As discussed recently at the TCG Retail Summit, the business landscape must move past the theoretical hype of Artificial Intelligence. Many organizations have spent the last few years painting little “dots of colors” across their operations with isolated AI use cases. However, companies do not truly transform simply by executing isolated AI experiments. The ground has shifted, and the industry is now entering the era of Agentic AI.
Unlike traditional AI, which relies on historical data to classify and predict, or Generative AI, which creates new content, Agentic AI is a piece of software that combines reasoning and action. It can perceive information (like reading an email or a document), reason to understand the context of a problem, and act by triggering an action in another system. Agentic commerce brings this revolution to internal processes, helping salespeople, enhancing retail engagements, and fundamentally altering how work is done.
To move from theory to reality, one can look at the recent transformation at Carrefour. A market research process that typically required 20 people to spend one to two months generating PDF reports for 400 projects a year was completely reimagined. By placing a market study agent directly into the hands of the real estate development team, 200 unviable projects were instantly filtered out, and comprehensive studies for viable projects were generated in just two minutes. This achieved 50% to 70% FTE savings, allowing talented teams to be relocated to higher-value tasks.
Transformations like this require a new playbook. Based on live implementations in the trenches, here are the ten key insights defining agentic business transformation:
1. Agentic is a catalyst, not the objective
AI agents must be used to drive broader business transformation rather than being the end goal itself. For example, a car manufacturer might leverage agents to gain speed in production, while a telecommunications provider uses them to deliver a better customer experience. Agentic technology is simply the supporter of these overarching strategic visions.
2. Set a “crazy” North Star
Drive the transformation with highly ambitious targets regarding speed, full-time equivalents (FTEs), millions of euros in performance, and lead times. Aiming for a fluffy 5% or 10% cost reduction means the economics of building the agents will not even pay for themselves. A crazy North Star is required, such as a retailer aiming for a strict 30% cost efficiency gain or a manufacturer aiming to be 50% faster across all processes.
3. It’s a “horizontal” energy
Companies naturally grow in vertical silos by function. Agentic AI, however, is fundamentally about breaking down these silos and focusing on end-to-end processes and workflows across the enterprise. Because it cuts horizontally across the organization, this transformation must be driven by the CEO; leaving it to individual departments will freeze the organization with disconnected agents everywhere.
4. Don’t automate a flawed process
There is a massive risk of using agents just to automate a broken or inefficient process. Agents should not just speed up bad habits. They are both a pretext and an accelerator to completely reinvent processes from the ground up using a zero-based logic.
5. Prioritize with the right criteria
To find true “agentic value,” leadership must look for tasks involving multiple systems, high repetition, and numerous team interactions. Excellent candidates for agentic reinvention include reviewing promotional catalogs for pricing errors or managing finance closing anomalies month after month.
6. Reorganize, don’t just automate
Real value comes through reorganizing the work and the interfaces, both upstream and downstream, rather than just automating existing, isolated tasks. For example, in procurement or market research, putting the agent directly into the hands of the person who has the initial need completely reshapes the workflow around them, rather than just automating the middle of the task.
7. Invest in “agentic” readiness
Before scaling, functions must be assessed for data readiness, semantic readiness, process readiness, and trust. Perfect semantic data is non-negotiable. If a company has never defined what a product hierarchy is or established a standard language for employee roles across countries, an agent cannot navigate the systems.
8. New tech requires new talents
Traditional data science skills are no longer enough. Organizations need Product Owners with deep process skills who can act as “AI Catalysts” to frame use cases, alongside “AI Builders” who can leverage low-code platforms to implement everyday solutions.
9. Make or buy? It depends!
The approach must scale with the impact. A “Make” approach is required for highly transformative, company-wide initiatives that cut across multiple functions (like promotion management). Conversely, off-the-shelf “Buy” solutions are suitable for boosting individual employee productivity, and “Low/No Code” platforms are ideal for building team-level efficiencies.
10. Agentic AI is “no magic”
At the end of the day, a successful agentic transformation requires a strict balance. It is only 20% technology, while heavily relying on 30% change management and 50% end-to-end process reinvention.
If leadership is not ready to picture how functions will be drastically disrupted, or if they are not prepared to do the hard work of fixing underlying processes, the technology alone will not save the business. But when an organization is willing to fundamentally rethink how work gets done, the agentic revolution offers an unprecedented opportunity to unlock productivity and growth.
