As businesses pour billions into artificial intelligence, one company's 50% success rate reveals what separates genuine transformation from expensive experimentation
Nine out of ten AI projects never make it to production. Born with fanfare and nurtured through proof-of-concept stages, they simply fall by the wayside in what Nathan Marlor, Head of AI at technology firm Version 1, terms "random acts of AI": lots of activity delivering minimal impact.
It's a sobering statistic, confirmed by McKinsey's recent report showing fewer than 10% of generative AI use cases progress past pilot stage. Yet Version 1, a Dublin-founded company transforming businesses since 1996, is hitting closer to 50%. Their success rate points to a profoundly different approach to weaving artificial intelligence into the fabric of business.
READ MORE: Archibald announces 40 high quality training places on two Assured Skills Academies with Version 1
The shift in client expectations has been stark. "If you rewind two or three years, we were talking to a lot of clients who wanted to really just explore the art of the possible," Marlor explains. "But now, when we go and speak to clients, they're asking: how do we get this into production? Who's going to maintain it? How do we measure ROI?"
This transformation mirrors a broader industry maturation. Where businesses were once content with pilots and proof-of-concepts, they now demand tangible results and Version 1has had to evolve accordingly.
What sets the company apart is Version 1’s very ownBusiness Challenge Analysis Framework, a methodology that deliberately inverts the traditional technology-first approach. "Only when we've got a real deep understanding of the customer and what they're trying to achieve, do we look at where AI can help," says Marlor. "That starts with operational questions first, then commercial considerations including ROI, and only then do we look at technology."
It's a philosophy reflected in Version 1's recent high-profile contracts, including a €102.7 million deal to transform Ireland's school employee payroll systems and a partnership with the International Schools Partnership to develop AI-driven teaching tools.
As Marlor points out, if you walk into most organisations today and you'll find AI experimentation everywhere. Marketing teams are testing ChatGPT for content creation, HR departments are trialling Copilot for recruitment emails, business development teams are exploring automated transcription tools.
But this "scattered experimentation" rarely delivers transformation. "Countless times we'll speak to organisations who have just started to use Copilot, maybe ChatGPT, Claude or the Otter notetaker," Marlor says. "But customers that are successful are avoiding scattered experimentation. They're striving for systematic transformation."
When departments operate in silos, experimenting independently without coordination or strategic oversight, organisations end up with random acts of AI. The result is impressive demos in departmental meetings and enthusiastic early adopters, but minimal impact on the bottom line or operational efficiency.
Companies that break through this barrier share common characteristics. They resist the temptation to let AI initiatives flourish organically, instead implementing what Marlor describes as "a strategic, structured, centralised approach." More critically, they're willing to adapt existing processes rather than simply overlaying AI onto current workflows.
"Organisations that are successful typically are open to adapting and changing internal processes or workflows, rather than just sticking AI over the top and hoping for the best," he explains.
The distinction matters because scattered experimentation ties up resources and creates organisational fatigue that makes future, more strategic AI initiatives harder to sell internally.
Marlor is particularly excited about agentic AI, intelligent systems that perform specialist tasks rather than generalist content creation. He describes it as "dynamic RPA," referencing the robotic process automation tools that have been quietly revolutionising back-office operations for years.
"RPA has been around for years, automating workflows and processes, but it tends to be quite brittle and quite fixed in how it can operate," he says. "But when we overlay the fluidity that generative AI brings into that, we suddenly get a much more compelling proposition."
His favourite example involves a US insurance broker where Version 1 developed an agent that automated the tedious process of checking renewal policies and cross-referencing policy numbers. What once took a team of ten people hours of manual work now happens in minutes, with the AI system not only checking policy numbers but understanding the semantic meaning of terms and conditions.
It's the kind of transformation that doesn't make headlines but genuinely changes operations by freeing human workers from repetitive tasks to focus on work requiring human judgement.
This pragmatic approach extends to Version 1's 'AI for good' initiatives."Working with the Children and Family Court Advisory Support Services, the company has automated the generation of personalised communications to children and families working through complex legal situationsreducing the time caseworkers spend on administrative tasks from 20-25 minutes per letter to seconds. For Dyslexia Ireland, they've built a browser plugin that simplifies complex text, and with Encephalitis International they've developed tools to make medical information accessible across 75 languages.
These projects matter to Marlor on a personal level given his wife is a teacher, and he can see firsthand how AI tools can genuinely support educators.
If there's one conversation Marlor has more often than others, it's about data quality. "At the end of 2022, there was this rhetoric across the market that we must collect data, that one day it will be valuable," he recalls. "This was the start of the data lake and organisations just pooling their data in the understanding that one day it could have some real value to it. But only a very few, from my experience, did that with any kind of structure around governance or data quality."
Rather than attempting wholesale data transformation projects that can take years and cost fortunes, Version 1 takes a surgical approach. When working with a major Tier One bank on a data remediation project involving 15 years of disparate data across multiple systems, they didn't try to fix everything. Instead, they built a focused data model designed specifically for the task at hand, processing around 300,000 policies in a fraction of the time a complete overhaul would have required.
Marlor is passionate about AI governance, viewing regulation not as a constraint but as an accelerator. "Organisations with robust governance end up deploying AI faster, and that's because they've got the trust and the infrastructure to move more confidently and you can move at speed," he says.
Version 1 has partnered with Credo AI to develop governance services helping organisations navigate the increasingly complex regulatory landscape, from the EU AI Act to emerging US legislation. Working with another Tier One bank, they helped establish a responsible AI board that has now signed off on over 100 AI use cases.
"Cybersecurity is treated as first-class citizen in every organisation," Marlor observes. "Over the next year or two, we'll see governance become the same."
As Version 1 prepares to invest £40 million in UK expansion, generating over 100 new jobs across their technology hubs, Marlor is thinking about the next wave of developments. In the future Marlor expects to see small language models fine-tuned for specific industries becomingmuch more prevalent. Multi-agent systems will enable AI applications to communicate across organisational boundaries and governance will evolve from a compliance afterthought to a core business capability.
Sync NI's Autumn 2025 magazine explores how Northern Ireland is moving AI from pilot to production
This issue features exclusive insights from industry leaders on real-world AI implementation, responsible adoption, enterprise infrastructure, and how data-driven innovation is transforming businesses across the region.
Read the Big Data Special Edition free online →
Stay connected with NI's tech community:
Subscribe to our newsletter – Get the latest tech news, job opportunities, and events delivered to your inbox
Visit Sync NI – Your daily source for Northern Ireland technology news
