Agentic AI

– the new
frontier in
GenAI

An executive playbook
Harnessing AI isn’t just about
technology— it’s about unleashing
unprecedented potential.
In an era where speed, efficiency, and customer centricity dictate market leadership, organisations need to
harness every tool at their disposal. Over the past couple of years, artificial intelligence (AI) has exploded onto
the world stage, with companies and individuals across the globe rapidly adopting the technology. The GCC is
playing a lead role in the space, with business leaders in the region exploring ways of integrating this rapidly
developing technology into their operations.

Generative AI (GenAI) is being recognised as a game-changer for innovation in the region, empowering
enterprises by automating routine tasks, enhancing customer experiences and assisting in critical
decision-making processes. Insights from our 27th Annual CEO Survey: Middle East findings have shown that
73% of CEOs in the Middle East believe GenAI will significantly change the way their company creates, delivers
and captures value over the next three years1. GenAI is poised to make a significant economic impact, with
estimates indicating that it could contribute between $2.6 trillion and $4.4 trillion annually to global GDP across
various industries by 2030. In specific sectors, such as energy, investments in GenAI are expected to triple,
from $40 billion in 2023 to over $140 billion by the end of the decade. This surge in investment reflects the
transformative potential of GenAI, particularly in enhancing productivity, streamlining business processes, and
reshaping value chains across industries2.

Against this backdrop, multimodal GenAI agentic frameworks has emerged as transformative catalysts,
enabling businesses to accelerate process automation at an unprecedented scale. This technology involves
multiple AI agents working together, each specialising in different tasks or data types, to solve complex
problems and automate processes. By collaborating and constantly learning, these agents enhance
decision-making, optimise processes, and drive innovation. It combines range of advanced AI techniques to
process diverse data types and automate complex tasks.

The central question isn’t whether to adopt this technology, but how swiftly organisations can integrate it to
stay ahead of the competition. This executive playbook explores how organisations can leverage this
technology to boost operational efficiency, enhance customer experience, and drive revenue growth. It provides
real-world success stories spanning industry sectors and organisational functions, strategic insights, tactical
blueprints, and best practices to guide your journey into this revolutionary landscape.

Key insights

● Agentic AI, differentiated by its advanced human-like reasoning and interaction capabilities, is
transforming the manufacturing, healthcare, finance, retail, transportation, and energy sectors, among
others.
● Organisations’ AI strategies should leverage multimodal GenAI capabilities while ensuring ethical AI
safeguards to drive autonomous process re-engineering and enhanced decision-making across all lines
of business.
● Integrated effectively, agentic AI can enhance efficiency, lower costs, improve customer experience, and
drive revenue growth.
What is agentic AI?
Agentic AI generally refers to AI systems that
possess the capacity to make autonomous
decisions and take actions to achieve specific
goals with limited or no direct human
intervention3.

Key aspects of agentic AI

Autonomy: Agentic AI systems Goal-oriented behaviour: These Environment interaction: An
can operate independently, AI agents are designed to pursue agentic AI interacts with its
making decisions based on their specific objectives, optimising surroundings, perceiving changes
programming, learning, and their actions to achieve the and adapting its strategies
environmental inputs. desired outcomes. accordingly.

Learning capability: Many Workflow optimisation: Agentic Multi-agent and system

agentic AI systems employ AI agents enhance workflows and conversation: Agentic AI
machine learning or business processes by integrating facilitates communication
reinforcement learning techniques language understanding with between different agents to
to improve their performance over reasoning, planning, and construct complex workflows. It
time. decision-making. This involves can also integrate with other
optimising resource allocation, systems or tools, such as email,
improving communication and code executors, or search
collaboration, and identifying engines, to perform a variety of
automation opportunities. tasks.

Environment Learning
interaction capability

Goal-oriented Workflow
behaviour optimisation

Multi-agent
and system
Autonomy conversation
Evolution to multimodal GenAI agents
In AI, the only constant is change—embrace a culture of perpetual innovation.

The journey of agentic frameworks began as simple, rule-based systems designed to perform specific tasks. Over time,
these systems have evolved into sophisticated, multimodal agents capable of processing and integrating information from
various sources, such as text, images, and audio. Multimodality capabilities allow AI agents to understand, employ
reasoning, and interact like humans, enhancing their effectiveness and versatility to solve a wide range of business
problems4.

The evolution can be broken down into three key phases:

1.
(2000s) Integration of Machine Learning (ML)
○ Learning from data: The integration of ML allowed agents to learn from large datasets, improving their ability
to make decisions and perform tasks. This was a significant step forward from rule-based systems, as agents
could now adapt to new information and improve over time.
○ Natural Language Processing (NLP) enabled user interactions: Advances in NLP enabled agents to
understand and generate human language more effectively, making interactions more natural and intuitive.

(2010s)
2. Introduction of multimodality
○ Combining text, images, and audio: Multimodal agents emerged, capable of processing and integrating
information from various sources. For instance, an agent could analyse a text description, recognise objects in
an image, and understand spoken commands. This multimodality made agents more versatile and capable of
handling complex tasks.
○ Enhanced user interactions: Multimodal agents could interact with users in more dynamic ways, such as
providing visual aids in response to text queries or understanding context from a combination of spoken and
visual inputs.

