AI in a Product’s Journey

Deep Learning Assignment

Group 9
i|AI in a Product’s Journey

Abstract
ccwcw
ii | P a g e

Table of Figures

Figure 1: Market Research.....................................................................................................................4

Figure 2: Product Design........................................................................................................................5

Figure 3: Product Developement...........................................................................................................6

Figure 4: AI in Apple..............................................................................................................................6

Figure 5: AI in Product Design................................................................................................................7

Figure 6: AI in Retail.............................................................................................................................10
iii | P a g e

Table of Contents
Abstract................................................................................................................................................1

Market Research.................................................................................................................................4

Conversation AI chatbots...............................................................................................................4

Social Media Listening...................................................................................................................4

Collating data..................................................................................................................................4

Product Design:...................................................................................................................................5

Product Development:.......................................................................................................................5

Product Design....................................................................................................................................7

Production:......................................................................................................................................7

Logistics:..........................................................................................................................................8

AI in Marketing...................................................................................................................................9

AI in the retail space.........................................................................................................................10

1|Page

Market Research
Market research is a vital step because it lays foundation for a successful product launch. It is
the process of understanding people’s expectations of the product and how much they are
willing to pay. When the organization does not invest in market research ahead of the time,
the product is set to fail. Knowing the importance, many companies fail to see it as a
significant step.
There is an array of market research techniques
that an organization can use to learn more
about their customers including interviews,
survey, focus groups and market reports but the
downside to the traditional market research is
that it is expensive and time-consuming.
Figure 1: Market Research

Artificial Intelligence (AI) is the solution to this problem, in the following report, real world
examples of how Apple can use AI to optimize market research will be discussed

Conversation AI chatbots
For Apple, feedback is vital for a continual improvement of their iPhone models but filling
out a lengthy survey can be frustrating. Conversational AI chatbots can make the process
engaging and not boring. AI chatbot survey by an organization CONVRG had 70-80%
customers answering questions including open ended questions which is 3 times higher than
traditional routes. Also, chatbot surveys analysis helps the organizer understand who left the
survey midway and when helping the researchers identify the point of conflict and make
changes for better conservation flow.
Kantar’s conversational AI chatbots have a speech-based survey with the respondent as a part
of the survey. It can also have human like conversations on social media platforms like
WhatsApp, Facebook and voice assistants like Alexa.

Social Media Listening

Apple can use the power of AI to understand what users of existing Iphone are saying about
the product. AI social media monitoring tools can help track global brand mentions,
understand changing consumer trends, keep a tab on brand mentions and find better avenues
for social media promotions. These insights can help Apple identify the discrepancies users
are facing on a larger scale and make these changes in the upcoming model of the iPhone.

Collating data
The last part of the Market research is making reports of the data collected which can be a
tedious process because of the number of data sources. AI can simply use an algorithm to
make certain judgments and conclusions about the data and make a report. This frees up the
researchers and they can focus on high value tasks like making hypotheses, validating the
data collected and communicating to the top management of Apple.
2|Page

Product Design
Apple uses Core ML to take advantage of powerful hardware technology such as the CPU,
GPU, and Neural Engine in the most efficient way possible, maximising performance while
minimising memory and power consumption. The team creates the best-suited hardware in
order to maximise its potential.

Figure 2: Product Design

There's a persistent limiting element when it comes to mobile device performance, and it's the
power of the hardware inside the devices themselves, regardless of how precise or optimised
our machine learning model.
Neural processing units, or NPUs, are specialised pieces of hardware that improve the
performance of machine learning models. NPUs are AI-specific processors that are rewriting
the rules of what is possible when deploying neural networks on-device.
When we run machine learning models to make predictions (inference), we may need to
perform massive calculations that are both compute and memory-intensive, depending on the
model and task at hand. We can push algorithms to their limits if we have a lot of powerful
cloud-based CPUs and GPUs to perform this heavy data processing (and a lot of money).
With CPUs, we can already run fairly powerful and accurate neural networks on-device (and
sometimes GPUs). However, in order to fully realise the undeniable promise and potential of
on-device machine learning across industries, chip and smartphone manufacturers have had
to collaborate in recent years to develop chipsets that are custom-made for running advanced
neural networks that can track real-time user activities such as iPhone, iPad, Apple Watch,
and so on.

