haiped. impact of AI in marketing comms and CX

Will AI completely ruin everyone’s in advertising’s
day?

No. Not everyone’s.
And not completely.

Introduction to AI
Marketing Impact
Getting Started

Introduction to AI
1. We are likely in a hype cycle, but the impact will still be notable
2. Broadly speaking, AI excels at:
• Capturing information from data
• Extrapolating from available data
3. AI struggles with (probably ALWAYS) working outside of explicit rules where knowledge of the underlying structure
of the environment is needed
Marketing Impact
4. Marketing covers a range of outputs from Low to High Cognitive Complexity, with Analysis, Decision-making, and
Execution key steps in any task.
5. AI’s impact ranges from automation to augmentation. High Cognitive Complexity tasks can’t be automated.
• A large portion of marketing communications activities are low complexity and will be automated
6. AI will have a large impact on marketing communications and digital touchpoints, especially on operations.
7. But be aware of limitations in
• Analysis: lack of explainability
• Decision-making: lack of common sense
• Execution: lack of creativity (distinctiveness) and production accuracy
Getting Started
8. Structure an AI strategy against clear business value and role of technology (why should AI be used here?)
9. Use current known applications as a guide for innovations in AI
10.Start using it in an individual capacity

Execution
Decision-making
Analysis

Execution
Decision-making
Analysis
Low Cognitive Complexity High Cognitive Complexity

Execution
Decision-making
Analysis
Automated Augmented
Limitations:
No
model
of
the
world

Execution
Decision-making
Analysis
Volume
of output
Automated Augmented
Limitations:
No
model
of
the
world

Execution
Decision-making
Analysis
Volume
of output
Higher
impact
Lower
impact
Higher
impact
Automated Augmented
Limitations:
No
model
of
the
world

Execution:
Strong with generic
applications in text,
generative image and
videos can struggle
Decision-making:
Strong with rules-based,
decisions, ones
requiring implicit
knowledge will
underperform
Analysis:
Strong potential in
finding patterns that
traditional techniques
would not
Limitations:
No
model
of
the
world
Volume
of output
Automated Augmented
Higher
impact
Lower
impact
Higher
impact

Execution:
Strong with generic
applications in text,
generative image and
videos can struggle
Decision-making:
Strong with rules-based,
decisions, ones
requiring implicit
knowledge will
underperform
Analysis:
Strong potential in
finding patterns that
traditional techniques
would not
Limitations:
No
model
of
the
world
AI is probably not the
first step. Supporting
infrastructure is needed.
Make sure it’s
controllable, be mindful
of biased data /
endogeneity.
AI might be better than
current solutions, but
comes with drawbacks in
explainability.
Lacks common sense
AI can hallucinate or create
generic results (bad for
brand)
Volume
of output
Automated Augmented
Higher
impact
Lower
impact
Higher
impact
AI can assist, with
Automation Paradox as a
potential drawback.
Novel thinking remains out of
reach
Image and video output still in
its infancy, not necessarily
suitable for production

High
Complexity
Work
Increases
Low Complexity Work Increased
Investment Increases:
Impact of Marketing
Advances in analysis
demonstrating the value of High
Complexity work, and efficiency of
Low Complexity work warrants
increased investment in both.
Investment stays the same:
Creativity takes centerstage
AI and Automation allows more
time to be spent on high
complexity work, with existing Low
Complexity work increasingly
automated, but not expanded.
Digitalization of Marketing
automated touchpoints or
activities to be created, while
current High Complexity work is
seen as sufficient.
Investment Decreases:
Efficiency of Marketing
Current High Complexity work is
seen as sufficient, with Low
automated but not expanded,
overall investment decreases
Four possible futures
The
same
Increases
The same Increases

Introduction to AI
Marketing Impact
Getting Started
Interest in AI
AI in Brief

https://www.gartner.com/en/articles/what-s-new-in-artificial-intelligence-from-the-2022-gartner-hype-cycle

haiped. impact of AI in marketing comms and CX

https://www.aiiottalk.com/types-of-artificial-intelligence-details-that-everyone-should-know/
• Body Level One
• Body Level Two
• Body Level Three
• Body Level Four
• Body Level Five

