We’re losing the battle against complexity. So, can artificial intelligence (AI) rise to the rescue? A new survey of 800 IT decision-makers by Camunda found IT teams deal with an average of 50 endpoints (applications, APIs, robotic process automation) in their efforts to satisfy business processes. AI could help reduce complexity, but 84% say a lack of transparency in AI creates new headaches.
“Most organizations say that as their business becomes more complex, digital, interconnected, and automated, there’s an increased risk of core processes failing,” the survey authors pointed out. “It’s increasingly difficult to effectively analyze and optimize them.”
Also: Why the ‘Bring Your Own AI’ trend could mean big trouble for business leaders
AI is being hyped as “a means to manage and automate processes from end to end,” they continued. “With many organizations increasing their investment in AI, this will need to be orchestrated like any other endpoint within automated business processes.” At the same time, 85% of respondents said they face challenges when scaling and operationalizing AI across their organizations.
<!–>
This is where things get messy
Thinkers and doers across the industry agree with this prognosis of AI as a double-edged sword. “This is the crux of what I call the productivity paradox – AI’s benefits can’t be fully realized until organizations gain control over their existing tech stack,” Faisal Masud, president of worldwide digital services at HP, told ZDNET.
“The surge in apps, APIs, and endpoints has created a complex environment for IT teams, developers, and employees,” he continued. “Although AI can simplify processes, automate routine tasks like system updates, and eliminate the need for manual IT requests, it can also introduce new complexities if not managed properly, potentially leading to employee disruption and burnout.”
AI can reduce complexity in many areas, “such as automating repetitive tasks similar to what RPA did, or provide more reasoning for simple tasks, use agents to work on scheduling tasks, create code, provide API mapping, predict system failures, and provide insights into system optimization – which are all great,” said Andy Thurai, principal analyst with Constellation Research. “However, it can also add complexity that normal engineers can’t handle.”
For starters, AI “can require specialized infrastructure that only very advanced skilled AI/ML engineers can solve,” Thurai continued. “It can also introduce more dependencies both on the data side or on the model side which might be very difficult to manage.”
Also: Managing AI agents as employees is the challenge of 2025, says Goldman Sachs CIO
Adding AI to applications “can really complicate things and make long-term maintenance more challenging,” agreed Nathan Brunner, CEO of technology specialist Boterview.
“Whether you choose to host the AI model on your own servers or use an external API, it introduces another layer of complexity that requires monitoring, maintenance, and troubleshooting, which can lead to higher operational costs. In the end, customers care more about the value the product provides than the technical intricacies behind it. Therefore, it’s crucial to ensure that any AI integration delivers genuine benefits to the user.”
Cybersecurity is another Pandora’s box that AI opens: “Challenges that existing software and hardware can’t solve,” Thurai warned. Such challenges include “increases in the attack surface so the hackers can easily attack models either to explore the data or poison the decisions thereby leading to disastrous results later on.”
Also: AI transformation is a double-edged sword. Here’s how to avoid the risks
However, other experts said AI can be a major force in reducing complexity. “There’s a lot of work around documenting APIs, endpoints, and code that developers do not enjoy doing and that AI is very good at,” said Komninos Chatzipapas, founder at HeraHaven.AI.
“Especially now that we’re seeing it become more reliable with scaling test-time compute, it can be used as a tool to reduce complexity by increasing clarity.”
Making sense of the madness
An analogy for the complexity of AI can be found in the biological world, said Paul McDonagh-Smith, senior lecturer of IT at MIT Sloan Executive Education, who said he sees ongoing beneficial “mutations” altering the composition of these systems.
“In biological evolution, natural selection preserves beneficial mutations, gradually building complexity into the organism. Similarly, in AI, simple rules are applied to extract information from data, which is then fed into the model. As this process is repeated, the complexity of the data is transferred into the model itself.”
This pattern of creating complexity from simplicity “has important implications for our organizations,” McDonagh-Smith continued. “If we think of our organizations as organisms, we can identify opportunities to better adapt to AI, which is continuously evolving the business environments and systems we all operate within today.”
Also: Your AI transformation depends on these 5 business tactics
He added: “By defining simple organizational rules – or algorithms – that define our organizational natural selection, we can gain a deeper understanding of, and more effectively respond to, the mutation process we’re all experiencing as AI technologies and techniques transform our environments.
“By viewing our companies through this lens and leveraging the pattern seen in both biological and AI evolution – where simple rules generate complexity – we can uncover ways to refine and manage that complexity, ultimately driving better outcomes.”
To address the challenges and make AI a productive force in reducing complexity, Thurai urged greater adoption of “centralized governance frameworks, having robust observability and monitoring tools, and continuously training the unskilled workforce to be AI-ready.”
Masud said a successful AI deployment may require actively simplifying and integrating existing systems and applications: “This requires a targeted approach to technology adoption, ensuring that AI enhances rather than complicates employees’ work. Additionally, leaders must prioritize a successful rollout to ensure employees are well-versed in using the new tools – and ultimately find them useful.”
Also: How your business can best exploit AI: Tell your board these 4 things
For instance “offering interactive demos or white-glove training can help address any questions and ensure smooth adoption,” he added.
“The bottom line is AI is more than just technology,” said Thurai. “It is new thinking, new ways of working, new ways of strategy, new brain power, and new processes which means you need to design systems that can adapt to those challenges. The systems have always been designed until now for deterministic decision making and now we move into the era of probabilistic decision making. Most things have to change and move away from legacy thinking.”