Product Development
The behaviour of a machine learning app is determined by the data it receives, and it
responds to changing information and conditions. Because of this dynamic behaviour,
designing the UI and user experience of a machine learning app is uniquely difficult: rather
than designing specific reactions to a static set of scenarios, you design it by teaching it how
to interpret data and react accordingly. The most recent advancement in product development
came in the form of various new features such as:
Sound Recognition in iOS 14: iOS 14 introduces a number of useful new features, some of
which are intended to be Accessibility options but are also useful to everyone. Sound
Recognition is another Accessibility feature that can provide users with hearing impairments
with a visual cue for a variety of sounds that are occurring around them. This accessibility
3|Page

feature was not mentioned on stage, but it allows an iPhone to listen for and identify sounds
such as doorbells, sirens, dogs barking, or babies crying.

Figure 3: Product Development

Native sleep tracking for the Apple watch: Users can set their own bedtime regimens with
the sleep app on Apple Watch. To track sleep, users simply need to wear their Apple Watch
to bed and it will do the rest. It's possible for them to track their sleep patterns over the past
two weeks when they wake up.
Facial Recognition for Homekit: Apple Homekit is a smart home solution that lets users
control and communicate with connected devices using an app. Users can provide a
mechanism to configure accessories and develop actions to control them using the HomeKit
framework.
App Library Suggestions: A major change in how we manage apps came with the release of
iOS 14. "On-device intelligence" is used in a new folder in the App Library to show users
programmes they're likely to require next.

Figure 4: AI in Apple

Handwriting Recognition: If you use the Apple Pencil to scribble something, the iPad can
read it and, with Scribble, turn it to typed text. There was no mention of AI in this feature, but
it does work like most machine learning.
4|Page

Product Design
Every day, we observe tremendous technical advancements in big data, algorithmic
development, connection, cloud computing, and processing capacity. AI's performance,
accessibility, and prices have never been better thanks to these new technologies.

Figure 5: AI in Product Design

The adoption of cutting-edge technologies such as artificial intelligence, machine learning,

and blockchain has transformed the previously unorganized and fragmented production and
logistics sector. Predictive analytics, zero defect manufacturing, autonomous vehicles, and
smart roads are all affected by these technologies.

Production:
Machining is the process of shaping and sizing raw materials like metal, wood, and plastic
into the final shape and size needed. For tasks like cutting, drilling, boring, milling, and so
on, modern machine shops use computer-controlled precision machining equipment.
Machining is an important aspect of the manufacturing process. The time-consuming task of
cost assessment has long been a major concern for the machining business. When a machine
shop receives an order for manufacturing parts, for example, it must carefully examine the
blueprint provided by the customer to determine the individual cuts and holes that must be
done for each item This manual procedure necessitates machining experts committed to the
cost estimation process, which can be a bottleneck for the machine shop's overall efficiency.
5|Page

Handling the data management of the machining process using AI is a new solution for this
problem. To begin, a blueprint is submitted to a tool, and a region for estimation is specified.
The user then just presses the Cost Estimation button to begin the process. Every object that
has to be processed is detected and highlighted by the tool. We can also view the number of
detected processing steps, how much each one costs, and the overall anticipated expenses for
the entire portion. Users will also be able to submit extensive information for each part or
have specialists update any part that requires minor repairs using the tool. The tool can detect
hundreds of processing elements on a design in less than a second, considerably increasing
the efficiency of machining specialists' cost estimation.
Customers will have a better experience and additional business prospects if machine shops
can reply swiftly to cost estimating queries. Furthermore, these types of systems can serve as
a central repository for each machine shop's specialist knowledge. This allows the company
to readily transfer expertise with junior employees and learners without creating bottlenecks,
allowing the business to go up.
These technologies enable users to create and maintain orchestrated machine learning
pipelines, from data pre-processing and validation to model training, validation, and
deployment. Instead of repeatedly requesting a manual rerun of the entire training process
from AI experts, everything can be automated and continually incorporated simply by calling
the pipeline, making the entire service resilient and ready for production.
With AI platform pipelines, a tool can create a positive feedback loop between machining
experts and the model training pipeline, resulting in increasing model accuracy and
productivity as the experts continue to use the tool. In machine learning, this loop of
application and training is referred to as active learning.