• Body Level One
• Body Level Two
• Body Level Four
• Body Level Five

https://www.gartner.com/en/articles/what-s-new-in-artificial-intelligence-from-the-2022-gartner-hype-cycle
Overall, we’re
probably here

“machines that mimic human intelligence in
tasks such as learning, planning, and problem-
solving through higher-level, autonomous
knowledge creation.”
De Bruyn, A., Viswanathan, V., Beh, Y. S., Brock, J. K.-U., & von Wangenheim, F. (2020). Artificial Intelligence and Marketing: Pitfalls and Opportunities. Journal of Interactive Marketing, 51, 91–105. https://doi.org/10.1016/j.intmar.2020.04.007

Branches of AI

Burgess, A. (2018). The Executive Guide to Artificial Intelligence. Springer International Publishing. https://doi.org/10.1007/978-3-319-63820-1
Why is it happening
What is happening
Capture information
Turning unstructured data into structured data.
Find patterns, or clusters, in data that could
inform useful business insights.
Deriving meaning from data to influence action.
Creates a view of what needs to be done.
Understanding the underlying concepts.
The AI capabilities that we have today and will
have in the near future (and maybe even ever)
do not have the ability to understand.

Why is it happening
What is happening
Capture information
On par or surpasses human capabilities in certain use cases
Assisted or replaced by AI

It doesn’t know what hands are or how they work.

Why is it happening
What is happening
Capture information
Not currently feasible, but
generative AIs mimic this
Assisted or replaced by AI Performed by Human, assisted by AI
https://www.wired.com/video/watch/tech-support-ai-expert-answers-ai-
questions-from-twitter

Why is it happening
What is happening
Capture information
Search
Data Analysis / Clustering Prediction
Understanding
Structured
data
Unstructured
data
Optimisation
Speech Recognition
Image Recognition
Natural Language
Understanding

Introduction to AI
Marketing Impact
Getting Started
Broad Applications
CX & Marketing Framework
AI Examples
AI Limitations
Impact on Operations

Why is it happening
What is happening
Capture information
With this

Why is it happening
What is happening
Capture information
• Author a new strategy
• Create a new, relevant, surprising
concept that answers the brief
• Get buy-in for the above
• Identify what changes to make on
existing activities / executions
• Make sense of current strategies’
performance
• Optimize existing activities /
executions
• Measure performance of existing
activities
• Mine useful data from existing activities
• Find patterns within and between
executions
With this
You can do this
(Assuming you have access to data) (Assuming you can make use of the data)

Adapted from: Hays, C., & Wind, Y. (2017). Beyond advertising: Creating value through all customer touchpoints. Gildan Media.
Digital Marketing Touchpoints: What business and customers see
Customer
Customer
Journey
Mapping
Stories and
Content: Defined by
Brand and Comms
Strategy
Systems and Services:
Defined by Marketing
Operations
Data and Models:
Defined by Data
Strategy
Interfaces and
Touchpoints:
Defined by
Platforms Strategy
Technologies and
platforms through
which customers
interact with the brand
Communications
from and about the
product or brand
Technologies used
to activate or
manage digital
touchpoints
Data from digital
touchpoints that
needs to be
gathered, managed
and analysed
Trigger Initial
Consideration
Active
Evaluation
Momentof
Purchase
Product
Experience
Advocacy LoyaltyLoop
How much will AI contribute here?
And how will it improve on what’s happening here?

High cognitive complexity
Average cognitive complexity
Low cognitive complexity
• Mostly the same, but different in some
minor way.
• Single creative route from an existing
strategic and creative brief.
• Something new, but in line with what’s
been done previously.
• Ongoing deliverables from existing
strategic brief but involving a new creative
brief.
• Something entirely new and original.
• New strategic briefs and multiple creative
routes.
Adapted from Farmer, M. (2017). Madison avenue manslaughter - an inside view of fee-cutting clients, profit. Lid Publishing Inc.

Execution
Decision-making
Analysis
Mostly the same, but different in some minor way. Something new, but in line with what’s been done previously. Something entirely new and original.