3.
2020s- Advanced autonomy and real-time interactions
present ○ Advanced autonomy: Agents can operate independently, rationalise and set their own goals, develop path(s)
to attain these goals, and make independent decisions without constant human intervention, leveraging data
from multiple sources or synthetic datasets. In a multi-agentic orchestration system, the first set of agents
focus on mimicking human behaviour (e.g. ChatGPT-4o), that is, thinking fast to come up with solution
approach, while the second set of agents focus on slow reasoning (e.g. ChatGPT-1o) to come up with a vetted
solution5. Combining thinking fast and slow reasoning, agents can process information and make optimal
decisions in real-time – crucial for applications like autonomous vehicles, real-time customer service, and
various mission-critical business processes. This autonomy makes agentic AI particularly powerful in dynamic
and complex real-world environments.
○ User interactions within an ethical and responsible AI-controlled environment: With increased
capabilities, there has also been a focus on ensuring that agentic systems operate ethically and responsibly,
considering factors such as bias, transparency, and accountability.

Integration of ML (2000s)

NLP enabled user

Learning from data
interactions

Integration
IntroductionofofMachine Learning
multimodality (2000s)
(2010s)
Natural Language Processing
Combining text,
Learning from Dataimages, and (NLP) Enabled
Enhanced userUser
interactions
audio Interactions:

Goal-oriented Integration of Machine Learning (2000s)

AI agent Advanced autonomy and real-time interactions (2020s-present)
behaviour
Natural
User Language Processing
interactions within an
Human-like
Learning fromreasoning
Data and (NLP)
ethicalEnabled User
and responsible AI-
advanced autonomy Interactions:
controlled environment
Why organisations should pay attention
In the fast lane of technological evolution, missing the AI turn today means being outpaced tomorrow.

Agentic AI offers significant advantages in efficiency, decision-making, and customer interaction. By automating routine
tasks and providing intelligent insights, agentic AI can help organisations save time, reduce cost, and improve overall
productivity. Moreover, organisations who adopt an agentic AI system can gain a competitive advantage by leveraging its
capabilities to innovate and enhance their business operations. Lower cost to entry and economies of scale makes it
favourable for organisations to fully harness the capabilities it offers compared to its predecessors like traditional ML and
Robotic Process Automation (RPA)-driven automations.

Agentic AI systems can significantly enhance an organisation’s competitive edge by automating complex workflows,
reducing operational costs, and improving decision-making processes. These systems are designed to adapt to changing
business environments, driving higher productivity and enabling organisations to stay competitive. For example, agentic AI
can predict market trends and customer preferences, allowing businesses to tailor their strategies proactively. This
adaptability not only improves efficiency but also fosters innovation, giving companies a significant edge over competitors6.

Moreover, agentic AI systems can handle large volumes of data and extract actionable insights, which can be used to
optimise operations and enhance customer experiences. By automating routine tasks, these systems free up human
resources to focus on more strategic initiatives, thereby increasing overall organisational agility and responsiveness7.

Enhanced decision-making

Agentic AI systems can analyse vast amounts of data quickly and accurately, providing valuable insights to inform better
decision-making. Businesses can leverage these insights to optimise revenue and operations, identify market trends, and
make data-driven decisions. For instance, in the financial sector, AI can analyse market data to predict trends, inform
investment strategies, and boost investment ROI. In retail, it can streamline inventory management by predicting demand
and optimising stock levels.

Boosted efficiency and productivity

Agentic AI can significantly enhance business efficiency and productivity by automating routine tasks and processes. This
allows employees to focus on more strategic and creative activities. For example, in customer service, agentic AI can
handle common inquiries, freeing up human agents to tackle more complex issues. In manufacturing, AI-driven robots can
manage repetitive tasks with precision and consistency, reducing errors and increasing output.

Improved customer experience

By integrating agentic AI, businesses can offer personalised and responsive customer experiences. AI-driven chatbots and
virtual assistants can provide instant support, answer queries, and even recommend products based on customer
preferences and dynamic interactions. This improves customer satisfaction, builds loyalty, and drives sales. For example,
e-commerce platforms use AI to recommend products based on browsing history and purchase behaviour.
 How to conceptualise agentic AI
solutions for future business operations
Agentic AI systems are redefining customer service centres and are gaining popularity as a game-changing capability
for both government entities and private sector organisations. While traditional rule-based chatbots
(software-as-a-service) provided basic 24/7 support, and Retrieval Augmented Generated (RAG)-based chatbots
enhanced human-like interactions (enhanced software-as-a-service), agentic AI surpasses both in terms of accuracy,
contextual coherence, and problem-solving ability.

In terms of accuracy, rule-based chatbots are limited to programmed responses, causing inaccuracies when queries
fall outside of predefined rules. RAG-based chatbots depend on retrieved data that may not match user intent. In
contrast, the novel approach of agentic AI allows it to understand nuances in language, generating accurate responses
even to complex or unseen queries. Its ability to learn from vast datasets enhances precision and adaptability, making
it superior for customer interactions.

