The 2-Day Applied AI Workshop
Day 1 teaches your team how to work with AI — prompting, iterating, challenging outputs, picking the right tool. Day 2 turns that skill into a real MVP: a RAG system, a fine-tuned model, an agent, or a WhatsApp lead engine, built on your data.
Workshop architecture
The workshop is sequenced the way the work actually goes: learn the tool, then ship something with it. Day 1 is the same for everyone — Day 2 is your chosen capstone.
Prompting, iteration, and the AI-powered tools that have actually changed how work gets done.
Pick one capstone track. Apply Day 1 skills to build a working prototype on your data.
Day 1 · AI Foundations
The same Day 1 for every team, regardless of industry or capstone. No ML theory, no code-heavy prerequisites — just the prompting, iteration, and tool-selection skills that decide whether AI is useful or a waste of credits.
Your team has used ChatGPT a few times, maybe tried a copilot, and still treats AI like a clever search box. The result is shallow outputs, wasted credits, and a quiet suspicion that the tools aren’t as useful as the demos suggest.
A working shared vocabulary for what AI can and can’t do — and a hands-on tour of the tools that have changed the most: Lovable, Genspark, Cursor, and the prompting patterns that actually move the needle on output quality.
A single prompt is rarely the answer. The first response is a draft, not a deliverable — and most teams stop too early, ship hallucinations, or give up because the model "didn’t understand" what they wanted.
A repeatable loop for turning rough prompts into usable outputs: draft, critique, refine, verify. Participants push back on the model, demand sources, and learn to spot when an answer sounds confident but is wrong.
Day 2 · Choose your capstone
These are the four builds we’ve shipped the most often. Pick one during scoping — Day 2 is built around it. If none fit, the Custom Integration tier takes the same skill and applies it to whatever you need.
Ground AI answers in your own documents — policies, SOPs, product catalogues — with a working retrieval pipeline that cites its sources.
When prompting hits its ceiling and your outputs still look wrong. Train a small open-source model on your examples and measure whether it actually beats the baseline.
Single prompts can’t run your processes. Real work is multi-step — look up, decide, act, check, escalate — and a working agent turns one prompt into a real automation.
Your customers already live on WhatsApp, but enquiries arrive at all hours and qualification depends on whoever happens to be holding the phone. Leads leak every week.
What each capstone delivers
Same principle as Day 1: the work has to ship. Every capstone ends Day 2 with a running MVP, code your team owns, and a roadmap to take it further.
Pick this track when your team keeps saying "the model doesn’t know our stuff" and your knowledge lives in PDFs, wikis, or shared drives.
A retrieval pipeline over your real documents: chunking, embeddings, vector search, and a grounded answer layer that cites where each claim came from. By the end of Day 2, you have a system your team can demo internally the following Monday.
Pick this track when prompting gets you 80% there and the last 20% is the same correction on every output — tone, format, or domain language the base model keeps missing.
A real fine-tuning run on a small open-source model: dataset preparation from your examples, training, and an evaluation set that proves whether the tuned model actually beats the prompted baseline. Measured, not vibes.
Pick this track when a real workflow in your team needs more than one step and you want automation that knows when to stop and ask a human.
A multi-step agent wired to the AI tools and patterns from Day 1: tool use, planning, retry logic, and human-in-the-loop checkpoints at the steps where mistakes are expensive.
Pick this track when inbound enquiries are coming through WhatsApp and the bottleneck is response speed or consistent qualification — not demand.
An end-to-end lead qualification flow on WhatsApp: capture the enquiry, qualify it with an AI layer grounded in your offers, route hot leads to sales, and log everything for audit.
Who should attend
Mixed teams get the most out of the two days — the Day 1 prompts are accessible to anyone who writes, the Day 2 build needs engineers who can read code.
You need to separate feasible from hype, cost it honestly, and walk out with an architecture you can defend to the board.
You write code daily but haven’t shipped LLM features yet. Day 1 levels up how everyone uses AI; Day 2 gives you the build pattern to start.
You own the upskilling mandate and need training that produces visible, reportable outcomes — not another attendance certificate.
You own the workflows being automated — especially WhatsApp. Your input during scoping decides whether the engine qualifies real leads or imaginary ones.
Engagement options
Three ways to work with Byte Forge. Every engagement is confirmed with a written quote after a scoping call — no surprises.
Join a scheduled public workshop cohort
Delivered at your premises, scoped to your systems
Workshop, then we build the real system with you
// indicative pricing — final quote follows the scoping call, in writing
The workshop is split into two parts. Day 1 is the same for every team and covers how to work with AI — prompting, iteration, challenging model outputs, and using tools like Lovable, Genspark, and Cursor. Day 2 applies those skills to a chosen capstone build: a RAG knowledge system, a fine-tuned model, an agentic workflow, or a WhatsApp lead qualification engine.
No prior machine-learning experience is required. Engineers should be comfortable reading code; non-technical participants focus on the prompting and evaluation work that decides whether AI outputs are usable. Every capstone ships as a working MVP, so all four roles leave with evidence they can act on.
Before day one
The two days land harder when the exercises are built from your reality. Here’s what happens between “yes” and the first morning.
Thirty minutes on WhatsApp or a call. We learn your team composition, your industry, and what’s driving the AI push — budget pressure, competition, or a board mandate.
A lightweight review of what you already have: document stores, CRM, WhatsApp Business setup, cloud accounts. We identify what’s usable on day one and what needs prep.
Together we pick the capstone track — RAG, fine-tuning, agent, or WhatsApp lead engine — based on business value and achievability inside a two-day build.
We reweight the Day 1 prompts and exercises around your industry. Heavy on customer enquiries? WhatsApp-leaning examples throughout Day 1. Mostly internal docs? RAG-leaning.
One week before, every participant gets a setup checklist — accounts, tooling, and sample data prepared so the first morning starts with building, not installing.
Message us your team size and your target month — we’ll reply with available dates and a scoped quote within two working days.
// replies within working hours · no obligation · scoping call is free