Execution
Decision-making
Analysis
• More precise measurement
• Measure ROI by channel, campaign, and overall
• Predict content, campaign performance
• Realtime awareness of customer needs
• Determine what offers will motivate users to
action
• Adapt audience targeting
• Refine the online customer experience
• Generate targeted, personalized messaging
• Create data-driven content
• Answer customer queries
• Uncover insights from data
• Predict content, campaign performance
• Identify Messaging themes that drive engagement
• Direct customer queries
• Create additional data-driven content
• Uncover insights from data with complex patterns
• Deeper behavioural segmentation
• Construct buyer personas
Forbes, Quantcast. (2021). Lessons of 21st-Century Brands: Modern Brands & AI Report.,
Marketing AI Institute. (2022). 2022 State of Marketing and Sales AI Report—Marketing AI
• LLMs and Generative AI Impact?
Mostly the same, but different in some minor way. Something new, but in line with what’s been done previously. Something entirely new and original.
• LLMs and Generative AI Impact?

Personalization
• Recommending highly targeted content to users in real-time
• Determine offers that will motivate individuals to action
• Present individualised experiences on the web or in-app
• Customise email nurturing workflows and content
• Engage website visitors in conversations through chatbots that learn and
evolve
• Optimise email send time at an individual recipient level

Walmart Text-to-Shop
• “When different things pop into your mind you’re usually out and about running
errands. I don’t have time to log into the app and add to the cart.”
• Text to Shop is seamlessly connected to your Walmart account, so we know
your usual ordered items.
• Underlying technology and use case seems reasonable, but issues in UX,
Chatbot memory, and limitations in text-based interfaces point to the difficulty
in building a successful Chabot product (Perez, 2023)

https://landscape.brxnd.ai/companies/logoai
Virtual Models
Deep Agency offers virtual photo studio services with advanced AI technology for
professional photos without leaving home. Hire virtual models and elevate your
photo game and say goodbye to traditional photo shoots.
Adobe Firefly
Firefly, a family of creative generative AI models coming to Adobe products.

https://www.pcmag.com/how-to/negotiating-a-vendor-deal-with-walmart-you-may-be-talking-to-ai
Walmart AI Price negotiation
• The retailer tells the AI chatbot its budget, and the chatbot communicates with a
human vendor.
• The chatbot will negotiate with human vendors to finish deals.
• And a majority of suppliers – three out of four – actually like dealing with the robot
more than a Walmart negotiator.

Stable Diffusion OpenAI Midjourney
Brainstorm, visualisation, and analysis assistants
• While not a replacement for people, generative AI tools will help the ideation and
visualisation process.
• It might replace components of the production process
• Deep learning can work in service of more advanced marketing analysis, but might
not be worth the effort.

While static images have massively
improved over the last decade, video
requires a more fundamental
understanding of underlying structures
that current approaches might not
overcome.

• Body Level One
• Body Level Two
• Body Level Four
• Body Level Five
AI can be extremely useful for art direction,
testing out creative routes, or brainstorming.

But the levels of insight needed to make the
following ad are out of reach of current AI:

Farmer, M. (2017). Madison avenue manslaughter - an inside view of fee-cutting clients, profit. Lid Publishing Inc.
Creative agency output tends
toward Low complexity,
adaptive work
Sample workload splits for an
agency from Madison Avenue
Manslaughter
Distribution of Deliverables
Broadcast Print / OOH Print / BTL Print / DM Digital / Social
Strategic /
Research
Total % SMUs
Adaptation -
Low
2% 16% 2% 16% 6% 4% 47%
Adaptation -
Average
0% 2% 1% 0% 2% 2% 7%
Adaptation -
High
- - 0% 0% 0% 1% 1%
Origination -
Low
2% 1% 1% 1% 5% 3% 12%
Origination -
Average
6% 2% 1% - 10% 2% 21%
Origination
High
5% - - - 3% 4% 12%
% of total
SMUs
15% 20% 5% 18% 27% 15% 100%