One of the biggest limitations of chatbots has been contextual coherence. Rule-based chatbots struggle to maintain
context in extended interactions due to linear scripting, leading to disjointed responses that harm customer
experience. RAG-based chatbots may produce inconsistent replies if retrieval mechanisms don't consider previous
interactions. Whereas agentic AI’s orchestration capability helps it excel at tracking conversation history,
understanding dialogue flow, ensuring responses remain contextually appropriate and coherent, significantly boosting
customer engagement.

Thus far, both rule-based and RAG-based chatbots have limited autonomous problem-solving ability. The former can't
handle problems outside their scripts while the latter provide information but can't synthesise data and prepare the
human-live problem-solving logic to solve complex issues across integrated sources such as CRMs, ERP, or IVR
systems. The agentic AI performs dynamic reasoning and decision-making, leveraging a series of autonomous agents,
analysing customer issues, considering multiple factors, and applying learned knowledge to resolve problems more
efficiently. The outcome is quicker, solution-oriented, and fluid conversations that enhance customer experience and
set new standards for efficiency and responsiveness in automated customer service.
Master agent

Customer support agent

Orchestrator agent

Customer support agent

Micro-agents

User Issue Feedback

Status updates
experience FAQ agent resolution collection Nth agent
agent
agent agent agent
Agentic AI business imperatives
Organisations managing day-to-day operations stand to gain significantly from agentic AI systems, embracing the
emerging "service-as-a-software" model. This innovative approach transforms manual labour into automated, AI-driven
services. Rather than purchasing traditional software licences or subscribing to cloud-based software-as-a-service
(SaaS), businesses can now pay for specific outcomes delivered by AI agents. For example, a company might employ
AI customer support agents like Sierra to resolve issues on their websites, paying per resolution rather than maintaining
a costly human support team. This model allows organisations to access a wider range of services – whether it’s legal
support from AI-powered lawyers, continuous cybersecurity testing by AI penetration testers, or automated CRM
management – at a fraction of the cost. This not only drives efficiency but also significantly reduces operational
overheads.

By leveraging the service-as-a-software model, businesses can automate both routine and highly specialised tasks that
were once time-consuming, required skilled professionals, and typically involved expensive software licences or cloud
solutions. AI applications with advanced reasoning capabilities can now handle complex tasks, from software
engineering to running customer care centres, enabling companies to scale their operations without a proportional
increase in cost. This transition expands the services available to organisations of all sizes, freeing them to focus on
strategic priorities while AI systems manage the operational burden. Adopting these AI-driven services positions
businesses to stay competitive in an ever-evolving marketplace8.

Transitioning from copilot to autopilot models

Service-as-a-software represents an outcome-focused, strategic shift, enabling organisations to transition from their
current state to operating in "copilot" and ultimately "autopilot" modes. Sierra, for instance, offers a safety net by
escalating complex customer issues to human agents when necessary, ensuring a seamless customer experience.
While not all AI solutions offer this built-in fallback, a common strategy is to initially deploy AI in a "copilot" role
alongside human workers. This human-in-the-loop approach helps organisations build trust in AI capabilities over time.
As AI systems demonstrate their reliability, businesses can confidently transition to an "autopilot" mode, where AI
operates autonomously, enhancing efficiency and reducing the need for human oversight. GitHub Copilot is a prime
example of this, assisting developers and potentially automating more tasks as it evolves.

Outsourcing work through AI services

For organisations with high operational costs, outsourcing specific tasks to AI services that guarantee concrete
outcomes is an increasingly attractive option. Take Sierra, for example: businesses integrate Sierra into their customer
support systems to efficiently manage customer queries. Instead of paying for software licences or cloud-based
services, they pay Sierra based on the number of successful resolutions. This outcome-based model aligns costs
directly with the results delivered, allowing organisations to harness AI for specific tasks and pay solely for the
outcomes achieved.

This shift from traditional software licences or cloud SaaS to service-as-a-software is transformative in several ways:

Targeting service profits: Traditional SaaS focused on selling user seats, whereas service-as-a-software taps into
service profit pools, delivering solutions that focus on specific business outcomes.

Outcome-based pricing: Instead of charging per user or seat, service-as-a-software adopts a pricing model based on
the actual outcomes achieved, directly aligning costs with results.

High-touch delivery models: Service-as-a-software offers a top-down, highly personalised approach, providing
trusted, tailored solutions that meet the specific operational needs of businesses.
Why should organisations consider early
adoption and avoid being late movers?

Early adopters Late movers

Set industry benchmarks

Market Struggle to catch up and miss out on
and gain first-mover market advantage.
position creating competitive advantage.

Leverage AI to innovate business Slow to innovate business processes and

Innovation processes, deploy the AI solutions take full advantage of AI solutions to create
effectively and create differentiation. differentiation.

Build deeper customer relationships

Customer Play catch-up to match the personalised
through personalised and newer
relationships services of early adopters.
experiences.

Operational Streamline operations and reduce Higher lost opportunity cost due to late entry
efficiency operational cost early on. and adoptions.

Benefit from the initial learning curve and Miss out on early learning opportunities and
Learning curve
shape industry standards. industry influence.

Increase market share and profitability

Market share Struggle to achieve similar market share.
through early adoption.

