LIVE WORKSHOP · HOSTED BY ALLAN SENDAGI
Make AI Work
Go from AI experiments to one AI task you can build next week.
Bring the process that eats your week. In 2.5 hours, find the work AI and agents should do in it, and leave with a one-page specification.
Join Cohort 21
LIVE WORKSHOP
2.5
HOURS
7
AI TASK ELEMENTS
1+
COMPLETED CANVAS
10
PARTICIPANTS
THIS IS FOR YOU IF...
What you walk
away with.
Not another list of AI ideas. One workflow mapped, and one task specified well enough to build.
Your workflow, mapped
Every step labelled: AI assists · AI performs · Agent runs it · Stays human · Not worth changing.
A one-page AI Task Specification
Precise enough to hand to a developer or vendor, without losing the business intent between strategy and build.
Agent design
Which steps an agent takes on, the systems it uses, its hand-offs, and what it must never do.
A first test to run
What to try in the next two weeks, with what data, and what to measure.
A measurable outcome
How you will know whether the intervention actually improves the work.
A verifiable credential
A certificate naming the specification you built, with its own page you can add to LinkedIn and anyone can verify.
THE AI TASK CANVAS
7 elements.
One session.
Every element forces one more decision you can't leave undefined — what AI's job is, what data it needs, where authority sits, and how you'll know it worked. You leave with all seven answered for your own task.
The Canvas combines the strategic and development decisions needed to define an AI intervention.
An AI idea describes a possibility. An AI Task Canvas specifies an intervention.
OWNER
Who owns the task and uses the output?
BASELINE
How does the task perform today?
TARGET
What needs to change?
ELEMENT 1
Action
What business activity are you improving?
e.g. "Send personalized follow-up messages after appointments."
ELEMENT 2
AI's Job
What should AI predict, classify, generate, or act on?
e.g. "Predict the likelihood a customer leaves feedback without a reminder."
ELEMENT 3
Judgment & Authority
What decision follows, and who or what is authorised to make it?
e.g. "Send a reminder if predicted feedback likelihood is under 40%."
Every condition falls into one of four classes: machine can check · machine can check with proof · needs a person’s judgment · needs more authority
ELEMENT 4
Input
What does AI need to see at the moment it acts?
e.g. "Customer ID, last visit date, service type, feedback history."
ELEMENT 5
Training Data & Context
What examples, documents, instructions, or context does AI need?
e.g. "Past appointments, whether feedback was left, reminders sent."
ELEMENT 6
Feedback & Record
How does the system learn what happened, and what record is kept?
e.g. "Track whether reminder recipients actually left feedback."
ELEMENT 7
Outcome
What measurable change tells you the intervention created value?
e.g. "Increase feedback completion rate without increasing reminder volume."
Based on the AI Canvas by Ajay Agrawal, Joshua Gans and Avi Goldfarb, Prediction Machines.
THE OUTPUT
One task. Seven decisions.
The AI Task Canvas turns a vague AI opportunity into a specification for what the intervention should do, what it requires, and how its value will be measured.
This is the artifact you leave with.
START WITH AN AI IDEA
“Use AI to improve patient follow-up.”
FROM HIGH POTENTIAL TO HIGH CLARITY
You start with a business task. The Canvas forces the decisions that turn it into something that can actually be built, tested, and measured.
STRATEGY ↔ AI TASK CANVAS ↔ DEVELOPMENT
SESSION BREAKDOWN
COHORT 2 PRICING
Cohort 2 of AI Value / Sandbox
10 Seats · 1 Working Session
$275
AED 1,000 · QAR 990
Saturday, October 24 · 10:00 AM Doha (GMT+3) / 11:00 AM Dubai (GMT+4) · 2.5 hours · Live online
Limited to 10 seats so I can work on your actual workflow with you. This is a working session, not a webinar.
- →Live 2.5-hour working session
- →Your workflow mapped and labelled
- →Completed AI Task Canvas
- →One-page AI Task Specification
- →Agent design: steps, systems and limits
- →A first test to run
- →Direct working feedback
Full refund if you don't leave with a specification you'd use.
Pay by card, or ask for an invoice.
What you'll need: One workflow to work on. No technical background required.