1 De Bruyn, A., Viswanathan, V., Beh, Y. S., Brock, J. K.-U., & von Wangenheim, F. (2020). Artificial Intelligence
and Marketing: Pitfalls and Opportunities. Journal of Interactive Marketing, 51, 91–105.
https://doi.org/10.1016/j.intmar.2020.04.007
• Biased Artificial Intelligence – training data can have embedded
bias that the model learns from and then uses
• Explainable Artificial Intelligence – risk increases when acting on
model recommendation that you fundamentally don’t understand
• Controllable Artificial Intelligence – unforeseen decisions made
by the AI can be damaging and must have controls in place
• The Paradox of Automation – the more a system is Automated,
the more it relies on experts, but doing repetitive tasks is a a form
of training en route to becoming an expert.
Dangers Limitations
• Lack of Common Sense – difficult to cover all common sense
• Objective Functions – difficult to define completely
• Safe and Realistic Learning Environment – can’t cover all
marketing variables in a simulated environment
• No theory of mind / concept of the world – Needed to make
calls that don’t conform to prior data or examples

“There’s a process of the economy churning, destroying a
lot of jobs, then creating new kinds of jobs, and people
reallocating to new roles in a more productive economy, in
which people are on average considerably richer and live
longer.”
Ben Jones, Kellogg School of Management (May 2023)

Technology adoption takes much longer than predicted, certain companies and industries
are already affected. Whether it’s a net win or loss is hotly debated.

Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An Early Look at the Labor Market Impact Potential of
Large Language Models (arXiv:2303.10130). arXiv. http://arxiv.org/abs/2303.10130
“80% of the U.S. workforce could have at least 10%
of their work tasks affected by the introduction of LLMs,
while approximately 19% of workers may see at least
50% of their tasks impacted.”

2022 State of Marketing AI Report

Distribution of Deliverables
Broadcast Print / OOH Print / BTL Print / DM Digital / Social
Strategic /
Research
Total % SMUs
Adaptation -
Low
2% 16% 2% 16% 6% 4% 47%
Adaptation -
Average
0% 2% 1% 0% 2% 2% 7%
Adaptation -
High
- - 0% 0% 0% 1% 1%
Origination -
Low
2% 1% 1% 1% 5% 3% 12%
Origination -
Average
6% 2% 1% - 10% 2% 21%
Origination
High
5% - - - 3% 4% 12%
% of total
SMUs
15% 20% 5% 18% 27% 15% 100%
The creative agency model
will need to change
• 59% significantly affected by AI
• 28% partially affected by AI

“The biggest barrier to AI achieving
escape velocity in my opinion is the
over-inflation of expectations.”
(Burgess, 2018, p. 23)

Introduction to AI
Marketing Impact
Getting Started For agencies
For individuals

Consider the challenges in AI Implementation
Dangers
• According to surveys, 70% of projects see no or
minimal impact from the introduction of AI
systems2
• Generic produce generic outputs which dilute
brand distinctiveness
• More complicated does not mean better. Deep
learning analytics only yielded 2% improvement
over traditional marketing analytics3
Barriers to success4
• Lack of education, awareness, and strategy cited
as some of the key people shortfalls
• Lack of technology, budget, data, and
infrastructure cited as key investment shortfalls
2. Westenberger, J., Schuler, K., & Schlegel, D. (2022). Failure of AI projects: Understanding the critical factors.
Procedia Computer Science, 196, 69–76. https://doi.org/10.1016/j.procs.2021.11.074
3. Urban, G., Timoshenko, A., Dhillon, P., & Hauser, J. R. (n.d.). Is Deep Learning a Game Changer for Marketing
Analytics?
4. Marketing AI Institute. (2022). 2022 State of Marketing and Sales AI Report—Marketing AI

Companies that succeeded in the deployment of advanced digital
technologies did an assessment of where they were in terms of
the nine performance indicators:
1.Strategy,
2.Opportunity focus,
3.Budget,
4.Results,
5.Data execution,
6.People,
7.Partnerships,
8.Deployment, and
9.Governance
Crucial for successful delivery

Identify a problem, don’t build thing and then go looking for a problem

Current applications are dominated by tactical executions,
somewhat safe place to start:
Important for
Omnichannel
marketing
Keep in mind, these use
cases are built on an
existing data and
technology infrastructure.
AI comes after, or at least,
in tandem.