Barriers to Create barriers for competitors through Face higher barriers to entry due to
entry deep AI integration. established competitors.

Pay relatively higher cost of entry and

Pay relatively lower cost of entry and lower
Cost to entry iterative test-and-learn due to new AI
learning and experiments.
solutions.
Real-world success stories
Catalysing change across all industries

Manufacturing: Siemens AG

Siemens transformed its maintenance operations by deploying AI models that analyse sensor data from machinery. The
system predicts equipment failures before they occur, scheduling maintenance proactively. The multimodal framework
processes data from various sources – vibration, temperature, and acoustic signals – providing a holistic view of
equipment health and proactive maintenance orchestrated by the agentic AI models.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Regression and deep ● Savings: Reduced maintenance ● Enhanced equipment reliability
learning models costs by 20% ● Improved worker safety
● Platforms: Siemens ● Revenue growth: Increased
MindSphere9 production uptime by 15%
● Tools: Scikit-learn, TensorFlow,
Keras, IoT sensors

Healthcare: Mayo Clinic

By integrating AI into its radiology workflows, Mayo Clinic allows for quicker and more accurate diagnoses. The multimodal
AI processes imaging data alongside patient history and lab results, offering comprehensive insights that aid radiologists in
decision-making, automating documentation and process automation across the radiology value chain.

Technology stack: Financial impact: Non-financial benefits:

● AI Models: Regression and ● Efficiency gains: Reduced ● Improved diagnostic accuracy

Convolutional Neural Networks diagnostic times by 30% ● Enhanced patient outcomes
(CNNs) models ● Cost reduction: Lowered
● Frameworks: NVIDIA Clara unnecessary procedures by
platform10 15%
● Tools: Scikit-learn, PyTorch,
Medical Imaging Data

Finance: JPMorgan Chase

JPMorgan’s Contract Intelligence (COiN) platform uses AI to analyse legal documents, extracting key data points in
seconds. The multimodal framework interprets complex legal language, images, and tables, streamlining a process that
once took thousands of human hours.

Technology stack: Financial impact: Non-financial benefits:

● AI models: NLP with Generative ● Savings: Saved 360,000 hours ● Enhanced accuracy in
Pre-trained Transformers (GPT) of manual review annually document analysis
● Frameworks: COiN platform11 ● Risk mitigation: Significantly ● Improved employee productivity
● Tools: Python, Hadoop reduced compliance risk
 Retail: Amazon

Amazon leverages AI to analyse browsing behaviour, purchase history, and even visual preferences. Multimodal AI models
generate personalised recommendations, orchestrate tasks across order fulfilment value chains, and enhance the
shopping experience to drive sales.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Regression and deep ● Revenue boost: Increased sales ● Enhanced customer satisfaction
learning Models by 35% through personalised ● Increased engagement time on
● Frameworks: Amazon recommendations and one-click the platform
Personalise12 and Amazon order fulfilment
Order Fulfilment ● Customer retention: Improved
● Tools: AWS SageMaker loyalty rates by 20%

Transportation and logistics: DHL

DHL utilises AI models to predict and orchestrate shipping demands, optimise routes, and manage warehouse operations.
The system processes data from various sources, including traffic patterns, weather conditions, and order volumes.

Technology stack: Financial impact: Non-financial benefits:

● AI models: ML models and route ● Cost savings: Reduced ● Enhanced customer satisfaction
optimisation algorithms operational costs by 15% ● Reduced carbon footprint
● Frameworks: DHL Resilient ● Efficiency gains: Improved
supply chain platform13 delivery times by 20%
● Tools: IoT devices, ML models

Energy: BP (British Petroleum)

BP uses AI to analyse seismic data, generating 3D models of subterranean structures. The multimodal approach combines
geological, geophysical, and historical data to identify favourable drilling sites and orchestrate drilling equipment settings
for optimal outcomes.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Regression and ● Savings: Reduced exploration ● Reduced environmental impact

GenAI models costs by 20% ● Improved safety measures
● Frameworks: Azure cloud ● Revenue growth: Increased
services14 successful drilling operations by
● Tools: Microsoft AI 15%
 Education: Pearson

Pearson’s AI models tailor educational content to individual learner needs, adjusting difficulty levels and content types
based on performance and engagement data.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Adaptive learning ● Revenue increase: Boosted ● Improved student outcomes

algorithms course enrollment by 25% ● Enhanced user engagement
● Frameworks: Multimodal content ● Cost reduction: Lowered
delivery systems15 content development costs by
● Tools: Python, TensorFlow 15%

Media and entertainment: Netflix

Netflix uses AI models to recommend and orchestrate content by analysing viewing habits, ratings, and even visual
content features. The multi-modal AI ensures that users find content that resonates with their preferences, keeping them
engaged.

Technology stack: Financial impact: Non-financial benefits:

● AI models: ML and GenAI ● Subscriber growth: Increased ● Personalised user experiences

models retention rates by 10% ● Improved content strategy
● Frameworks: Netflix multimodal ● Revenue boost: Enhanced
user interaction analysis16 engagement leading to higher
● Tools: AWS, Apache Spark subscription renewals

Telecommunications: AT&T

AT&T’s AI models analyse and orchestrate network performance data and customer interactions to optimise network
operations and personalise customer service through chatbots.

