Building a Unified AI Agent Strategy: Beyond the Single Platform (and Beyond Just Technology)

TL;DR

In this recent webinar, Vala Afshar, Chief Digital Evangelist at Salesforce, emphasizes the crucial role of a unified AI agent strategy in achieving an autonomous enterprise.  “The superpowers of AI and autonomous business will elude you unless you can map your process in real-time. Identify where you need to have AI agents implemented, and then try to do that, not just for a particular task, but tasks that extend beyond your line of business.”

Ronald Starreveld, Distinguished Enterprise Architect at Salesforce, agrees. In the Salesforce Posse podcast  “What does AI strategy mean to you? In a business context, it’s really to identify the reason why you’re considering AI. So, the use cases implementation and the tools that you can select.” 

The prioritization of AI agents depends on precise real-time business process mapping that is automatically generated using Elements.cloud Configuration Mining. Emerging standards like MCP and A2A are crucial for integrating diverse AI agent platforms, reducing vendor lock-in. The AI strategy necessitates robust governance, encompassing vision, risk management, process-led design, and education, to ensure successful and impactful AI adoption.

Now is the time to get started building agents. And Elements.cloud has launched free Agent Designer capabilities, which support the design, build, and governance of reliable agents. 

AI Agents are real

As CIOs, the buzz around AI agents isn’t just hype; it’s a critical inflection point for how organizations operate and compete. Many in the industry might suggest picking a single AI agent platform and going all in. However, that approach is often too narrow for the complex, integrated enterprise. An AI strategy isn’t about choosing one platform; it’s about building a cohesive ecosystem that leverages the strengths of multiple agent platforms, seamlessly integrated to drive maximum business value.

The idea of a monolithic “one-stop shop” for all AI needs is tempting, but it’s often unrealistic and limiting. Just as an organization wouldn’t choose a single cloud provider for every application, selecting one AI platform when they are evolving so quickly can stifle innovation and create unnecessary dependencies. Instead, the focus should be on an integrated, multi-platform approach, a strategy now more feasible than ever thanks to evolving standards and integration capabilities – MCP and A2A. More on this later.

Starting with the “Why”: A Strategic Imperative

Before even considering a single tool or pilot project, an AI strategy must begin with the “why”. As Ronald Starreveld, Distinguished Enterprise Architect at Salesforce, pointed out on a recent Salesforce Posse podcast, “What does AI strategy mean to you? In a business context, it’s really to identify the reason what you’re considering AI. So, the use cases implementation and the tools that you can select.” 

He notes that many companies rush into AI adoption driven purely by technology curiosity, leading to “pet projects” or proofs of concept that fail to gain business buy-in or identify true value. “Don’t do this in isolation,” advises Starreveld.

Organizations must clearly define their business objectives and understand where AI genuinely fits into their overarching strategy. This isn’t just about technical implementation; it’s about transforming the organization to become a digital company. Without this foundational understanding, even the most advanced AI tools will struggle to deliver tangible ROI.

Prioritizing the Best Processes to Agentify

In a recent webinar, Vala Afshar, Chief Digital Evangelist for Salesforce, suggested agents were like pro-athletes.  “Imagine an AI agent performing like a professional athlete, consistently delivering peak performance when it matters most. That’s the aspiration for many organizations diving into the world of AI,” he comments. But as he said, “How do you actually get there?”

He boiled it down to two critical factors: capability and reliability. And the secret weapon? Your business process. Think of it this way, on a two-axis graph:

• Champion: High capability, high reliability. These are your consistent winners, performing complex tasks flawlessly every time.
• Prodigy: High capability, but lower reliability. They show flashes of brilliance on difficult tasks but aren’t always consistent.
• Workhorse: Low capability, high reliability. They reliably handle simpler, repetitive tasks without fail.
• Generalist: Low capability, low reliability. These are your all-rounders, but not specialists.

So, where do your AI agents need to land on this spectrum? The answer is revealed in the clarity and accuracy of your business process maps. They help you identify the agentic opportunities and the type of agent that you need.

