Blog

Generative and agentic AI

What is AI actually delivering right now? And where will we be in 2027?

18 Nov 26
5
min read

In this blog, we explore one of the biggest IT trends of the moment: AI. We focus primarily on trends within generative AI and the rise of agentic AI. What is AI actually delivering right now? And where will we be in 2027? We asked Technical Director Rik van Berendonk and gained some interesting insights.

The current state of affairs: AI between hype and reality

Since the AI train left the station, almost every company has jumped on board. AI has become a top priority for most organizations. Tools like Gemini, ChatGPT, and Copilot have become indispensable, and the stream of new AI applications seems endless. Over the past year, GenAI has become even smarter and better at understanding context with GPT-5.

However, the view of AI as a silver bullet turns out to be more nuanced. The anticipated productivity gains have yet to materialize. MIT Media Lab concluded that 95% of organizations have not yet achieved a return on their AI investments. According to the Gartner Hype Cycle, AI is also still in the peak of inflated expectations. We already predicted this last year, although back then we thought we’d be a bit further along by now.

“AI as a fun gimmick isn't enough”

Technical Director Rik van Berendonk on the current state of affairs:
“AI as a fun gimmick isn't enough. Sometimes it actually creates bureaucratic clutter: AI-generated texts that add little value and create extra work. Often, that output still needs to be manually checked or rewritten.”

No productivity boost, no profit, and no time saved: what good is AI, then? Still, it remains promising. Rik: “Although there are no independent studies yet demonstrating a significant impact on productivity or profit, users do experience tangible improvements. Especially in healthcare, where administrative and repetitive tasks are common. Objective figures are still to come, but the initial signals are positive.”

Agentic AI: autonomous agents

A key development is the rise of autonomous agents: AI systems that independently carry out tasks, manage workflows, and make decisions without continuous human intervention. By 2027, the expectation is that we will no longer need to babysit AI. You give an instruction, and the agent executes it from start to finish.

A prerequisite for agentic AI, however, is having your data governance and data quality in order; in this area, almost all organizations still have work to do. Also read the trend: data governance as a prerequisite for AI.

Agentic AI raises questions about the role of humans in their work and the balance between human and artificial decision-making. Organizations would do well to get their policies and data infrastructure in order now to make this transition in a controlled manner.

Specialized AI models

There is a shift from generic to specialized AI models, trained on reliable and sector-specific data within domains such as healthcare, finance, and government. These specialized models can be both large and small language models (LLMs and SLMs), with SLMs becoming increasingly popular for local use due to their lower computational costs and better data control.

Organizations that want to use such specialized AI models must choose between cloud solutions and local implementations. Local options are becoming increasingly attractive due to data privacy and technological independence.

From hype to strategy

Truly integrating AI into business processes takes time and requires the right strategy. Rik: “The organizations that don't see AI as an end in itself, but as part of a clear strategy, will reap the rewards. Think beyond the hype and focus on sustainable value: optimize processes instead of individuals, set the right KPIs, and actively involve employees. Ultimately, it’s not the companies with the most tools that survive, but those with the best integration between people, data, technology, and culture.”

Ultimately, it’s not the companies with the most tools that survive, but those with the best integration between people, data, technology, and culture.”- Rik

Other AI developments to watch:

  • Multimodal AI: continues to develop in breadth: text, image, video, audio, and code. The expectation is that multimodal AI systems will continue to improve and will soon be able to roll out a concept across different media.
  • Human-AI collaboration: new roles are emerging, such as AI coordinator or prompt specialist, where people learn to guide AI. This way, people are not replaced by AI, but the work they do changes.
  • Sustainable AI: growing attention to energy consumption and environmental impact; organizations are becoming more conscious of unnecessary usage.

Agentic AI and specialized models as key trends

AI is still in the hype phase, but fortunately, it is becoming increasingly clear how it can deliver real value for organizations. Autonomous agents and specialized models hold great promise, provided your data and data governance are properly set up.

By 2027, it will be up to organizations to formulate a clear AI strategy and foster seamless human-AI collaboration: technology, data, people, and corporate culture must align to make it a true success. If done right, AI can evolve from mere hype into structural value.

This blog was previously published on ogd.nl

‍

Featured posts

We don't do off-the-shelf job openings. We create your ultimate role based on your personal ambitions and expertise.

Blog

Data governance as a prerequisite for AI

Want to use AI successfully? Then your data really needs to be in order. In 2026, data governance will become even more important; and not just because it is a prerequisite for effective AI use.

Blog

Data product as a measurable business asset

Value-driven cost management in a data mesh environment.

Blog

The art of reducing complexity in software

Software development is often about adding new functionality. But as a solution grows, a different challenge arises: how do you ensure it remains understandable, maintainable, and extensible? The biggest obstacles to successful software projects are often not new features or technologies, but the complexity that builds up over time.