Image generated by AI: © Anna/Stock.adobe.com
BearingPoint Ireland and Kreoh discuss the major considerations that organisations need to keep in mind when integrating AI-driven business solutions.
Last month, tech management and consultancy firm BearingPoint announced a new strategic collaboration with Kreoh, a Dublin-based AI agency and development lab. The partnership will enable the two companies to provide their clients with generative AI capabilities as well as access to the latest research in the area and guidance on how to maximise its impact.
With plans to provide improved AI services to clients in a variety of sectors, we spoke to representatives from both BearingPoint and Kreoh about the benefits and challenges of AI-driven business solutions.
AI in business processes
When it comes to AI adoption in the business sector, reports over the last year indicated that businesses in Ireland are increasing their focus on AI, with increased budget considerations to integrate the technology as well as its overall influence on long-term strategies.
In regard to the ways businesses are utilising the innovative tech, Kreoh’s Garry Tiscovschi highlighted how the company’s report generator has seen a lot of use by R&D tax credit specialists.
“R&D reports win billions in savings for Irish and UK companies and promote much-needed R&D in our ecosystem, but require the diligent, detail-oriented work of experts,” he says. “[Generative AI] systems really excel when it comes to detail and following the well-defined writing structures you would expect when reporting to the state.”
BearingPoint Ireland’s Stephen Redmond described how his company helped global NGO Plan International implement an AI system to help compile reports for stakeholders, which normally consists of a “mammoth task” of collating text reports from all of its worldwide projects.
“The volume of data made it a challenge to structure into the briefs that were necessary to create,” he says. “We implemented a gen AI system, including a human-in-the-loop validation, to free up time for their staff from reporting and focus on their humanitarian work.”
Both Tiscovschi and Redmond have a background in data, which led to a transition into AI. Tiscovschi, who is managing director and co-founder of Kreoh, taught analytics labs to biomedical students at Trinity College Dublin, before working in release management at Mastercard. While in the latter role – which preceded the release of Chat-GPT – he says he would work away on early generative AI freelance projects with his future Kreoh co-founders.
Meanwhile, Redmond used to work in business intelligence and reporting, particularly with Qlik products, on which he published three books. After obtaining a master’s degree in data analytics, he spent more than six years working with AI at Accenture, before taking on his current role as director and head of data analytics and AI at BearingPoint Ireland.
Careful considerations
Redmond believes that one of the key benefits that generative AI offers to businesses is the ability to “better mine your own IP”.
“You have a ton of data throughout your business that can be usefully mined and sometimes rediscovered,” he explains. “Being better able to bring your own IP to solve business problems is a key competitive advantage.”
However, there are some things that businesses also need to be mindful of when using this technology, as pointed out by Tiscovschi. “Malicious actors can map out unprotected systems and manipulate the AI agent within to ‘mine’ data, expose sensitive prompts, spawn inappropriate responses on an organisation’s behalf or cause havoc with the databases/tools that a company may have connected to their AI agents to make them more useful in the first place.”
Along with spotlighting other “sources of mischief” like prompt injections and jailbreaking, Tiscovschi described specific examples of AI pitfalls, such as a blunder where Air Canada’s AI assistant misrepresented the company’s policies (which resulted in a lawsuit), as well as an early incident experienced by Kreoh.
“When starting development of a more advanced chatbot for a new customer, during prototyping the AI agent looked to offer the products of our client’s competitors. Not optimal.”
Redmond adds that in instances where companies release AI agents that are not fully tested and result in bad responses, they can experience reputational damage. He stresses that companies can avoid this by following best practices and ensuring that the AI is effectively tested.
“Your own employees can be a great ‘Red Team’ to try and break things, so releasing internally first is a good strategy. But always make sure that people know that they are dealing with an AI and not a human.”
With various risks and concerns following the implementation of AI business solutions, it’s important that businesses keep some key considerations in mind when diving into this area.
Redmond emphasises that businesses need to think about the source data that they’re feeding into their AI systems, as well as what they ultimately want out of it.
“You can use Microsoft 365 Copilot to search across all of your documents in SharePoint, then that could be useful for discovery, but you could also be worried that people have ‘overshared’ documents in the past and then take a more curation-based approach and only index specific SharePoint stores,” he says. “The same if you have a helpful chatbot that answers questions on a particular topic – you need to make sure that it only has access to the most up-to-date sources and that old versions are removed.”
Moving forward
With the integration of generative AI in business culture still in its early stages, what does the future have in store?
Redmond believes that the tech will quickly become embedded in normal business practice. “We won’t even think about asking gen AI to draft emails or documents or to generate images for our presentations.”
He’s also looking forward to seeing how AI-driven video technology plays out, particularly OpenAI’s Sora. “I know that a lot of people in content generation are nervous about these tools replacing them, but I don’t think we hire an artist for their ability to draw, we hire them for their ability to draw what is in their imagination, and that is where their genius lies,” he says.
“I am not sure that artists will ever stop creating wonderful works, and these technologies will just enhance that.”
Tiscovschi agrees with Redmond’s outlook, stating that “this is just the beginning”.
“We will continuously see more teams of humans and their AI agents or tools working together to achieve tasks,” he says. “A human quickly mining their organisation’s IP, automating repetitive tasks and then collaborating with their AI copilot on a report or piece of code will have a constantly growing multiplier on their productivity.”
However, he adds that humans will always need to “program or provide instruction for the technology that serves us”.
“We will still need to coordinate our organisations and then produce and interpret instructions that we’ve chosen to agree upon,” he says. “And if we’re not producing media as much via digital pens and recording studios, we will still need to design, curate, envision and deliver art, culture and knowledge for those around us.
“Now, more of the repetitive parts will be automated and the friction reduced. This help is fortunate as there’s still a near infinite amount of work left for us to do!”
Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.
