Management Memo to
Leverage AI for Cannabis
Edibles Product
Development
Student #: 996758450

RSM2008H-LEC0101
CDL Intro
April 9th, 2021
Ollibrands.com
[DOCUMENT TITLE]

1
Introduction
Olli Brands Inc. (OBi) is a premium cannabis edibles manufacturer who focuses on
creating quality and great tasting edible products. Our company’s core point of
differentiation is its focus on creating products with a food first mentality and
cannabis simply an ingredient. Management’s goal for 2021 is to maximize sales and
increase our product suite to 20 SKUs. In order to do this, we need to be a thought
leader with innovative products to build and maintain a strong market position.

Our product development and R&D team spend countless hours researching
product trends, consumer insights, flavours and innovative ingredients to keep up
with the ever-evolving pace of the cannabis industry while maintaining the
company’s core mission to create delicious tasting products that do not taste like
cannabis. This is a key piece to our differentiation strategy as our target market is not
a cannabis connoisseur but rather a new or renewed cannabis consumer.

This development process is both costly, causing management to have to choose a
limited selection of products to focus on, and risky, coming with a lot of uncertainty
given the regulations restrict us from easily testing the infused product in-house. It is
also incredibly time consuming often resulting in a lost opportunity to be first to
market and stay ahead of the trends. Given the time and there being no way for us to
test our products for taste aside from doing costly clinical trials or waiting until after
we have already launched a product, OBi can leverage AI to advance the product
development process and eliminate some of the uncertainties.
Summary of the AI Canvas
We can leverage AI to predict what type of ingredients and flavours to use that will
mask the cannabis taste and which types of products will be successful in the current
market. The double-action machine will only accept ingredients and flavours that
mask cannabis taste and from here, it will narrow down a list of potential products
even further by analyzing the probability of success of such product in the market.
The measure of success will be a product free of cannabis taste and that achieves a
ROI of at least 35% in year one of product sales (see Appendix II for ROI
calculation). The machine will be rewarded each time it successfully achieves the
desired outcome to help further its potential for future successes. We can input data
from various sources including consumer insight reports, cannabis palatability
studies, employee experiences, terpene profiles of food and cannabis, and market
trends to help train the machine.

We will not have to make any immediate changes to our workforce, and I foresee a
reduced need to hire as rapidly as planned for our product development team. This
will also allow our current team to become more experienced in building lab
[DOCUMENT TITLE]

2
prototypes and focusing on innovations in cannabis or processes by reducing the
need for countless hours of research of food ingredients and trends. A smaller team
will be able to accomplish producing more new products in a given year. Refer to
Appendix I for the AI Canvas and Appendix III for a diagram of how the machine
will work the to further describe the process.
Benefits
The cost savings for this type of investment will be massive. I would plan for a six-
month set-up, similar to our ERP implementation at which point we can begin to
use AI in the development process. Each time we develop a new product, we can
input more data on the outcome of each product to help it get better all the time.
There would be cost savings of ~$26,000.00 in the product development process by
reducing the labour hours required for market research and product testing,
increasing ROI by 35% (Appendix II). In addition to this, a reduction in labour
hours required in the development process by 52% (Appendix II) frees up time for
the product development team to focus on additional products and develop more
products in a given year. It will also allow the company to stay ahead of trends,
reduce uncertainties about cannabis taste in our finished product and avoid having to
test products in market only to find them not succeed. I have estimated this process
will reduce uncertainty and risk by 15% - 20% and we will save over $360,000.00
over a five-year period in failed product costs.
Risks
There are few risks that come along with this investment in AI. Firstly, we may not
be able to access enough data to train the machine given the newness of the cannabis
industry and publicly available data. This could restrict us in how quickly we can train
the machine and get to the level of accuracy I have predicted (95%). I advise we
allocate a $15,000 budget to acquire data from Headset, Business of Cannabis and
various retailers offering industry statistics for a fee. Secondly, the machine may
predict flavours that mask the cannabis taste but do not work well together or result
in a great tasting product. To mitigate this risk, we will have to test the flavours and
ingredients in a non-infused prototype before approving it for the machines second
action. Lastly, and most importantly, we may miss truly innovative opportunities
with new flavours or ingredients where data is not available to feed and train the
machine. Human judgement from both the management and product development
teams will continue to be crucial in making the final decision on product launches to
pursue.
[DOCUMENT TITLE]

