“GenAI integration doesn’t need to mean IT overhaul,” says AI entrepreneur Berken Sesen

60% of IT leaders say GenAI can’t integrate with their current tech stack, according to a recent Salesforce report. Yet, as Deloitte research shows, 67% of organisations are increasing investments in GenAI.  

These conflicting statistics raise an important question: How can GenAI projects succeed when you have underlying IT infrastructure challenges? Companies must invest in legacy application modernisation to get value from anticipated AI integrations.   

The additional challenge is that — despite AI projects being a huge priority for most businesses right now — overhauling IT to accommodate/accelerate these initiatives spells disruption and downtime. And yet, stakeholders don’t want to pause day-to-day money-making operations for any time at all.  

Seeking a solution, we spoke to Berkan Sesen, AI entrepreneur and founder of Tinie.ai, a startup currently working on the delivery of several healthcare AI projects. He spoke to us about the perception of IT as a value-adding asset, eliminating data siloes, breaking down giant leaps into small steps, and winning stakeholder support. 

Cost centres vs revenue drivers

Many senior leaders still don’t recognise the tremendous value their IT professionals bring to the business. “The biggest problem for IT leads is that they’re seen as a supportive function that doesn’t produce revenue,” says Berkan. 

This creates a two-tier hierarchy between business functions that can afford downtime and those that can’t. “In that context, the daily operations of revenue-driving departments and functions cannot be disrupted. This makes it really difficult for IT to update or overhaul any legacy infrastructure.”  

If you can demonstrate the value of IT — for example, how historical infrastructure upgrades have helped your business succeed — it helps build your case to make the necessary upgrades. 

Data siloes vs data discovery  

Gaining approval for potential downtime isn’t the only challenge. Berkan highlights that the data you need for GenAI isn’t always in the right place, saying, “With old-school legacy platforms, you’ll also have data siloes.” Which, according to Berkan, is an industry-wide issue. “Every organisation I’ve ever worked in has had ongoing projects to try and unify siloed data, but it’s not easy.” 

Berkan explains that most of these issues “arise from the fact that big businesses have multiple business units using different data storage methods and data providers who use different solutions.”  

The issue of data siloes isn’t a new problem, but “it’s relevant again now we’re talking about GenAI.” Some associative AI models can help unpick the challenge of siloed data, but for GenAI success, “it’s important to think about using unifying data services and looking at where your data is stored, because the two go hand- in- hand.” 

Upgrade to AI one module at a time 

Your AI overhaul needn’t be overwhelming. Preparing for AI “doesn’t mean IT leads need to scramble to upgrade, update, and overhaul their entire tech landscape in one fell swoop.” Berkan recommends breaking it down, saying, “It can be done in manageable chunks … Modularisation is where you modernise and migrate chunks of your tech stack at a time and therefore minimise the probability of disruption.”  

For example, “Migration to the Cloud is a potential solution which means companies don’t need to update or overhaul all their infrastructure.” While easier, these solutions don’t make this complex task easy.

“This itself comes with its own challenges, and it’s not always straightforward.” AI may promise advanced computing for all, but AI integration still requires specialist skills and experience, and a great deal of time. 

Bring stakeholders with you 

While expert teams can see the IT infrastructure challenges ahead, they need to articulate them in ways other roles can understand. “Technology folks have stakeholders in each business who need to be on board,” says Belkan. Getting them aligned is hard but necessary.

“For successful migration, they need to make sure that stakeholders are well informed, provide feedback and help design the infrastructure with the IT person.” This collaboration stage ensures “the new platform overlaps with the expectations of the business.” 

If possible, pointing to bottom-line benefits helps to bring senior leaders along for the ride. “Always try to put a number to your project delivery, because at the end of the day, everything that’s a bit intangible and qualitative always ends up being a bit fluffy,” which matters because otherwise “it’s difficult to get backing and kudos for the work you’ve done.” 

Don’t wait for perfect: Start! 

GenAI won’t work without some IT fundamentals in place: A source of reliable data, stakeholder buy-in, some level of in-house AI skills. But that doesn’t mean you need to wait for everything to be perfect before you start implementing AI. You can’t clear out all legacy systems at once, so taking a modular approach helps you test use cases, avoid downtime, build a case for scaling up AI based on real results, and get moving quickly.