The excitement surrounding AI agents has been growing steadily over the past year. Despite all the buzz, many organisations struggle to realise the full benefits. A recent study from Boston Consulting Group found that while 75% of executives rank AI as a top-three business priority, only about a quarter of companies see ROI from their AI investments. So, what’s going wrong?
The problem with AI agents today
AI agents have become popular tools, but they often disappoint when it comes to enterprise deployment. Many AI systems currently function as passive assistants – chatbots or digital assistants that only respond when prompted. But when AI acts as a bystander, it can’t drive real change or add value.
The problem isn’t AI itself. It’s how we use it. Most companies treat AI as an isolated tool outside of their core business processes. Or they fail to provide enterprise AI for employees, leaving individuals to use whatever AI tools they prefer, which compromises data security and privacy. This fragmented approach fails to unlock the full value AI can provide.
Why AI needs to be part of business processes
For AI to reach its potential, it should be part of a business process. This is where work gets done – whether it’s making decisions, serving customers, or scaling operations. We worked with US behavioural healthcare provider Acclaim Autism to accelerate new customer onboarding time by 80%. Previously, signing up new patients took up to six months, involving manually processing medical documents and lengthy compliance checks. By automating the extraction of patient data and the key onboarding steps, we reduced waiting times to only 10 days.
It’s within such workflows that AI can add real value. When AI is part of a structured business process, it gains clear goals, purpose, and accountability. This enables it to assist with tasks, provide insights, and automate repetitive work, freeing employees to oversee more complex tasks.
Making AI more accessible and effective
Second, AI needs structure to be effective. Embedding it in a process ensures AI works within a defined flow of tasks, making it easier to complement human efforts and automation tools. For example, academic advisors of the University of South Florida (USF) use conversational Gen AI assistants for faster, more personalised support. This saves 15 minutes in prep and wrap-up for each 30-minute meeting, resulting in up to 12,500 working hours saved annually.
Another key benefit is access to data and data fabric. When integrated into processes, AI can access necessary real-time data across systems, providing more business value. Data fabric unifies data across systems without requiring migration, enabling fast, secure, low-code access to enterprise data wherever it resides. It simplifies connecting and modeling data while ensuring high performance and record-level security.
Finally, processes provide the infrastructure necessary to scale AI across an organisation. This simplifies secure, effective AI deployment, with the flexibility to grow as needed.
Improved transparency and security
AI is powerful but must be carefully managed. By embedding AI into a workflow, businesses can implement safety measures like human approval steps for high-risk actions. This ensures that AI doesn’t cause harm or errors, while activity logs enable easy auditing and compliance. Data privacy and governance also become simpler to manage, ensuring AI uses the right data while maintaining security.
In addition to safety, processes make AI measurable. Many businesses struggle to track how AI is performing, particularly when it’s used in isolation. However, when AI is part of a process, its actions are tracked, providing businesses with valuable insights into its performance.
The complexity and challenges of integration
This all sounds great in theory, but many businesses struggle to integrate AI agents into their existing software stacks and workflows. This is especially challenging when dealing with legacy systems and siloed data. If it’s challenging for a human to access data from multiple systems and data repositories, an automated agent isn’t going to have more success. Other difficulties frequently include poor data compatibility and quality, gaps in monitoring and observability, versioning and compatibility drift, as well as cost and resource constraints and compliance issues.
Careful planning is imperative. Integrating agentic technology often means reimagining and rebuilding software stacks and processes. It’s a significant undertaking, which may need to involve a top-to-tail rethink of how the business stores and processes data.
How businesses can maximise AI’s value
To get the most out of AI, businesses must have a strong data foundation. This means ensuring that AI has access to high-quality, real-time data across systems. Companies should also focus on explainability, only automating processes that can be easily explained and audited.
AI agents can’t work in isolation. They need to be part of a broader, interconnected system that includes human employees, other automation tools, and various data systems. To achieve that, organisations need to orchestrate AI within a wider business process. This orchestration is crucial for making sure AI agents are part of an efficient, impactful process.
Building a smart AI strategy
AI can’t reach its full potential without a clear, strategic approach. It should align AI adoption with business priorities, helping companies pinpoint where AI will have the most impact. This ensures that AI investments are well-directed and lead to tangible improvements.
It’s also crucial to be adaptable. Processes change over time, and AI solutions need to evolve along with them. For many organisations, this means undergoing a cultural shift, with a focus on upskilling employees and fostering a mindset of continuous learning and innovation.
AI agents aren’t a magical solution for every business challenge. But with the right strategy, businesses can get them to live up to the hype. The key is embedding AI into business processes. When done right, AI can offer real value, making processes more efficient, scalable, and intelligent. So, rather than deploying isolated AI projects, businesses should integrate AI into their core operations and start seeing the results they’ve been promised.
Medhat Galal
Medhat Galal is SVP Engineering at Appian where he steers the engineering endeavours, with a focus on business process design and harnessing the power of technology to optimise operations.
