Four weeks that take a team from "AI is a black box" to a working AI workflow of their own, with measured ROI. Two weeks of short online classes build real understanding; two weeks of live coaching turn it into shipped work.
The shape of the program
The program runs in two deliberately different formats. Understanding doesn't need an hour a day: it needs a clear idea, a demonstration, and a check that it landed. Implementation is the opposite: it needs time, real work, and a coach in the room.
One 20-minute class per day. Each class is the same three moves:
Doable beside a full workload. By Day 10, every participant submits a structured capstone brief: a real workflow from their own job, ready to build.
One 60-minute live session per day, roughly 70% hands-on:
Every participant leaves with a working workflow, documentation a teammate could run it from, and before-and-after ROI numbers.
Curriculum · Part I
What a language model actually does: pattern-completion, not a database, a search engine, or a logic engine. A live demo of confidently wrong answers next to nailed ones. This mental model is the spine of the whole course, and every later class calls back to it.
Language models, image generation, speech, computer vision, and classic machine learning: what each one is for, and how they differ. Ends the "AI = ChatGPT" confusion on day two.
Training data, tokens, context windows, and knowledge cutoffs. Why a model doesn't remember you between chats. Fine-tuning and RAG in one sentence each. The mechanics, without the math.
Real deployments, from Nobel-level science down to "we killed a spreadsheet": AlphaFold and drug discovery, materials science, legal review, radiology triage, customer service, code generation, and Aslan's own client projects.
Hallucination in depth. Bias, data privacy, and what should never be pasted into a public model. AI-enabled phishing and scams. The common myths, busted. The goal is balanced skepticism: neither hype nor dismissal.
OpenAI, Anthropic, Google, Microsoft Copilot, Meta and open-source, plus the specialists: Perplexity, Midjourney, ElevenLabs. Strengths, character, and specialties, with the pricing tiers, the "your data trains our model" fine print, and the shadow-IT warning.
A decision framework: the task, the data sensitivity, the integration needs, the budget. Scenario walkthroughs, including when "data can't leave the building" points to a local or open-source model.
The same prompts run across two or three providers, side by side. Personality and strength differences made visible, and proof that "which AI is best" has no single answer.
Make the model ask you clarifying questions. Have it critique its own answer. Role and audience framing. Why long chats degrade and when to start fresh. Ask for ten options. Paste your writing to match your voice. The "make it worse first" trick. Voice dictation for richer prompts. AI writing prompts for AI. When regenerating beats arguing.
Why every company needs an AI use policy now, and what it covers: approved tools, prohibited data, review requirements, disclosure. Accountability stays human: whoever ships it owns it. Includes a one-page acceptable-use policy template. Then every participant submits a structured project brief: the workflow to transform, its current time cost, tool choice, data sensitivity, and success metrics.
Curriculum · Part II
Each capstone brief is reviewed one-on-one: scope pressure-tested, data-sensitivity issues flagged, and a realistic Day 20 target set. Re-scoping happens now, not on Day 17.
A 15-minute group opener on a shared lesson from the previous day's sessions, then hands-on building of prompts, templates, and workflows on real work, with coaching.
Same structure. Peer pairing begins: stronger builders are matched with the ones who are stuck.
Same structure. By end of day, every participant has a functioning workflow, even if it's rough.
Five-minute work-in-progress demos from every participant, with group feedback. Stalled projects surface here, with time left to recover.
Editing AI output to sound human. Verification habits. Catching errors before anything goes external.
Converting what was built into reusable templates, and handling the inputs that break the workflow.
Writing the workflow up so a teammate could run it. A personal trick becomes a company asset.
Final polish, then each participant calculates before-and-after numbers against their Day 10 brief: time saved per week, error reduction. The ROI is locked in before demo day.
Every participant presents a working workflow and its ROI to the group, with leadership invited in. Projects sort naturally into "solved with off-the-shelf AI" and "this needs custom software."
Program mechanics
The literacy classes are 20 minutes: video, lesson, quiz. They fit beside a full workload, which is what makes daily completion realistic.
The implementation sessions are 60 minutes and roughly 70% hands-on, built on each participant's actual job. No toy exercises.
From Day 1, every participant logs each task AI helped with and the minutes saved. By demo day the ROI math has been accumulating for a month.
The coaching weeks are capped so every project gets individual attention, and peer pairing keeps nobody stuck alone.
An optional check-in after the program ends, timed to catch the post-program adoption dip before new habits fade.
Every capstone is scored against the success metrics its owner committed to on Day 10. Results are numbers, not impressions.
Outcomes
Delivered by Aslan Advisors. When a demo-day project lands in the "this needs custom software" pile, we build that too.
Train your team