One-day intensive
About six hours, on-site or remote
The course’s spine in a single day, ending with a working session on a workflow the team chose in advance.
Best for: Teams that can clear a day and want momentum.
What we see in most teams
Access is not ability.
- 01
Prompts typed like search queries. Three words in, a paragraph out, and the paragraph gets rewritten by hand.
- 02
Nobody is sure what is safe to paste. So people either paste everything or use the tool for nothing that matters.
- 03
One person on the team is good at it and cannot explain why. Their results do not transfer.
- 04
Usage plateaus at summarising emails. The tool could draft the board memo, check the spreadsheet logic, and review the contract clause. It is never asked.
How it's delivered
Two formats. One course.
Four weekly sessions
90 minutes a week, for four weeks
The course across four weeks. Between sessions, people apply it to real work and bring the results back for review.
Best for: Teams that want the habits to stick.
- Cohort
- 8 to 20 people
- Where
- On-site in the Greater Toronto Area, or remote
- Tools
- Whatever your company has approved: Microsoft Copilot, ChatGPT, Claude, Gemini
- You keep
- The playbook, the prompt and skill library your cohort builds in lessons 4 to 7, and the recordings if remote
Who it's for
People who write, decide, and analyse for a living.
- Operations
- Strategy
- Finance
- Marketing
- Human resources
- Legal operations
- Project management
- Customer experience
Whole departments, not only the technical ones. Analysts through directors; anyone whose week is documents, decisions, and email. It is not a course for engineers building models, and we will say so if that is what you need.
Who teaches it
We build AI systems for a living.
Arshveer Gahir Founder, lead facilitator
Trellis is a brand, digital, and AI studio in Toronto. We build AI automation for our clients and run our own studio on it, so the course is taught by people who use these tools every working day, not from a slide deck written last year.
We do not sell software and we are not tied to a vendor. If your company runs on Copilot, we teach Copilot. Where the tools differ in ways that matter to your work, we say so plainly.
Packs
Choose how far you go.
Every pack starts at lesson one; the foundations are where the gains come from. Stop at the depth your team needs, or go the whole way and leave with the decision made.
- All nine lessons Our recommendation
The full path
$12,500
The course is built to be finished. The decision in lessons 8 and 9 is what the foundations are for.
The whole course, through the Copilot-and-Claude decision and a before-and-after experiment your team runs that week.
Best for: Teams that also need the Copilot-or-Claude question settled.
Request this pack - Through lesson 7
Full craft
$10,000
Adds context engineering, skills, and advanced prompting: the complete desk, practised on your team’s real work.
Best for: Teams ready to build briefs and skills they reuse.
Request this pack - Through lesson 5
Working craft
$7,500
Foundations plus prompt engineering and agents: briefing the model well, and knowing what an agent is doing with your work.
Best for: Teams whose week is drafting, judging, and delegating documents.
Request this pack - Through lesson 3
Foundations
$5,000
The brain: what a model is, sizes and thinking effort, what it can see, remember, and make up. The shared ground truth a team needs before craft.
Best for: Teams that need shared ground truth before anything else.
Request this pack
Build, not only training
Harness development
We build what the course teaches: the prompts, skills, and agents for your team’s actual workflows, tested on live examples and handed over documented. The price is scoped by what you want built, typically four to six weeks per workflow.
from $15,000
Prices are per cohort of 8 to 20 people, in Canadian dollars, plus HST. Remote or on-site in the GTA. Either delivery format, same price. At a full cohort of 20, the whole course is $625 a person.
FAQ
Common
questions.
No. We teach in whatever your company has approved. The principles are the same across Copilot, ChatGPT, Claude, and Gemini. Where one tool does something the others cannot, we point it out and move on.
We ask for your acceptable-use and data-classification policies before the first session and teach inside them. No company data comes into the room unless you have cleared it. The lessons on what the model can see are taught inside what your policy allows, so people leave knowing the line instead of guessing at it. The course is built to be taught inside strict acceptable-use policies, the kind regulated employers actually have.
Eight to twenty per cohort. Smaller and the working sessions lose energy; larger and people stop getting their own work reviewed. Bigger teams run as several cohorts.
Yes. The weekly format works well remotely. The one-day intensive is better in a room, and we will tell you if we think your team should wait for one.
The close of the course always is: the final lessons build the harness for a workflow your team chose. If you want the earlier lessons to use your documents and examples as well, we do that with a short discovery call beforehand.
The playbook, first of all: the briefs, the context, and the rules your team wrote during the sessions, in their own words. In the weekly format, people apply each lesson to real work between sessions and bring the results back for review. The course ends with an experiment your team runs that week. And when you want what was learned built into the tools themselves, that is harness development.
The packs above show it: $5,000 to $12,500 per cohort of 8 to 20, in Canadian dollars, plus HST, the same price in either format. Harness development starts at $15,000, scoped by what you want built. Ask and we will send a proposal you can circulate internally.