https://chiefmartec.com/2018/01/5-disruptions-marketing-part-4-digital-everything-2018-update
Technology
Domain
Human
Domain
IT used to operate here Marketing used to operate here
Traditionalmarketing
Increasingdigitalsophistication
Marketing

Marketing
Technology
Domain
Human
Domain
Increasingdigitalsophistication
Technology
sophistication is
marching forward

Technology
Domain
Human
Domain
Increasingmarketingsophistication
Marketing
Technology
sophistication is
marching forward
Marketing
sophistication must
follow suite

Specialists who focus here Specialists who focus here
Generalists who can join the two
Technology
Domain
Human
Domain

Cassie Kozyrov, Chief Data Scientist at Google: https://www.youtube.com/watch?v=gjDRDrOrH94&list=WL&index=14
• The recent revolution in AI has been a UX
revolution rather than a breakthrough in
the underlying technology.
• AI models are for the first time, accessible
and relatively easy to use.
• Explore how your work is distributed in low
to high complexity, and explore how AI
tools can automate the low, or augment
the high complexity work.
• Use it in areas that your are an expert in to
understand some of the underlying
limitations.

Adapted from: Hays, C., & Wind, Y. (2017). Beyond advertising: Creating value through all customer touchpoints. Gildan Media.
Digital Marketing Touchpoints: What business and customers see
Customer
Customer
Journey
Mapping
Stories and
Content: Defined by
Brand and Comms
Strategy
Systems and Services:
Defined by Marketing
Operations
Data and Models:
Defined by Data
Strategy
Interfaces and
Touchpoints:
Defined by
Platforms Strategy
Technologies and
platforms through
which customers
interact with the brand
Communications
from and about the
product or brand
Technologies used
to activate or
manage digital
touchpoints
Data from digital
touchpoints that
needs to be
gathered, managed
and analysed
Trigger Initial
Consideration
Active
Evaluation
Momentof
Purchase
Product
Experience
Advocacy LoyaltyLoop
• AI will automate derivative works consisting of only minor variations (retail copy, content
calendars, emailers, SEO)
• AI will help but not replace the production process in all major components of integrated
marketing
• AI would replace high complexity work only through a lack of appreciation of the value of creative
or bespoke work (whether coding or concept)
• AI will automate decision-making in rules-based scenarios, and analysis from unstructured and
structured data
• Decision-making relying on Implicit knowledge remain out of reach
• A greater focus on data analysis and AI as a “magic black box” risks over-reliance on
quantitative data to dictate decisions, at this point marketing becomes an efficiency game
• There is work to be done to prove the value of human input and intuition

High
Complexity
Work
Low Complexity Work
Investment Increases:
Impact of Marketing
Advances in analysis
demonstrating the value of High
Complexity work, and efficiency of
Low Complexity work warrants
increased investment in both.
Creativity takes centerstage
time to be spent on high
complexity work, with existing Low
automated, but not expanded.
Digitalization of Marketing
automated touchpoints or
activities to be created, while
current High Complexity work is
seen as sufficient.
Investment Decreases:
Efficiency of Marketing
Current High Complexity work is
seen as sufficient, with Low
automated but not expanded,
overall investment decreases
Four possible futures
(although we’ll probably
bounce between them
depending on economic
or business cycles)
The
same
Increases
The same Increases

• We are in the middle of a hype cycle, some of it is relevant though:
AI has already had a large impact on marketing.
• Marketers are still at the beginning stages of understanding how to
use or integrate AI, but most are optimistic.
• AI has distinct weaknesses that current approaches might not be
able to overcome
• Agency: identify small use cases to automate and integrate into
processes, start to change costing model for automated
deliverables, define where your value lies.
• Individuals: experiment with the tools, identify small areas in the
day to day to trial AI, understand where and how it is not so great.

haiped. impact of AI in marketing comms and CX

More Related Content

What's hot (20)

Similar to haiped. impact of AI in marketing comms and CX (20)

Recently uploaded (20)

haiped. impact of AI in marketing comms and CX