Logistics
Logistics is another industry where AI has begun to make an impact by making supply chain
management more seamless. It is now an essential component of every business and its future
software system.
With cognitive automation, AI plays a critical role in saving time, lowering costs, increasing
productivity, and improving accuracy. It allows us to save time and money because it helps in
automating various time-consuming processes and helps in demand forecasting. AI aids in
logistics route optimization, which aids in lowering shipping costs, which in turn aids in
generating more profits. Computers with artificial intelligence can gather and analyse
information in seconds to make informed decisions, saving humans time.
Patterns in supply chain data are frequently discovered with the help of Machine learning by
relying on algorithms that can quickly pinpoint the most influential factors. Its algorithm and
the logistics companies that use it are capable of quickly analysing large, diverse data sets,
improving demand forecasting accuracy. In the collaborative supply chain and logistics
segment, machine learning helps to reduce freight costs, improve supplier delivery
performance, and reduce supplier risk.
6|Page

Machine learning can assist predict unforeseen scenarios, which can further reduce the
likelihood of any mishaps during the delivery of goods, thanks to the digitization of logistic
planning. In the logistics business, machine learning replaces the time-consuming tasks of
planning and scheduling, allowing for greater accuracy and efficiency, and therefore
streamlining the processes.

AI in Marketing
With the volume of data being generated by users growing at a breakneck pace, there needs
to be a way to properly process this data and find patterns within them. This is where AI
models (Deep Neural Networks) shine, as they can be built with the specific purpose of
detecting patterns. Before it comes to that, the huge data needs to be distilled and
compressed. AI models can be designed to filter irrelevant data (for the specific purpose).
As is the case with digital marketing, a company needs to be weary of the types of
advertisements they place in different offline locations as well. This choice must take into
account the people’s preferences there which comprises of many other factors. Deep models
can be trained to evaluate people’s preference for advertisements.
Most applications of AI can be seen in the digital sphere. Recurrent networks can be used for
text generation for content that needs to be posted on social media. Upper management of
companies can, however, be reluctant to leave the important task of content generation to
machines and they have content managers generate such content. AI can help in this aspect as
well. Transformer networks can be used for translation purposes as well.
Content curation activities can be augmented with the help of AI as well. Targeting of ads
requires bots to make thousands of micro-decisions every second. Some of these decisions
can be to decide whether to display a particular ad to a certain user; how much to bid for a
certain ad spot; which variant of an ad to display to the user etc. This is extensively used in
digital marketing for placing and choosing ads.
AI can be used to monitor the status of customer feedback. Much of the feedback that
customers post these days are on social media. NLP models can be deployed to perform
sentiment analysis on the reviews posted by the people. This can be used to measure the
impact of any public decision (especially product announcements) by the company.
7|Page

AI in the retail space

To compete in today's environment, retailers must deliver an exceptional customer
experience. They must also eliminate waste and improve their operations by reducing costs.
Retail digital transformation is about transforming data into actionable insights that drive
business outcomes. With
AI in retail, retailers can benefit from deep learning and machine learning to improve
processes and increase sales. The rapid emergence and evolution of digital technology has
greatly improved the efficiency and accuracy of every branch of the retail industry.

Figure 6: AI in Retail

Some of the components of retail where AI is/can be used extensively:

 With AI-driven point of sales solutions, retailers can turn huge amounts of data into
valuable insights, which can be used to create new products and services that cater to
specific customers.
 AI-powered customer service bots can identify the best representative to handle a
given issue and provide the necessary background information to enable the rep to
complete their work efficiently.
 AI used data to identify the best representative and thereby address customer needs. It
them provides the rep with the suitable customer thereby saving time and man power.
 AI helps plan inventory replenishment by forecasting and predicting demand.
 This highly tedious task of segregating products based on images and description was
earlier assigned to the company’s human workforce, resulting in a major investment
of time and resources, with less than satisfactory results. The AI-based system relies
on machine learning to analyse the large repository of product categorization data
already available and accordingly implement and further perfect the same.