Auto-generated Process Mapping – Configuration Mining

When you have all your org metadata and dependencies that make your operations tick, you can get Elements.cloud to build a living, breathing blueprint (process diagram) that can identify opportunities to build truly effective AI agents.

Consider a customer-facing process where reliability is paramount. A small error could lead to significant reputational damage. In such a scenario, your process map will clearly indicate the need for a “Champion Agent” – one that is both highly capable and unfailingly reliable. Conversely, a less critical internal task might only require a “Workhorse” agent, capable of consistent, lower-complexity execution.

Without a precise, dynamically updated process map, you’re essentially flying blind. You won’t be able to determine the exact level of capability and reliability needed for each agent, leading to inconsistent performance and potential risks to your brand and market share.

Many companies operate with significant “blind spots” in how they run their business. They lack a clear, accurate understanding of:

• Their end-to-end business processes
• The understanding of intricate interdependencies among their data and its governance.
• A live, up-to-date metadata library that allows for rapid change.

This lack of visibility severely limits their ability to compete effectively in an increasingly AI-driven landscape. If you can’t see precisely how your business operates, how can you expect to build AI agents that seamlessly integrate and enhance those operations?

The Path Forward: Strategy to Process to Champion Agents

Your capacity to compete in the age of AI is directly tied to the accuracy of your business process visibility and your agility in deploying agents based on specific capability and reliability needs. This isn’t a “nice-to-have”; it’s a fundamental requirement for success. Investing in robust business process analysis and maintaining a dynamic metadata library are no longer optional – they are the strategic plays that will enable you to build champion agents and accelerate ahead of the competition.

The AI strategy should not be longer technology-driven, but is grounded in the needs of the business: the processes and the types of processes that are going to be agentified. This helps define the AI platforms that are required. With the new AI standards of MCP and A2A the ability to choose a multi-platform strategy is a lower risk decision. The costs of integration are massively reduced, which also makes the choice of any vendor less strategic. There are costs to skill up on a particular agent platform, but the technology lock-in is reduced. This is crucial when the AI platform space is evolving so quickly.

The Power of Integration: MCP and A2A Standards

Key to this multi-platform strategy are advancements like the Model-to-Cloud-Platform (MCP) and Agent-to-Agent (A2A) standards. These emerging standards are game-changers, making it far easier to achieve true interoperability between different AI models and agent systems.

• MCP allows models to be trained on one platform and deployed across various cloud environments, maximizing flexibility and avoiding vendor lock-in. This means organizations can leverage cutting-edge models from developer-centric platforms like OpenAI or AWS Bedrock, and then seamlessly integrate them with low-code application environments.
• A2A standards enable different AI agents to communicate and collaborate with each other. Imagine a scenario where a pre-built customer service agent handles initial queries, and then intelligently hands off complex issues to a role-specific internal agent that leverages deep enterprise knowledge. This level of collaboration between specialized agents unlocks unprecedented efficiencies, especially as agents move towards more autonomous capabilities. 

This integrated approach means organizations no longer have to compromise. They can cherry-pick the best-of-breed solutions for specific needs without worrying about data silos or integration nightmares. Ronald Starreveld highlights Salesforce’s flexibility in supporting multiple LLMs, noting that “for this specific question or for this specific [use] case, I’m better off using this LLM”.

Understanding Types of AI Agent Platform

A robust AI strategy will probably involve strategically deploying different types of AI agent platforms across the organization. Here are the board categories:

