AI is not a quick win: Why AI winners act now, not tomorrow

While chat-based generative AI in particular has created high expectations of quick wins, competitive advantages are ultimately built on the same principles as any other IT development: a clear strategy, well-defined business objectives, high-quality data, and the courage to scale successful experiments.

Connecting the Dots: Scaling AI Across the Enterprise 

The first wave of AI adoption has been mainly through generative AI, improving individual productivity in a fragmented way. Some employees benefit significantly, others very little, and only the first steps have been taken toward scaling these benefits across the entire organization.

Now, the focus is shifting from individual productivity gains to team-, function-, and organization-level impact. The key question is how the entire organization can operate more intelligently.

Chat-based AI solutions could also have contributed to extremely high expectations. Because we are accustomed to using conversational AI models like ChatGPT in our everyday lives, it is easy to assume that AI solutions for entire organizations could be built just as effortlessly.

In reality, no AI solution is a silver bullet. Developing trustworthy, cost-effective AI tailored to an organization does not happen automatically; investments are required. For that reason, organizations must look beyond quick wins and, instead, focus on long-term resilience, structured development, and continuous learning.

Time to wake up to the potential of AI agents

Threat or Opportunity? AI agents unlock significant potential but carry also risks. When built and integrated in a controlled way, they remove bottlenecks and accelerate decision-making. Without clear governance, however, they can act faster than the organization can understand or respond, leading to errors and loss of trust.

One of the most significant developments in the AI landscape is the shift from traditional automation to intelligent, autonomous AI agents. Unlike chat-based solutions that are used to solve individual issues, AI agents can operate independently. They execute multi-step tasks and even entire processes, leverage multiple background systems, and can even make decisions based on predefined business objectives without constant human supervision.

It’s important to understand how AI agents differ from traditional automation. Conventional robotic process automation relies on clearly predefined if–then rules. While this approach is effective for streamlining repetitive tasks, its ability to adapt to changes and uncertainty is limited.

AI introduces a new level of intelligence to automation. AI agents are capable of more autonomous decision-making and of generating new information based on the given context. The greatest value often emerges when agents are chained together to support broader use cases. However, this is also where new risks lie. Creating a chain of AI agents means using one or more AI models (so-called black boxes) whose operation is determined by what they have been trained to do, but that aren’t similar to traditional automation.

For this reason, AI adoption requires clear process definitions and, additionally, a robust AI governance model. Organizations must define how risks are managed, how production use is monitored, how wrong or non-optimal use is tracked, who makes decisions, and when human oversight is required. Continuous monitoring is essential to ensure that AI solutions operate within the intended boundaries, cost estimates, and objectives.

The fundamentals of IT development apply to AI

The same principles apply to AI as to any other form of IT development. Organization-wide quick wins are not automatically guaranteed. Instead, companies must go back to basics and define what needs to be achieved and developed.

Scaling AI requires paying special attention to the following areas:

• Strategy and business objectives: What do we want to achieve, and why?
• Processes: Which functions and processes are we improving? Which processes can be simplified or even eliminated? Who is affected by the changes, and how do we bring people along?
• Data: Where is the data located? Can the data be located in cloud services, or is an on-premise service needed? Is it high-quality and secure? How accessible is it? What is the level of data security and data protection? It is also important to remember that AI is not omniscient; without deliberate investments, it has no access to internal terminology or organizational culture.
• Skills, roles, and collaboration: How are AI systems and their decisions governed and monitored? How do we ensure that people’s skills remain up to date?
• Operating and governance model: Without a common operating and governance framework, individual AI experiments within departments can lead to uncontrolled chaos.

The growing use of artificial intelligence shouldn’t and can’t be restricted. On the contrary, increasing its use should be encouraged. At the same time, it must be ensured that individual units’ own AI experiments do not lead to an uncontrolled situation, and a shared AI governance model must be established for the organization. Regulatory requirements must be met, issues of responsibility must be addressed, and standard productivity and cost metrics must be in place to track the progress of development projects.

Without a shared AI governance model, isolated initiatives across departments can quickly lead to unmanageable, fragmented chaos.

AI in practice: results across industries

When AI adoption is scaled into core business processes, the results can be significant across industries and functions:

• HR and internal processes: Complex backend systems can be replaced with AI assistants, enabling employees and managers to handle vacation requests, promotions, and salary adjustments through conversational interfaces. Solutions can be localized for different markets, thus ensuring country-specific compliance.
• Order management: AI agents analyze customer orders and requirements and propose suitable solutions.
• Public sector and expert work: Agents can process vast volumes of material, for example when preparing decision-making summaries, freeing experts to focus on analysis and decision-making.
• Predictive maintenance: In manufacturing, AI agents analyze large data sets and recommend maintenance actions before failures occur.
• Financial services: Beyond customer service bots, AI is widely used in fraud detection and risk management.
• Manufacturing industry: By using AI and mathematical models, not only is the production process itself optimized, but the batches to be produced are also selected based on different parameter sets.

What’s next?

Looking a few years ahead, the most radical change in organizations and business is unlikely to be a single new AI breakthrough. Instead, the deep integration of AI into everyday operations is predicted. AI will automate processes and ways of working in ways previously deemed impossible.

The key message for business leaders is clear: act now. Organizations must be resilient enough to tolerate experiments that do not always succeed immediately, while still delivering value with the tools already available. Those who learn to manage their data and build, for example, AI agents today will be the ones growing successfully tomorrow.

AI will not solve everything at once, but it will reward those who start the journey and scale their successes early.

Writers: 

Kyösti Laiho, Sales Representative, Data & AI software, IBM Finland 

Jutta Karjalainen, AI advisor & Manufacturing & Automotive Principal, Tieto Tech Consulting