AI workflows - Boomi: Fixing foundations with smarter data management
This is a guest post for the Computer Weekly Developer Network written by Ann Maya, EMEA CTO, Boomi.
Maya writes in full as follows…
Businesses across all industries are racing to tap into the potential of generative AI and agentic AI workflows to unlock productivity and efficiency gains. Yet many are hamstrung by a raft of technological challenges.
Poor-quality data, often stuck in silos or lacking a consistent structure, makes it difficult to safely provide meaningful information that AI solutions need to produce reliable results. This is compounded by outdated processes and legacy systems that simply can’t keep up with the speed and scale of today’s digital demands.
To succeed in their AI innovation plans, IT leaders must reevaluate what it takes to support the increasingly decentralised, digital processes proliferating across their organisation. To drive digital cohesion, it’s essential to have centralised visibility, control and governance before it’s too late.
Data, done right
Data is the catalyst for digital transformation. Outdated systems, accumulated technical debt and manual workflows make keeping up with the pace of technological change impossible without trying a new approach. Even if an organisation has good-quality data, it isn’t enough.
Expanding data with increased volumes and providing access to it through uncontrolled APIs can create weak points and elevate the risk of security breaches. Systems that are burdened by data sprawl and outdated API management practices are prone to failure under pressure, damaging trust among customers and partners.
Managing higher volumes of data securely and accurately means streamlining the processes involved in collecting, transforming and managing access – especially when building AI workflows. Poor or incomplete data compounds the chance of mistakes, hallucinations and poor decision-making. Ultimately, relying on rigid data systems risks mounting technical debt as AI initiatives scale.
Modern data infrastructure
An AI framework is only as effective as the data it can access. When models are drawing upon fragmented datasets, they operate within a limited scope, without the full picture. While the outputs may be technically correct and in line with the data that has been used, they can be misleading, as the wider context needed to make accurate decisions could be missing.
For instance, customer names might have been spelt differently in separate records, and while both systems are believed to be right, inconsistencies and duplications begin to breed. To overcome these challenges, organisations need to create a single source of truth for their data, which is synchronised across all teams and systems. It’s imperative to create a trusted data foundation to feed AI-driven workflows, based on consistent, accurate, and up-to-date insights.
Once that foundation is in place, organisations can unlock real-time insights to improve performance and spot inefficiencies. They’ll also be equipped to build and deploy AI agents faster using low-code solutions powered by their clean sets of data. And by continuously analysing and optimising automated processes through a unified view, they can easily scale AI-driven workflows with consistency and control while tracking cost implications.
As these systems mature, the priority must remain on protecting data and upholding ethical practices.
To safeguard AI agents that they build, organisations need to consider the agent’s “eyes and ears” – how it perceives what is asked of it and maintains awareness of its environment. As well as the “arms” – how it instigates actions. A trusted data foundation provides the business-rich contextual information to augment AI models. Files, automated processes, data queries and commonly, APIs, provide the tools to activate actions across AI workflows.
Beware, zombie APIs!
Since agents will function with some degree of autonomy, guardrails need to be applied to avoid misuse.
The good news is that APIs are already widely used across organisations, which means that future agents already potentially have access to ready-to -use tools. The bad news is that not all APIs are well formed and there are a lot of zombie and shadow APIs roaming around completely undetected by IT departments. Bad actors are aware and APIs have become the primary entry point for software attacks, with untracked or shadow APIs representing some of the weakest links.
Boomi’s Maya: When automation is orchestrated through cloud-native workflows, updates become more efficient & oversight becomes more centralised.
Security should be integrated throughout the API lifecycle, not just applied after deployment. An agile API management framework provides central oversight across all integrations, enabling federated control and governance that protects innovation without introducing operational bottlenecks.
Laying the groundwork
Most transformation efforts were launched before the accelerated innovation cycles and AI-driven demands of today. To stay competitive, organisations must shift from outdated processes to systems built for adaptability, speed and scale.
A unified approach, where a system is capable of seamlessly managing and connecting data across a range of devices, apps and environments is needed to tackle today’s demands. To reduce the complexity and effort in building these connections, using low-code/no-code AI-driven integration strategies will be pivotal. When automation is orchestrated through cloud-native workflows, updates become more efficient and oversight becomes more centralised. As agent-based systems become increasingly integrated across businesses, strong, dynamic data governance and central control will become critical safeguards.
With platform-centric approaches to these capabilities, what once felt unmanageable is now within reach. By establishing a resilient, connected data foundation, businesses can turn complexity into clarity and data into an active driver of innovation. The organisations that invest in this ecosystem today will be the ones that reap the full advantages of the AI era, rather than be left behind.