Karli Kalpala, Head of Strategy & AI Agent Business at Helsinki-based Digital Workforce
How has public perception of AI shifted in 2025 compared to the initial ChatGPT hype?
The conversation around AI in 2025 has shifted from hype to an acceptance that the technology is no longer just a productivity tool. Large institutions, such as major banks, have introduced generative AI-based agents and solutions to their end customers, proving that we’ve passed the tipping point. Looking ahead, businesses are increasingly prioritising employee upskilling and the integration of AI tools and agents into day-to-day operations.
Previously, there was some hesitation about allowing end customers to engage in free-flow chat with generative AI solutions due to concerns around prompt injections and associated risks. Today, customer calls and chats can be driven by AI, with humans remaining in the loop as a safety measure, and organisations are embracing this more autonomous approach.
Upskilling employees to use and manage AI responsibly will become a key priority in 2026. Companies will need to ensure their teams are AI-literate across departments and workflows, able to collaborate effectively with autonomous agents, and able to oversee the platforms they adopt.
Which industries have moved fastest on AI adoption, and which are lagging behind?
Large institutions and organisations in financial services and healthcare have been the most visible movers this year, particularly in deploying customer-facing AI solutions. The key shift has been moving from internal process automation to customer-facing applications, which signals a commitment to transformation.
This momentum will accelerate in 2026 as AI becomes further embedded across customer interactions, compliance and decision-making processes. However, business and tech leaders should recognise that speed of adoption isn’t solely about technology deployment. The fastest-growing industries are those whose leaders understand that progress requires change management, process redesign, and people transformation. Those lagging behind are often stuck in the pilot phase, treating AI as a productivity tool instead of recognising that most of the work involves people and processes, not just the technology itself.
What’s driving the current conversation around AI agents in business?
The rise of reasoning models is likely the defining AI development of 2025. These models have brought agent-like capabilities into everyday use, enabling features such as document analysis, automated scheduling, data querying, and other autonomous task handling. More importantly, they support the creation of entirely new information networks, allowing AI agents to connect tools, data sources, and workflows in previously impossible ways.
Until now, information has always flowed between people, or between people and documents. Reasoning models change that by being able to interact with documents and tools, without needing human involvement.
The impact is that we can now begin to automate industries that were previously out of reach, such as asset management, accounting, insurance and legal services. Humans have traditionally been essential in these areas because only human reasoning could interpret complex, unstructured information and make decisions. In insurance, for example, much of the work involves reading documents, interpreting policies, and making judgments. Reasoning models can now handle these document-to-document flows, and that’s why AI agents are at the centre of the discussion.
What’s stopping companies from scaling successful AI pilots to full deployment?
The main barrier keeping companies from scaling successful AI pilots into full deployment is a misunderstanding of what real transformation requires. Take Gartner’s prediction that over 40% of agentic AI projects will be cancelled by 2027, and MIT’s discovery of a 95% failure rate at enterprises for generative AI pilots. But when you look at the data, the issue isn’t the technology. The tools work. The challenge is that turning technology into bottom-line impact depends primarily on people and processes. Around two-thirds of the effort is in traditional change management, such as redesigning workflows, redefining roles, and helping people adapt.
No pilot or proof of concept set up in a corner can deliver that value because it’s not designed to transform how the business operates. It’s like giving a self-driving feature to an Uber driver; yes, it helps, but the driver is still doing the same job, just with a bit more convenience. Real transformation means rethinking roles, responsibilities, and processes from the bottom up. Without that, there are no real P&L gains. The difficult part isn’t waiting for the technology to mature, but leading organisational change and ensuring it remains.
If you had to sum up 2025’s year in tech, what would you say?
For the first time in human history, we see technological change beyond new software or automation tools. Generative AI, reasoning models, and autonomous agents are emerging, reshaping how industries create value and generate and apply knowledge.
The rise of reasoning models, the use of AI directly with end customers, and the growing acceptance that some operations can run without humans in specific roles are significant signals. This is a transformation in how work gets done. We may overestimate the impact in the next year, but underestimate how much will change over the next decade. Transformation takes time, not because of the technology, but because it requires people and systems to evolve with it.
What other trends will shape financial and insurance providers’ operations in 2026?
The insurance industry will evolve from seeing AI as a department-specific tool to recognising it as essential technology transforming the entire value chain. The era of the traditional pilot mentality will end as more organisations’ pilots mature and focus shifts to ROI-driven use cases. By 2026, the widespread success of agentic AI is expected to become common. We’ll see fewer isolated applications and more systemic redesigns in policy underwriting, claims assessment, risk modelling, and customer service.