1. Low-Code Agent Builder Platforms (e.g., Salesforce Agentforce, ServiceNow Agents, Microsoft Power Virtual Agentst): These serve as primary tools for empowering business users and departmental IT teams. Platforms like Salesforce Agentforce, with their drag-and-drop interfaces and deep integration with existing CRM data, are ideal for rapidly automating customer-facing processes and internal support workflows. They enable “Agentblazers” to quickly build solutions tailored to specific business units, leveraging existing Flows, Apex, and Prompt Templates.
2. Developer-Centric Builder Platforms (e.g., Google Cloud AI Platform / Vertex AI, Azure AI Platform, OpenAI API, AWS Bedrock): For more complex, custom AI needs, particularly those requiring highly specialized models or advanced data science, development teams can leverage these robust platforms. This gives organizations the flexibility to build unique AI capabilities that provide true competitive differentiation, without being constrained by the out-of-the-box functionalities of low-code tools.
3. Role-Specific Agents (e.g., Elements.cloud, Writer.com, Copado Agents,  Midjourney/DALL-E): Organizations can actively integrate and utilize specialized applications that have AI tools embedded.. Ronald Starreveld notes that AI often maximizes its value when “supporting somebody who’s already an expert or good… in the field”. He states, “It’s, it’s, it’s augmenting, it’s, it’s working together with a human”. Elements.cloud, for instance, can support Salesforce architects, consultants, and business analysts by automating documentation, drawing process maps (even from org metadata, sketches or text), and generating use cases for agents, significantly accelerating implementation cycles and ensuring governance.
4. Productivity Agents (e.g., Notion AI, Zapier AI, Gong.io, ElevenLabs): To enhance daily operations across the board, organizations can embed productivity-focused AI within their existing collaboration and workflow tools. Whether it’s Notion AI assisting with document creation, Gong.io streamlining meeting analysis, or Zapier AI automating data classification across various apps, these agents can free up valuable employee time from mundane tasks. 
5. Pre-built Agents (e.g., Intercom’s Fin, HR Helpdesk Bots, Drift Chatbots, Shopify AI): For common, standardized functions like initial customer support inquiries or basic HR policy questions, pre-built agents can be deployed. These offer immediate value and can handle a high volume of routine interactions, allowing human teams to focus on more complex and empathetic tasks. Intercom’s Fin, for example, successfully resolves 86% of Intercom’s support queries on its own for Element.cloud.

Autonomous agents and multi-agent orchestration represent a significant milestone. Organizations are investing in the capability to design and orchestrate autonomous agents that can reason, plan, and execute multi-step tasks across integrated platforms. This involves agents collaborating, seamlessly handing off tasks, and proactively solving problems, transforming entire business processes with minimal human intervention.

The Path Forward: Governance, Risk, and Education

This multi-platform approach isn’t a free-for-all. It demands a robust AI governance framework. As Ronald Starreveld states, “Things are moving so quickly” in the regulatory landscape, requiring continuous review. But his non-negotiables for CIOs are clear:

• Vision: Always start with a vision. What is it that you’re trying to do? What is it that you want to do?
• Governance & Risk Management: AI introduces new risks, particularly around data privacy (GDPR still applies), bias, toxicity, and intellectual property exposure. With AI Agents, you need to consider the level of risk you are prepared to take for any use case.
• Process-Led Design: Every agent implementation, regardless of platform, should start with a clear, well-defined business process. Organizations must streamline and optimize human-driven processes before automating them with AI to avoid “your mess for less”. Elements.cloud has launched free Agent Designer capabilities, which supports the designing, building, and governance of reliable agents. 
• Education & Mindset Shift: Organizations must proactively educate everyone, from the C-suite to the front lines, on what AI means, how it augments human capabilities, and where its value truly lies. This involves transforming mindsets to understand that AI is about more than just chatbots or technical implementations.

 Ronald Starreveld’s advice for CIOs feeling too busy for AI strategy: “you’d have to make time. This is gonna happen”.

Final word

A multi-platform AI strategy needs to start with clear business objectives rather than just technology trends, to avoid “pet projects” that lack real value. The prioritization of AI agents, whether “Champion” or “Workhorse,” depends on precise real-time business process mapping that is automatically generated. Emerging standards like MCP and A2A are crucial for integrating diverse AI agent platforms, reducing vendor lock-in. The AI strategy necessitates robust governance, encompassing vision, risk management, process-led design, and education, to ensure successful and impactful AI adoption.

Vala Afshar, Chief Digital Evangelist at Salesforce, emphasizes the crucial role of a unified AI agent strategy in achieving an autonomous enterprise. I’ll leave the final word to him.

“The superpowers of AI and autonomous business will elude you unless you can map your process in real time. Identify where you need to have AI agents implemented, and then try to do that, not just for a particular task, but tasks that extend beyond your line of business.”