Making Data Smart: Key insights on how to create business value with AI

In this blog, Peyman Kor gives you his three key insights.

How do we realize business value from all the ongoing AI projects? Every day, we read about AI projects that lack return on investments made. A recent MIT report states that 95% of AI Agent projects fail in production.

The big question is: While AI models have the capacity at “PhD-level" intelligence, what stops them from creating value for businesses and organizations? 

We got some good answers participating in the Make Data Smart conference in Oslo in September. The conference was packed with inspiring talks on creating value with data and AI. Under, I have listed three key insights from attending this event:

AI first or product first?

If the AI model's performance is converging, and everyone has the same access to the same models, is a “model” a competitive advantage? The conclusion was that AI-powered “Products” that are deeply integrated into businesses can create competitive advantages, and we should go forward with a “Product First” approach. It seems that competitive advantage lies in how companies integrate LLM while providing a proper context to the AI by looking at their in-house data, knowledge-based documents, and business processes.

The paradox of automation

Another interesting concept I learned is that the more the system becomes automated, the human contribution becomes even more crucial. I can see this when I do AI-assisted coding. The more the assistant becomes powerful, my role in guiding and asking the right questions becomes more crucial. This is important to remember now that the AI is becoming a bigger player and can automate tasks with greater ease.

Three levels of value creation with GenAI

There are three levels of value creation with GenAI and Large Language Models (LLM):

• Level 1: Simple summarization or finding the insight in a text.
• Level 2: Involves connecting business data and internal knowledge, and asking LLM questions about it.
• Level 3: Given all the data, internal knowledge, and external knowledge, you ask LLM for suggestions.

In more technical terms, level 2 can be categorized as traditional “Descriptive Analytics”, which is useful when you have trend and dynamic data, and you want to be aware of trends. Level 3 can be considered as more “Prescriptive Analytics”, which affects how we take business actions and decisions.

AI agent projects at Antire

At Antire, we believe in the transformational power of AI Agents when integrated into the business process. Essentially, our view is to start by scoping and finding “low-hanging fruit” use cases that can create the highest value with an agentic approach. Then, building on that use case (easy win), going to more advanced projects such as having an AI agent as a personal assistant, which can collectively increase the intelligence of the organizations, leading to better actions and creating more value. Our AI agent project covers both “Automation” and “Recommendation” processes.

One notable example of Automation is “Expense Report Agent”, which processes receipt images shared through chat interfaces, extracts key information (date, vendor, amount, category), creates expense reports automatically in the ERP system, and validates entries against expense policies.

Another example of an agent that not only automates the process but also provides insight is the “Digital Marketing Agent”. Essentially, now the agent goes and extracts all the relevant data from Google Ads data, as well as the client's internal knowledge, and comes up with weekly recommendations on “What is the best course of action” to do for next week. The client has been very interested in the solution, as reading that AI Agent report provides good internal discussion on what decisions to take. We see this as “Collaborative Intelligence”, where you blend the “Data Crunching” capacity of the AI Agent with human judgment, for the best outcomes.

Pictured above: Peyman Kor and Jens Erik Kristiansen from Antire at the Make Data Smart 2025 Conference.

The path forward

The Make Data Smart conference ended with startup pitches from four startups in Norway, two of which were directly GenAI-related. As we can see from these startups, as AI matures, companies that embrace the technology early will gain a competitive edge. To embrace, I found that it is essential that companies completely “rethink” how work gets done. In the Software as a Service (SaaS) era, humans were interacting with the software. Now, you can have multiple agents, each planning, running some tasks, and delivering outcomes. Going forward, the key is to collaborate with AI agents, where you can drive business decisions together with them.

The journey to embrace AI in business is both exciting and complex. AI is a disruptive technology with incredible development speed. Businesses can really gain a competitive advantage with AI if they approach this challenge through the lens of curiosity, responsibility, and innovation. The key idea is to be “proactive” and start experimenting with AI while understanding the capacities of AI, find specific use cases that can create value for the organization, build trust, and pave the path for more company-wide adoptions.

If you would like to start a Data and AI Agent project, look no further than this initiative, “Proof of Value Sprint,” which is specifically designed for companies to prototype AI Agents.