YOUR HOST & INSTRUCTOR
Allan Sendagi
AUTHOR, THE AI ROADMAP · FOUNDER, SAFEHAVEN AI · BUILDER, NOMOS PROTOCOL
Allan Sendagi is the author of The AI Roadmap: Implement AI Profitably in 10 Steps and creator of the AI Navigator System, the methodology behind the AI Task Canvas used in this workshop.
He founded SafeHaven AI and co-founded Shapr, an applied AI agency based in Dubai. At SafeHaven, he built NOMOS Protocol, an infrastructure specification for machine-verifiable institutional authority (how an organisation can prove an AI acted within the authority it was given), and developed Computable Authority, a proposed runtime architecture for binding institutional authority to machine-executed action, put forward through the OECD's 2026 public consultation on Law as Code.
This workshop is the front end of that work. Before AI can be trusted with a decision, someone has to define exactly what it should decide, with what information, and how its results will be judged. That's what you'll do here, for one task of your own.
WHAT OTHERS SEE IN THE FRAMEWORK
“A clear, real-world framework that bridges strategy and execution.”
Akmaral Shamenova
Operations Leader (20 yrs); M.Sc. Blockchain & Digital Currencies
“Allan Sendagi delivers a clear, real-world framework for AI implementation that bridges the gap between strategy and execution. Highly recommended for leaders ready to move beyond the hype and deploy AI with precision.”
On The AI Roadmap — the book behind this workshop's framework.
THE PROMISE
You are not buying 2.5 hours.
You leave knowing exactly where AI belongs in one real workflow: what it should do, where it stops, and how you'll know it worked.
YOUR QUESTIONS
01
Do I need to be technical to do this?
No. This is not a coding workshop and you do not need to know how to build AI systems. You bring a real business task and work through the AI Task Canvas with guided instruction. The goal is to define the work clearly enough that the technical implementation becomes easier to understand, evaluate, and build.
02
How is this different from asking ChatGPT to plan my AI project?
ChatGPT can generate strategies, ideas, and recommendations. But a generated answer is not a specification of what your organisation intends to change.
The AI Task Canvas forces your organisation to make the decisions that matter: what task is being changed, what AI's job is, what judgment follows, what information, training data, and context are required, how the system receives feedback, and what business outcome defines success.
The result is specific to your task, your workflow, your decisions, your data, and your definition of value — not a generic answer that could be given to another organisation.
ChatGPT can help generate the ideas. The Canvas makes the organisation specify what it actually intends to build.
03
What will I actually have at the end?
Your workflow mapped with every step labelled, one completed AI Task Canvas for the task you picked, and a one-page AI Task Specification you can hand to a developer or vendor. You also leave with the agent design for your workflow and a first test to run in the next two weeks.
04
What if I don't have a specific AI idea yet?
That's fine. You do not need to arrive with a fully formed AI use case. You need a real business task that could be improved. We will help you identify and sharpen the task before working through the Canvas. Not sure which process to bring? Try the 5-minute Pick Your Process exercise →
05
What should I bring?
Bring one workflow you are responsible for: one that eats your team's time. After you register you will get a short pre-work email to help you list its steps (6–10 steps). You do not need to prepare a technical specification.
06
Is the Canvas a strategy tool or a technical specification?
Both.
The AI Task Canvas combines the strategic and development decisions needed to define an AI intervention. It connects the business task and the value it should create with AI's job, judgment & authority, inputs, training data & context, feedback & record, and outcomes needed to develop and evaluate it.
You are not creating a strategy document and then translating it into a technical specification. The Canvas does both in one framework.
07
I want AI across my whole business. Will this help?
Yes. We map one workflow and specify the highest-value task in it. The same method then applies to the rest of the business, one workflow at a time.
08
Does this cover AI agents?
Yes. Every step of your workflow is checked for agent work, and you leave with the agent's steps, systems, hand-offs, and what it must never do.
09
How do I pay?
By card, or ask for an invoice. Price is $275 per seat (AED 1,000 or QAR 990).
10
What if it is not useful?
Full refund if you don't leave with a specification you'd use.
You walked in with a workflow.
Leave with one task you can actually build.
From there, you can test it, hand it to your developer next week, brief a vendor, or repeat the method on the next workflow.
I Want to Make AI Work10 seats · Cohort 2 · Saturday 24 October
© 2026 SafeHaven LLC · Lusail Boulevard. All rights reserved. · Terms · Privacy · AI Readiness Quiz · Pick Your Process · Verify a credential · Training for Teams