3
Judgment
Determine the value of a false
positive (developing a product
that does not mask the cannabis
taste or that consumers like) vs. a
false negative (not using flavours
that would perfectly mask taste
and result in a product consumers
love).
Prediction
Predict flavours and ingredients
that will mask of the taste of
cannabis and result in a overall
taste consumers like (based on
consumer insights and success of
product in market).
Outcome
A product that does not have any
cannabis taste, tastes delicious
and achieves ROI of 35% in year 1
of product sale.
Action
Primary Action: Accept or reject a
product on the basis it masks
cannabis taste
Secondary Action: If it passes the
above test, choose the product that
will be a success in the market.
Training Input Feedback
How will this AI impact on the overall workflow?
Less research required by R&D team to determine popular flavours, food trends and flavours/ingredients that can mask cannabis taste
Less risk or assumption-based decision making oh what will work
Speed up the product development process and launch of new products
Requires less lower-level R&D employees (those who do the research) and more experienced in actual lab prototype development
Potentially seek employees who are more skilled in innovation of cannabis rather than food
Characteristics of a product that mask cannabis
taste (ingredients, flavours) and historical data
on what products consumers like (type of
product, number of units, flavours, quality, sales
in year 1, etc.)
Data from company’s prior products and other
industry products that have been tested for taste
and what products consumers like (consumer
insights, type of product, number of units,
flavours, quality, sales in year 1, etc.)
Cannabis taste in finished product, consumer
feedback, market share of product and sales in
year 1
What task/decision are you examining?
How to develop a product to mask the cannabis taste and create delicious products consumers will like.
The AI Canvas for Cannabis Edibles Product Development
© Agrawal, Gans, Goldfarb 2019
Appendix I: AI Canvas
[DOCUMENT TITLE]
4
Appendix II: ROI, Cost and Process Analysis
Improvements
Reduced Labour Hours (min) 52%
Labour & Other Cost Savings $ 26,030
Improved ROI (Base Case) 18%
Probability of Masked Cannabis Taste & Success of Product in Market
Success Metric
Probability of
Success
Cost of Risk Probability of
Success
Cost of Risk 5-year Savings
No Cannabis Taste in Finished Product 80% $ 41,169 95% $ 10,292 $ 154,384
Product Holds Listing for 1 Year+ 75% $ 51,461 95% $ 10,292 $ 205,845
*Note: Probability of success is based on 9 SKUs tested to date.
Return on Investment Before and After AI
ROI Before AI ROI After AI
Year 1 Sales ROI % ROI Payback Period (Months) ROI % ROI Payback Period (Months)
Best Case $ 500,000 143% 4.94 178% 4.32
Base Case $ 250,000 21% 9.88 39% 8.63
Worst Case $ 50,000 -76% 49.40 -72% 43.16
*Note: Above sales data is not exact due to confidentiality of company figures.
[DOCUMENT TITLE]
5
Product Development Process for One Product (w/o AI)
R&D Technologist
PD
Manager R&D Technologist
PD
Manager Other

Minimum
Hours
Maximum
Hours Hours
Minimum
Cost
Maximum
Cost Cost Cost
Phase 1: Ideation 20 30 5 $ 526.82 $ 790.23 $ 131.70 $ 100.00
Phase 2: Market Research 100 120 10 $ 2,634.10 $ 3,160.92 $ 263.41 $ 1,000.00
Phase 3: Theoretical Prototype 10 15 2 $ 263.41 $ 395.11 $ 52.68 $ -
Phase 4: Prototyping 80 100 20 $ 2,107.28 $ 2,634.10 $ 526.82 $ 15,000.00
Phase 5: Product Testing & Validation 40 50 10 $ 1,053.64 $ 1,317.05 $ 263.41 $ 80,000.00
Phase 6: Product Launch 0 0 0 $ - $ - $ - $ 100,000.00
Phase 7: Post Product Launch Analysis 10 20 25 $ 263.41 $ 526.82 $ 658.52 $ 1,000.00
TOTAL 260 335 72 $ 6,848.66 $ 8,824.23 $ 1,896.55 $ 97,100.00
TOTAL DEVELOPMENT COST RANGE PER PRODUCT $205,845.21
$207,820.79
*Other includes budget for raw materials, lab testing costs, lab equipment, sales costs, data reports, etc.
Product Development Process for One Product (w/ AI)
R&D Technologist
PD
Manager R&D Technologist
PD
Manager Other
Minimum
Hours
Maximum
Hours
Hours
Required
Minimum
Cost
Maximum
Cost Cost Cost
Phase 1: Ideation 3 5 5 $ 79.02 $ 131.70 $ 131.70 $ 100.00
Phase 2: Market Research 10 15 2 $ 263.41 $ 395.11 $ 52.68 $ -
Phase 3: Theoretical Prototype 0 0 0 $ - $ - $ - $ -
Phase 4: Prototyping 80 100 20 $ 2,107.28 $ 2,634.10 $ 526.82 $ 15,000.00
Phase 5: Product Testing & Validation 20 30 10 $ 526.82 $ 790.23 $ 263.41 $ 60,000.00
Phase 6: Product Launch 0 0 0 $ - $ - $ - $ 100,000.00
Phase 7: Post Product Launch Analysis 5 10 5 $ 131.70 $ 263.41 $ 131.70 $ 500.00
TOTAL 118 160 42 $ 3,108.24 $ 4,214.56 $ 1,106.32 $ 175,600.00
TOTAL DEVELOPMENT COST RANGE PER PRODUCT $179,814.56 $180,920.88
[DOCUMENT TITLE]
6
Appendix III
• Industry reports
• Consumer insight
reports (food or
cannabis)
• OBi product and
sales data
• Clinical trials or
palatability studies;
sensory panels
• human judgement to avoid lost
opportunities (i.e. innovations not
done before so the machine does not
have the data to feed it)
• human judgement to narrow down
list of potential products
• Action 1: Choose
flavours that mask
cannabis taste
• Action 2: Choose
products that will
succeed in market
• No cannabis taste
• ROI of 35% in year 1
• Product Development
and R&D team
• Launched product
performance (i.e. sales,
consumer responses,
etc.)
• Continually feed
updated inputs
Source: CDL Introduction: Economics of AI Slides and Prediction Machines by Ajay Agrawal, Joshua Gans and Avi Goldfarb.
!"# St. George Street
Toronto, Ontario, Canada M#S 6E8
T 9!8.:;<.#;"6
rotman.utoronto.ca
Final Page
Grade: _______