How businesses are accelerating time to agentic AI value

A recent survey involving 1,050 Chief Information Officers (CIOs) has shed light on the rapid adoption of AI agents within the IT sector. A staggering 93% of these IT leaders are gearing up to integrate AI agents into their systems within the next two years. The focus is on breaking down data silos to facilitate this integration, aiming for a more streamlined and efficient IT environment.

The survey also highlighted the extensive use of applications among these organizations, with an average of 897 apps in use. A significant 45% of respondents reported using over 1,000 applications, which poses a challenge for IT teams striving to create a cohesive user experience.

Currently, only 29% of enterprise applications are integrated and share data across the business. In preparation for the increased use of AI, CIOs are dedicating 20% of their budgets to data infrastructure and management, which is four times the amount they're spending on AI itself, at just 5%.

Understanding AI Agents

AI agents, as defined by ARK Invest, are set to revolutionize the way we interact with digital applications, marking a significant shift in human-computer interaction. These agents can:

• Understand user intent through natural language processing
• Plan actions using reasoning and contextual awareness
• Execute tasks using tools to fulfill user requests
• Enhance their performance through iterative learning and continuous improvement

ARK Invest predicts that AI will greatly enhance productivity in knowledge work. By 2030, they anticipate a substantial increase in software use per knowledge worker as companies invest in productivity-enhancing solutions. Depending on adoption rates, global software spending could grow from a 14% annual rate over the last decade to between 18% and 48% annually.

Accelerating Time to Value with Agentic AI

Valoir, a technology research firm, suggests that agentic AI can deliver exponential benefits by automating complex tasks and interactions without human intervention. However, building agentic AI that can manage these complex tasks efficiently presents its own set of challenges.

Using a platform specifically designed for agentic AI development, such as Salesforce Agentforce, can significantly speed up the process. Valoir's research indicates that organizations can deploy autonomous AI agents 16 times faster and achieve a 75% increase in accuracy compared to other methods.

Valoir's Seven Phases of Agentic Development

Valoir has outlined seven key phases for developing agentic AI, tailored to the varying complexity of tasks and data sources:

1. Model Setup: The foundation for building AI agents.
2. Data and Application Integration: Ensuring seamless data flow across systems.
3. Prompt Engineering: Crafting effective prompts for AI interaction.
4. AI Guardrails and Security: Implementing safeguards and security measures.
5. User Interface and Workflow/Application Development: Designing user-friendly interfaces.
6. Tuning: Fine-tuning AI models for optimal performance.
7. Data Accuracy: Ensuring the reliability of data used by AI agents.

Valoir's research highlighted significant differences between a Do-It-Yourself (DIY) approach and using a deeply integrated platform like Agentforce. For instance, DIY model setups typically take three to twelve months, whereas Agentforce's pre-integrated models can be deployed 7.5 times faster.

Similarly, integrating data and applications with open-source alternatives can take over a month, while Agentforce can accomplish this in weeks, three and a half times faster. The most notable difference was in establishing AI guardrails, trust, and security, where a DIY approach could take over a year, while a platform like Agentforce provides these layers much more rapidly.

Accuracy is crucial for the effective deployment of AI agents. For simple tasks, DIY approaches achieved 50% accuracy compared to 95% with Agentforce. In complex scenarios like sales coaching, DIY accuracy dropped to 40% while Agentforce maintained 95%. Overall, using a platform can boost agent accuracy by 75%.

Valoir's findings concluded that the total time for DIY projects averaged 75.5 months, while projects using Agentforce reached productive accuracy in just 4.8 months, making the platform approach 16 times faster. For more insights into Valoir's agent AI research, you can visit their detailed report.