Technology stack: Financial impact: Non-financial benefits:

● AI models: ML for network ● Cost savings: Reduced ● Enhanced network reliability

analytics operational expenses by 15% ● Improved customer satisfaction
● Frameworks: Edge computing ● Revenue growth: Improved
with multimodal data inputs17 upselling through personalised
● Tools: AI chatbots, data offers
analytics platforms
 Government and public sector: Singapore Government

Singapore utilises AI models to orchestrate and manage traffic flow, energy consumption, and public safety. The
multi-modal system processes data from various sensors and citizen feedback mechanisms to make real-time decisions.

Technology stack: Financial impact: Non-financial benefits:

● AI models: ML and GenAI ● Efficiency gains: Reduced ● Improved public services

models administrative costs by 25% ● Enhanced quality of life
● Frameworks: Smart Nation ● Economic growth: Attracted for citizens
platform18 US$12 billion in foreign
● Tools: IoT sensors, cloud investment
computing
Real-world success stories
Innovation within business functions

Human resources: Unilever

Unilever uses AI to screen candidates by analysing video interviews and responses, allowing recruiters to focus on the
most promising applicants.

Technology stack: Financial impact: Non-financial benefits:

● AI models: NLP and facial ● Cost reduction: Saved over ● Enhanced diversity in hiring
recognition algorithms US$1 million annually in ● Improved candidate experience
● Frameworks: Multimodal recruitment costs
candidate assessment ● Efficiency gains: Reduced hiring
platforms19 time by 75%
● Tools: HireVue AI platform

Customer service: Bank of America

Erica, an AI virtual agent, handles over a million customer queries daily – including snapshots of month-to-date spending
and flagging recurring charges – providing instant assistance and freeing human agents to tackle more complex issues.

Technology stack: Financial impact: Non-financial benefits:

● AI models: GenAI for ● Cost savings: Reduced ● Improved customer satisfaction

conversational interfaces customer service costs by 10% ● 24/7 customer support
● Frameworks: Multimodal ● Revenue growth: Increased availability
customer interaction platforms20 product cross-selling by 5%
● Tools: Erica, the virtual assistant

Marketing: Coca-Cola

Coca-Cola uses AI to generate marketing content, analyse consumer trends, and personalise advertising, resulting in more
effective campaigns.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Generative ● Efficiency gains: Reduced ● Innovative marketing strategies

Adversarial Networks (GANs) content creation time by 50% ● Enhanced customer
● Frameworks: Multimodal data ● Revenue increase: Boosted engagement
analysis for consumer insights21 campaign ROI by 20%
● Tools: Custom AI platforms
 Supply chain management: Walmart

Walmart employs AI to predict product demand, optimise stock levels, and streamline logistics, ensuring products are
available when and where customers need them.

Technology stack: Financial impact: Non-financial benefits:

● AI Models: Predictive analytics ● Cost Reduction: Decreased ● Reduced waste

for demand forecasting inventory costs by 15% ● Enhanced supplier relationships
● Frameworks: Multi-modal data ● Revenue Growth: Improved
integration from sales, weather, product availability leading to
and events22 higher sales
● Tools: Data lakes, Machine
Learning models

Research and development: Insilico

Medicine

Insilico Medicine, a biotechnology company focused on longevity, has developed inClinico, an AI platform that
predicts phase II clinical trial outcomes to enhance drug discovery and development.

Technology stack: Financial impact: Non-financial benefits:

● AI Models: In-house-developed ● Cost Reduction: 35% ● Accelerated drug discovery and

multimodal foundation model nine-month ROI in an clinical trials process
● Platforms: Multi-modal investment application ● 79% accuracy for clinical trials
integration of omics, text, ● Time Efficiency: Reduced drug
clinical trials, small molecule development time
properties, and disease
targets23
● Tools: Transformer-based,
in-house-trained AI model and
platform

Legal: Hogan Lovells

The AI platform analyses large sets of contracts and legal documents, extracting critical information, and identifying
risks.

Technology stack: Financial impact: Non-financial benefits:

● AI models: NLP and ML ● Efficiency gains: Increased ● Improved accuracy

● Frameworks: Kira Systems review speed by 40% ● Enhanced client satisfaction
platform with multimodal data ● Cost savings: Reduced billable
processing24 hours for clients
● Tools: Kira AI
 Procurement: Coupa

Coupa’s AI-driven spend management platform optimises supplier selection, contract management, and spend analytics,
transforming procurement processes into a strategic function.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Predictive analytics, ● ROI: Achieved an impressive ● Increased compliance and risk
machine learning, and spend 276% return on investment management.
forecasting. (ROI). ● Improved supplier performance
● Frameworks: Coupa ● Efficiency gains: Reduced and relationships
Source-to-Pay, Coupa Business procurement cycle and
Spend Management (BSM).25 significantly enhancing process
● Tools: Cloud computing, speed.
advanced sourcing optimisation,
real-time spend visibility.

IT Operations: Microsoft

Microsoft uses AI to monitor IT systems, predict failures, and automate support tickets, ensuring seamless operations.

Technology stack: Financial impact: Non-financial benefits:

● AI Models: Anomaly detection ● Cost Savings: Reduced IT ● Enhanced employee productivity

and predictive maintenance support costs by 20% ● Proactive issue resolution
algorithms ● Efficiency Gains: Improved
● Frameworks: Azure AI with system uptime by 15%
multi-modal data inputs26
● Tools: AI chatbots, Monitoring
tools

Sales: Salesforce

Salesforce’s AI analyses customer interactions, market trends, and sales data to provide actionable insights for sales
teams.

Technology stack: Financial impact: Non-financial benefits:

● AI models: Predictive analytics ● Revenue growth: Increased ● Improved customer relationships

with ML sales by 15% ● Enhanced decision-making
● Frameworks: Salesforce Einstein ● Efficiency gains: Reduced sales
with multimodal data cycle times by 25%
processing27
● Tools: CRM systems
Key GenAI agentic tools and their
differentiation

Commercial solutions Open-source solutions

LangGraph28 AutoGen29

● Target audience: Startups and established ● Target audience: Developers and researchers
enterprises ● Open-source framework: Facilitates cooperation
● Support: Offers robust customer support and among multiple AI agents
professional services ● Simplification: Orchestrates, automates, and
● Integration: Seamlessly integrates with existing optimises complex LLM workflows
enterprise systems ● Human-in-the-loop: Supports human-in-the-loop
● Customisation: High level of customisation and workflows for enhanced performance
control over workflows ● Community-driven: Encourages innovation and
● Features: Advanced features like statefulness collaboration within the community
(having a perfect memory or knowledge of previous
calls or requests), streaming support, and
moderation loops

CrewAI30 AutoGPT31

● Target audience: Fortune 500 companies and large ● Target audience: AI enthusiasts and developers
enterprises ● Autonomous AI agent: Executes tasks
● Ease of use: Provides no-code tools and templates independently using GPT-4 architecture
for quick deployment ● Task management: Breaks down complex goals
● Deployment options: Supports both self-hosted into manageable sub-tasks
and cloud deployments ● Capabilities: Utilises internet access and code
● Support: Comprehensive support and maintenance execution for task completion
services ● Versatility: Applied in various domains like content
● Efficiency: Designed for handling complex, creation and customer service
multi-agent tasks efficiently ● Popularity: Rapidly growing open-source project
with a strong community

When deciding between commercial vs open-source agentic AI tools, consider your organisation’s needs,
upstream/downstream integration capabilities, and accessibility to resources to build, deploy, and manage
these solutions.

Commercial solutions such as LangGraph and CrewAI offer robust support, seamless integration, and
advanced features, making them suitable for complex, large-scale deployments. Conversely, open-source
solutions like AutoGen and AutoGPT are excellent choices for rapid prototyping and proof-of-concept
development, providing flexibility, community-driven innovation, and low cost of entry for technology decision
makers and developers.

The agentic AI tools ecosystem is expected to witness a rapid surge over the next few quarters. Commercial
solutions will likely continue to enhance their enterprise capabilities, focusing on a wide range of integration
options, security, and developer-friendly features. Meanwhile, open-source tools will see increased community
contributions, leading to rapid innovation in depth and coverage of agentic AI features and increased adoption.
As commercial and open-source AI solutions evolve, organisations should stay agile, leveraging the strengths
of both types to remain competitive and innovative.
Formulating your GenAI strategy and crafting
the AI capability roadmap that works for
your business
A vision without execution is hallucination—align your GenAI strategy with actionable plans and meticulous execution.

Let's explore how to effectively integrate these principles into your AI roadmap:

Step 1 Step 2 Step 3

Assess
Vision Assess
Meticulous
alignment capabilities
execution

Step 4
Step 6 Step 5 Scale
up
Organisational Risk
change management

Vision alignment

● Define clear objectives: What do you aim to achieve – cost reduction, revenue growth, customer satisfaction, or
building an economic moat?
● Align AI initiatives with business goals: Ensure that AI projects are underpinned by your company's strategic
objectives. Whether it's cost reduction, increasing revenue, customer satisfaction, or creating a competitive
advantage, aligning AI efforts with business goals ensures relevance and maximises impact.
● Secure executive sponsorship: Having support from top management is crucial for securing resources and
driving organisational change. Executive sponsorship can also help align AI initiatives with broader business
strategies.
● Stakeholders buy-in: Ensure executive and departmental alignment.
● Start with high-impact use cases: Identify areas where AI can deliver significant value quickly. Prioritise projects
that address pressing challenges or offer substantial benefits, such as cost reduction or revenue growth, to
demonstrate AI ROI early on.
● Seek expert advice: Consult with AI experts or hire consultants to formulate your AI strategy and help you in
making informed decisions.

Assess capabilities

● Technology infrastructure: Is your IT environment ready for AI integration?

● Platform options: Weigh-in commercial and open-source AI solutions and make build-vs-buy decisions based on
your organisation’s requirements, budget, and technical expertise.
● Consider integration: Ensure the chosen platform can integrate seamlessly with your existing systems and
workflows, both upstream and downstream.
● Data readiness: Do you have access to quality, multimodal data?
● Talent pool: Do you have the skills in-house, or will you need external expertise?

