min
11 Anatomy of an agent Brain · Body · Hands · Notebook + the loop 4 min
12 AWS AI services The toolbox: Bedrock, AgentCore, Lex, Comprehend, Rekognition, Polly, Transcribe, Translate, SageMaker, Lambda+EventBridge 5 min
13 Demo intro NousResearch/hermes-agent — #1 OSS agent on OpenRouter 1 min
14 LIVE DEMO Install Hermes → web_search task → plug in AWS MCP → S3 write 13 min
15 Recap What just happened, step by step 4 min
Take-home (≈12 min)
16 Five Ways to Use Claude Study buddy · Writing partner · Decision sounder · Code companion · Learning accelerator 5 min
17 Tonight in 30 min 3 actions: open Claude · read the repo · spin up AWS 3 min
18 Career angle for SA devs Tight budget = unfair advantage 3 min
19 CTA — comment BYOS IG funnel + socials + photo break 2 min
20 Close + Q&A "You're not behind, you're early." 8 min

The analogies (so they stick)

These are the cognitive hooks the deck leans on. Memorise the analogy, the concept follows.

Concept Analogy
AI vs ML vs Gen AI Nested circles. AI = "transportation" (the whole field). ML = a kind of AI (toddler learning what a "dog" is from examples). Gen AI = a kind of ML (the librarian who became a novelist).
How AI learns from data Training a service dog. No rulebook — 10,000 examples, reward what's right, correct what's wrong. The dog figures out the pattern. AI training is exactly that, at internet scale.
How chatbots work Predictive text on your phone — but the phone has read the entire internet. The chatbot doesn't "know" things; it predicts the most likely next word, one at a time.
AI in everyday life A day-in-your-life: Face ID at wake-up, Maps ETA on the commute, Gmail autocomplete at work, Spotify Made-for-You at gym, Netflix thumbnails at night, fraud detection while you sleep.
Generative AI The librarian (sorts and recommends) became the novelist (creates new work in any voice).
Ethics & risks AI is a confident friend who sometimes makes things up. Trust, then verify.
Chatbot vs Agent Chatbot = intern stuck in the room answering questions. Agent = junior employee who can actually GO DO things — open the laptop, send the email, file the report.
Anatomy of an agent Brain (model) · Body (harness) · Hands (tools) · Notebook (memory) → wrapped in a loop.
AWS AI services A toolbox. Bedrock = engine room. AgentCore = harness. Lex = chatbot template. Comprehend = reading robot. Rekognition = eyes. Polly = voice. Transcribe = hearing. Lambda+EventBridge = schedule.

The demo — Hermes + AWS MCP

We use the #1 open-source agent on OpenRouter — NousResearch/hermes-agent. MIT licensed. Free. Ships with 40+ built-in tools (web_search, browser, terminal, file ops, vision). Persistent memory. Native scheduling. MCP support.

The live flow (no agent code written by us):

  1. Install Hermes — one curl command, handles Python/Node/uv/everything
  2. Configure the model — hermes model → pick OpenRouter → pick openai/gpt-4o-mini
  3. Enable tools — hermes tools → web_search, terminal, file ops on
  4. First task — open hermes, type "Search the web for 3 entry-level SWE jobs at OfferZen Cape Town. Return as JSON." Watch Hermes use web_search, return structured JSON
  5. Plug in AWS MCP — edit ~/.hermes/config.yaml, add 4 lines: aws_api server pointing at uvx awslabs.aws-api-mcp-server@latest (AWS's official MCP, GA'd 6 May 2026)
  6. Reload + multi-step task — back in hermes, run /reload-mcp, then type "Create S3 bucket akhil-hermes-demo · web_search for 3 jobs · save list as JSON to s3://…" Watch the loop. Verify with aws s3 ls.
  7. Schedule path — hermes cron runs the same task at 6am daily

The point everyone should feel: we composed. Hermes existed. The AWS MCP server existed. Our job was 4 lines of YAML and 1 plain-English sentence.

Claude Code alternative: if you want a more turnkey job-hunt agent on the Claude Code side, santifer/career-ops is the analog (46K stars, MIT, 14 skill modes). Different harness, same BYOS idea.

Five ways to use Claude in daily life

The take-home slide — works whether you're a builder or not.

# Use When Sample prompt
1 Study Buddy Confused by anything "Explain Bayes' theorem twice — once like I'm 12, once for a stats exam."
2 Writing Partner Emails, applications, posts "Rewrite this email warmer but still professional. Match my style — here are 3 past emails."
3 Decision Sounder Stuck between options "I'm choosing between a SWE role at X and a data role at Y. Make the case for each in 200 words."
4 Code Companion Even non-devs "Write a Python script that renames every PDF in this folder using the date inside it."
5 Learning Accelerator Papers, lectures, study plans "Summarise this 40-page paper into 10 bullets + 5 quiz questions I can review tomorrow."

Tonight — three actions before you sleep

  1. Open Claude (5 min) — claude.ai, free tier works. Pick ONE of the 5 uses above. Paste a real thing from your day.
  2. Read the demo repo's README (10 min) — github.com/santifer/career-ops. Just see what's possible.
  3. Follow the step-by-step guide (45 min) — → Run Your First AI Agent — every install command, every AWS step, every fix. By the end of it you have the exact demo I ran in the talk, on your own machine, with your CV in your own S3 bucket. Comment RUNAGENT on my IG post for the link.

By tomorrow morning you have used an agent. Not theory — your own.

Why this matters for SA devs in 2026

In a tight market, the composer outships the engineer.

  • Tight budgets are your unfair advantage — open-source tools + AWS free tier = $0 to start. A Bay-Area startup burns that on their AI bill before lunch.
  • "Knows AWS + uses AI" is the 2026 SA dev premium — every recruiter values AWS, every agent loves AWS CLI. You become more leverageable in both directions.
  • Build agents for yourself first — job-hunt agent, study agent, side-hustle agent. Each one is portfolio + résumé + interview story in a single artifact.

Resources

Find me · ask me anything

Two ways in — pick one or both:

  • Comment BYOS on my latest Instagram post → I'll DM you the link to this exact page (deck + write-up)
  • Comment RUNAGENT → I'll DM you the step-by-step "Run Your First AI Agent" guide — every install command, every AWS step, every fix. 45 min start to finish.

Other places to find me: