By Martin H. Morrissette, Sirocco CMO – Book time with me
As artificial intelligence continues to evolve, 2025 promises new challenges and opportunities, particularly with the rise of agentic AI and the further refinement of generative AI applications. While much of 2024 revolved around experimentation, IT leaders now have a clearer picture of what works (and what doesn’t) when scaling AI to achieve real business outcomes. According to the 2024 Gartner Hype Cycle for Emerging Technologies, generative AI is nearing the “Trough of Disillusionment” after peaking in inflated expectations. This shift reflects the practical challenges of implementing generative AI at scale, with businesses now focusing on tangible results. Meanwhile, agentic AI (Autonomous Agents) is rising up the “Innovation Trigger” slope, indicating growing interest in its potential.
Generative AI has already been transformative, enabling businesses to create outreach emails, source code, marketing strategies, and more. However, it typically requires human activation and decision-making to implement its outputs effectively. Agentic AI builds on this by focusing on autonomous execution, moving beyond generation to actively drive business outcomes. While generative AI excels at producing content like text, images, and code, agentic AI takes it a step further by emphasising action and autonomy. For instance, generative AI might draft an outreach email for a sales lead, but an agentic AI could autonomously send that email, track the recipient’s response, and schedule a follow-up meeting based on calendar availability. Similarly, while generative AI can design a marketing campaign, agentic AI can implement it by deploying ads, monitoring performance, and reallocating budgets in real-time. This ability to execute without human prompts makes agentic AI a transformative force across industries. These intelligent agents can autonomously activate workflows, leverage generative outputs, and execute tasks without constant human intervention. For example, an AI agent could independently manage customer onboarding, from creating accounts to scheduling meetings – tasks that previously required significant human effort. The key distinction between generative and agentic AI is their utility. While generative AI produces valuable outputs, agentic AI provides more measurable business value by directly reducing human costs and enabling greater efficiency. Additionally, agentic AI’s scalability allows businesses to handle seasonal or unexpected demand surges, such as Black Friday sales or disaster response, with cost-effective elasticity.
While the debut of ChatGPT in the Fall of 2022 was revolutionary in showcasing AI’s potential for natural, scalable interaction, much of agentic AI, including systems like Agentforce, leans towards being evolutionary. It builds on the foundations laid by predictive and generative AI, adding autonomy to decision-making and actions. The transformative impact of these systems depends less on the technology itself and more on how effectively organisations integrate and use them to drive meaningful outcomes.
So it’s clear, Agentic AI didn’t emerge in a vacuum. It’s part of an ongoing evolution in how businesses leverage artificial intelligence to improve customer interactions and operations. Understanding this progression offers valuable context for where we stand today and where the technology is heading. So here’s a (very short) AI history lesson.
From Chatbots to advanced Agents
The journey began with simple chatbots designed to handle predefined tasks, such as answering FAQs or routing customer inquiries. While these tools were effective for basic automation, they lacked the flexibility to tackle more complex issues or adapt to nuanced customer needs. Over time, these systems gave way to more sophisticated AI copilots, capable of assisting users with real-time insights and actionable recommendations. Now, with the rise of agentic AI, we’re seeing the emergence of fully autonomous agents that can not only understand and execute complex tasks but also make decisions and continuously improve through feedback loops.
Integration with enterprise data
The power of agentic AI lies in its ability to connect with and learn from enterprise data. By leveraging proprietary datasets and metadata, these systems can offer highly personalised solutions tailored to specific customer needs. This integration enhances the relevance and accuracy of AI-driven interactions, turning data into actionable insights and enabling businesses to deliver a superior customer experience. More on that in the next section.
Advancements in AI capabilities
The transition from basic automation tools to autonomous AI agents represents a significant leap in technology. Today’s agentic AI systems go well beyond simple task execution. They reason, adapt, and act as collaborative partners in achieving business goals. These advancements underscore the increasing sophistication of AI and hint at a future where these agents will play an even more pivotal role in driving innovation and efficiency.
The foundation of your Agentic AI success? Integration
Just as your overall AI strategy relies on a robust data foundation, the success of your agentic AI initiatives hinges on seamless integration with your existing IT infrastructure. Agentic AI agents, with their ability to automate tasks and make decisions, require access to real-time data and the ability to interact with various systems across your enterprise. Consider your customer service operations: an AI agent needs access to customer profiles, order history, product information, shipping updates, and even real-time inventory levels to effectively resolve queries. Siloed data severely limits the agent’s capabilities, leading to subpar customer experiences and increased operational costs.
True integration goes beyond simply connecting systems, of course. It requires a holistic approach that ensures your data is readily available and accessible to AI agents across all relevant departments and functions, that your data is clean, accurate, and consistent for reliable and effective AI decision-making, and that your AI agents have real-time access to data to respond quickly and effectively to changing conditions. To achieve this, prioritise integration as a core principle, treating AI agents as integral components of your IT ecosystem. This ensures seamless communication, collaboration, and data flow across all your systems. Invest in robust integration platforms that facilitate smooth data exchange and communication between your AI agents and existing systems. This will enhance the functionality, scalability, and overall effectiveness of your AI strategy. Furthermore, foster a data-driven culture by ensuring data accessibility and breaking down barriers between data sources. This will unlock the full potential of AI by providing your agents with a comprehensive and integrated view of your business. By prioritizing integration, you can ensure that your AI agents are equipped with the necessary data and connectivity to perform at their best, delivering exceptional customer experiences and driving significant business value for your organisation.
Agentic AI systems generally follow a structured process to address complex tasks and challenges. While specific implementations vary across platforms, most solutions adhere to these key steps:
- Data Gathering and Processing
Agentic AI starts by collecting and processing data from various sources, such as sensors, databases, CRMs, and user inputs through digital interfaces. This step involves extracting meaningful features, recognising objects, and identifying entities relevant to the task. For example, an agent might analyse customer data from a CRM, cross-reference it with recent purchase history, and provide tailored recommendations or insights. - Task Understanding and Solution Generation
At the heart of these systems is a large language model (LLM) acting as the reasoning engine. This orchestrator understands tasks, generates solutions, and coordinates specialised models for specific functions, such as predictive analytics, content creation, or visual processing. By leveraging techniques like retrieval-augmented generation (RAG), agentic AI systems can tap into proprietary data sources to ensure outputs are accurate and contextually relevant. - Task Execution and Integration with External Tools
Once a solution is formulated, the AI executes tasks through integration with external tools and software using application programming interfaces (APIs). To maintain precision, organisations can implement guardrails that define parameters for AI actions. For instance, an AI agent might be programmed to process routine transactions autonomously, while complex or high-stakes scenarios are escalated for human oversight. - Learning and Continuous Improvement
The ability to learn from experience is a defining feature of agentic AI. Data generated through interactions is fed back into the system, refining its models and enhancing performance over time. This feedback loop enables agentic AI to adapt to new challenges and improve decision-making capabilities, offering businesses a continuously evolving toolset for operational efficiency.
Beyond these core steps, agentic AI systems are designed to scale and adapt to a variety of organisational contexts. Advanced platforms, such as Salesforce’s Agentforce, allow businesses to customise workflows, integrate with enterprise systems, and ensure solutions align with unique requirements.
Some key AI trends shaping 2025
As AI continues to evolve, several key trends are emerging that will define the landscape in the year ahead. These trends will significantly impact the development and deployment of AI in 2025, driving innovation and shaping the future of this transformative technology:
- Integration will be paramount: Just as data strategy is crucial for AI success, seamless integration will be the cornerstone of effective agentic AI. AI agents need to interact seamlessly with existing systems and data sources across the enterprise to function effectively.
- Large action models will take center stage: While large language models (LLMs) excel at language understanding, Large Action Models (LAMs) focus on translating that understanding into real-world actions. By integrating with APIs, LAMs empower AI agents to automate complex tasks and streamline operations.
- The rise of Small Language Models: Small Language Models (SLMs) offer a compelling alternative to their larger counterparts. Their compact size makes them cost-effective, fast, and ideal for resource-constrained environments, such as edge computing and mobile devices.
- Energy efficiency will be critical: As AI’s environmental impact grows, organisations must prioritise energy-efficient models that minimise their carbon footprint. Techniques like model compression and efficient training methods will play a crucial role in reducing the environmental impact of AI.
- Multiple AI models will optimise efficiency: LLM routers will optimise AI usage by directing prompts to the most suitable model based on factors like complexity, performance requirements, and cost, leading to more efficient and cost-effective AI operations.
Several companies are at the forefront of agentic AI
Leading players are emerging at the forefront of agentic AI development, offering businesses powerful tools to automate tasks, enhance customer interactions, and drive operational efficiency. However, it’s important to remember that the names and features of AI platforms in this space are constantly evolving as new innovations emerge.
Salesforce Agentforce: Salesforce’s Agentforce enables businesses to create autonomous AI agents that integrate seamlessly with the Salesforce ecosystem. These agents can handle tasks across sales, service, marketing, and commerce, operating 24/7 to enhance efficiency and customer satisfaction. The platform offers a low-code Agent Builder, allowing users to customize agents to meet specific business needs. Additionally, Agentforce includes a Testing Center for simulating interactions and refining agent performance.
HubSpot Agent.AI: HubSpot’s Agent.AI is a network of AI agents designed to collaborate with human teams across marketing, sales, customer service, and operations. These agents assist in automating routine tasks, providing insights, and enhancing team productivity. By integrating with HubSpot’s CRM platform, Agent.AI ensures a cohesive experience for users, enabling businesses to deliver personalized and efficient customer interactions.
Pegasystems Customer Service and Sales Automation Suites: Pegasystems offers AI-powered solutions that enable the creation of intelligent chatbots capable of autonomously handling customer service and sales tasks. These AI agents can manage inquiries, process transactions, and provide personalized recommendations, all while learning from interactions to improve over time. Pegasystems’ platform emphasizes scalability and adaptability, allowing businesses to tailor AI agents to their specific requirements.
Microsoft Azure OpenAI Service and Power Automate: Microsoft provides functionalities for building and deploying intelligent agents through its Azure OpenAI Service and Power Automate platforms. The Azure OpenAI Service offers access to advanced language models, enabling the development of AI agents capable of understanding and generating human-like text. Power Automate facilitates the automation of workflows, allowing AI agents to perform tasks such as reviewing customer returns or checking supply-chain invoices autonomously. These tools are designed to integrate seamlessly with Microsoft’s suite of applications, enhancing productivity and operational efficiency.
Google Agent Builder: With its Gemini model and extensive research in AI, Google is a major force in agentic AI. Their focus on multimodal understanding and complex reasoning aligns well with the goals of agentic systems. Google Cloud provides tools for designing and developing conversational AI agents through its Agent Builder platform. These tools enable businesses to create AI agents that can engage in natural language conversations, understand context, and perform tasks across various domains. By leveraging Google’s machine learning capabilities, Agent Builder allows for the development of sophisticated AI agents that can be integrated into websites, applications, and other platforms to enhance user engagement and support.
On the research side we have a few more players. Known for its groundbreaking work with GPT models, OpenAI is actively exploring ways to enhance the agency and autonomy of AI systems. A subsidiary of Alphabet, DeepMind conducts cutting-edge research in AI, including areas like reinforcement learning and game playing, which are crucial for developing agentic capabilities. Research company Anthropic is developing AI systems with a focus on safety and interpretability, which are critical considerations for building responsible and trustworthy agentic AI.
Now what?
One thing is clear, the future of AI is here. It’s more powerful than ever and getting stronger with each release. By combining the creative potential of generative AI with the autonomous execution of agentic AI, businesses like yours can unlock unprecedented levels of innovation, efficiency, and productivity. As agentic AI continues to evolve, it’s crucial for IT leaders to prioritise integration, embrace ethical considerations, and develop a robust AI strategy. To help you navigate this exciting new era, we encourage you to contact us to Assess your AI readiness and/or Partner with our integration experts to develop a comprehensive AI strategy that aligns with your business objectives and seamlessly integrates with your existing IT infrastructure. By working together, we can help you harness the power of AI to drive meaningful business outcomes and achieve your strategic goals.