Meticulous execution

● Start small: Begin with small pilot projects to test the effectiveness of agentic AI in your business environment.
● Measure success: Define clear metrics for success and monitor the performance of the pilot projects. Gather
feedback from stakeholders and make necessary adjustments.
● Agile methodology: Be flexible, nimble and adaptive in your implementations.
● Iterate and improve: Use the insights gained from pilot projects to refine your approach and address any
challenges.
 Scale up

● Gradual expansion: Once the pilot projects are successful, gradually scale up the implementation of agentic AI
across more areas of your operations.
● Ensure support: Provide adequate training and support to your team to ensure a smooth transition and adoption of
the new technology.
● Monitor and optimise: Continuously monitor the performance of agentic AI systems and optimise them for better
results.

Risk management

● Ethical considerations: Address potential biases and compliance issues.

● Security protocols: Protect sensitive data and align AI governance with national and global standards.

Organisational change

● Educate and upskill: Begin by familiarising your workforce with the core concepts of data and AI. Understand
what it is, how it works, and its potential applications in your organisation, business function and/or industry32.
● Foster innovation: Encourage a culture of innovation within your organisation by promoting experimentation and
collaboration.
● Adapt and evolve: Be prepared to adapt your strategies and processes as the technology evolves and new
opportunities arise.
● Stay informed: Keep up with the latest developments and trends in AI by reading industry reports, inviting
experts to all-hands sessions, attending conferences, and participating in webinars.
Top 10 do’s and don’ts for maximising ROI
from AI investments
Avoid the GenAI hype trap—focus on pragmatic steps that deliver real value.

Do’s

● Ensure a customer-centric approach: Always prioritise the end-user experience, eventually it

pays off in both financial and non-financial results.

● Conduct thorough research: Before implementing AI solutions, research the available

technologies to find the best fit for your business needs. Understand the capabilities and limitations
of agentic AI to set realistic expectations.

● Start with small projects: Begin with pilot programmes to test the effectiveness of AI solutions.
Small-scale implementations allow you to measure impact and make adjustments before a full-scale
rollout.

● Monitor performance and iterate: Regularly track the performance of your AI systems using key
metrics aligned with your business goals. Use this data to refine models, adjust strategies, and make
data-driven improvements over time.

● Build cross-functional teams: Assemble teams that include members from various departments,
such as IT, operations, finance, and marketing. Cross-functional collaboration ensures that AI
initiatives are well-rounded and consider different perspectives and expertise.

● Invest in employee training: Equip your team with the necessary skills to work alongside AI
systems. Training ensures smooth integration and helps employees leverage AI tools effectively.

● Invest in quality data: High-quality data is the backbone of effective AI solutions. Invest in data
cleaning, integration, and management processes to ensure your AI systems have accurate and
reliable data to work with.

● Prioritise data security and privacy: Implement robust security measures to protect sensitive
data. Ensure compliance with relevant regulations to maintain customer trust and avoid legal issues.

● Invest in scalable AI platforms: Choose AI platforms and tools that are scalable and can grow
with your business needs. Scalable solutions allow you to expand AI capabilities without significant
additional investments.

● Invest in continuous learning: Stay curious and updated with AI advancements and industry
trends.
Top 10 do’s and don’ts for maximising ROI
from AI investments
Avoid the GenAI hype trap—focus on pragmatic steps that deliver real value.

Don’ts

● Ignore customer feedback: Pay attention to how your customers interact with AI solutions. Use
their feedback to refine and enhance the user experience.

● Underestimate complexity: AI projects are not plug-and-play.

● Rush implementation: Avoid hastily integrating AI without a clear strategy. A rushed

implementation can lead to wasted resources and suboptimal results.

● Neglect human oversight: While AI can automate many tasks, human oversight remains crucial.
Maintain a balance between automation and human input to ensure quality and accountability.

● Ignore user adoption: Ensure that the AI solutions are user-friendly and meet the needs of those
who will interact with them daily. High user adoption rates lead to better data input, more accurate
outputs, and higher ROI.

● Overlook ethical considerations: Be mindful of the ethical implications of AI use. Ensure your AI
systems are designed to prevent biases, respecting privacy laws, promote fairness and
transparency. Adhere to ethical guidelines and legal regulations related to AI use.

● Ignore change management: Prepare your workforce for AI adoption through training and change
management programmes. Educated employees are more likely to embrace AI tools, leading to
better utilisation and ROI.

● Underestimate costs: Be realistic about the investment required for AI integration, including
infrastructure, maintenance, and training costs. Plan your budget accordingly to avoid financial strain.

● Ignore partnerships: Collaborate with trusted technology providers, consultants, AI experts, and
academic institutions. External expertise can accelerate implementation, provide valuable insights,
and help avoid common pitfalls.

● Overlook for long-term sustainability: Develop a long-term AI strategy that considers future
needs and technological advancements. Sustainable planning ensures that your AI investments
continue to deliver value over time.
Looking ahead
By harnessing the unprecedented capabilities of agentic AI systems, both government entities and organisations can
achieve significant efficiency gains, enhanced customer experiences, and superior business outcomes. While agentic
AI will play a transformative role in both sectors, their specific objectives, contexts, and goals will shape the distinct
applications and their benefits.

Government entities can prioritise large-scale initiatives such as policymaking, governance, public welfare, economic
stability, and sustainability, leveraging agentic AI to orchestrate complex systems. In contrast, organisations aiming
for profitability growth, cost optimisation, and competitive advantage can focus on developing agentic AI point
solutions to address specific challenges within defined domains.

Most entities are expected to begin by experimenting with low hanging use cases. A smaller number will see the vast
opportunity window with agentic AI solutions and adopt a strategic approach, recalibrating AI strategies to fully
harness agentic AI solutions across a broader spectrum of business use cases and processes. Only a handful—like
Amazon, Google, Meta etc.—will embrace an AI-first mindset, reimagining products, services, and processes to
redefine value creation mechanisms. In this transformation, agentic AI systems will take the lead role with humans as
co-pilots, optimising speed, accuracy, contextual coherence, and cost-efficiency. Human oversight will evolve,
shifting focus toward more strategic planning and innovations rather than operational management.

The shift from human-driven, labour-intensive processes to AI-managed operations will see autonomous agents
handling tasks with unprecedented speed, precision, and adaptability. This transformation will not only reduce costs
but also unlock new revenue streams and growth opportunities, allowing businesses and governments to deliver
services faster and at a much larger scale.

As agentic AI systems become more integrated, they will redefine how we work, pushing the boundaries of possibility
and enabling a smarter, more agile world. The future of agentic AI is nearer than anticipated, propelled by rapid
technological advancements. However, realising its full potential requires greater commitment and adoption across
both government and industry.

Leaders at the forefront of AI adoption aren't just leveraging technology—they're redefining what's possible. Success
in this arena won't be accidental; it demands strategic vision, meticulous planning, and relentless execution. For
C-suite executives and senior leaders, embracing agentic AI is not just an option but a strategic imperative to stay
ahead in an increasingly competitive and AI-driven world.

The future belongs to those who prepare for it today—make AI the

cornerstone of your strategic arsenal.

Get in touch

Akif Kamal Dr. Mohammad Tanvir Ansari Kaushal Chapaneri

Partner Director Senior Associate
Technology Consulting Technology Consulting Technology Consulting
PwC Middle East PwC Middle East PwC Middle East
[email protected] [email protected] [email protected]
References
1. Agentic AI: A deep dive into the future of automation | VentureBeat
2. GenAI will be worth trillions. Here's a roadmap for harnessing it | World Economic Forum
3. What is Agentic AI? Key Benefits and Use Cases
4. Agentic AI: The Next Evolution of Enterprise AI | Moveworks
5. From Bots to Agents: The Next Great Leap Forward Is Autonomous AI
6. Agentic AI: The Next Evolution of Enterprise AI | Moveworks
7. From Bots to Agents: The Next Great Leap Forward Is Autonomous AI
8. Generative AI's Act o1: The Reasoning Era Begins | Sequoia Capital
9. Insights Hub
10. Mayo Clinic’s Healthy Model for AI Success
11. How JPMorgan Chase’s COIN is Revolutionizing Financial Operations with AI | by THE AI ZONE | Medium
12. Recommender System, Recommendation Engine - Amazon Personalize
13. Gen AI - DHL - United Arab Emirates
14. Working better, safer, faster: how AI can help the energy transition | News and insights
15. AI-powered study tool | Digital Learning Platforms | Pearson UK
16. Machine Learning Platform
17. AT&T Labs | Our Work | Analytics and AI-based Automation
18. Artificial Intelligence in Singapore | IMDA
19. The Amazing Ways How Unilever Uses Artificial Intelligence To Recruit & Train Thousands Of Employees |
Bernard Marr
20. Erica - Virtual Financial Assistant | Bank of America
21. Coca-Cola Scaling GenAI Marketing Campaigns With Digital Twins | Consumer Goods Technology
22. Walmart’s Element: A machine learning platform like no other
23. Prediction of Clinical Trials Outcomes Based on Target Choice and Clinical Trial Design with Multi‐Modal Artificial
Intelligence - Aliper - 2023 - Clinical Pharmacology & Therapeutics
24. Hogan Lovells Enhances Transactional Services with Kira
25. The Total Economic Impact™ Of Coupa For Source-To-Pay
26. Rethinking device management internally at Microsoft with AI - Inside Track Blog
27. Salesforce Einstein AI Solutions
28. LangChain Academy
29. AutoGen
30. CrewAI
31. AutoGPT
32. Agentic AI Overview, Applications and Use Cases

About PwC At PwC,

our purpose is to build trust in society and solve important problems. We’re a network of firms in 152 countries with nearly 328,000 people
who are committed to delivering quality in assurance, advisory and tax services. Find out more and tell us what matters to you by visiting us
at www.pwc.com.

Established in the Middle East for 40 years, PwC has 24 offices across 12 countries in the region with around 8,000 people.
(www.pwc.com/me).

PwC refers to the PwC network and/or one or more of its member firms, each of which is a separate legal entity. Please see www.pwc.com/
structure for further details. © 2024 PwC. All rights reserved