Is execution what makes up your “secret sauce”? The power of a formalized business logic in the age of Agentic AI4 min read·Just now--In recent months, a new set of questions has been surfacing with increasing urgency during due diligence processes — especially when looking at service companies: “how likely is it that Artificial Intelligence (AI) will disrupt the core business in the coming years”, usually combined with “how do we quantify the opportunity associated with AI for this asset”? While I would not dare to make any bold crystal ball predictions in that fast-evolving space, I have found it helpful to look at two critical enablers to start shaping an answer[1]:(i) How fragmented/connected are the systems providing the data points needed to perform core business tasks?(ii) Do key business workflows live in people’s mind, or are they systematically documented and shared?Too often, teams apply a repeatable reasoning process when dealing with a specific business scenario, but those steps fail to be captured in writing, or are inconsistently applied across individuals performing the same task. Think about business activities ranging from “determining what is the final price that a customer should pay” to “making a recommendation based on qualitative input from a client”: the more these answers rely on informal human judgement, the more vulnerable incumbents become to nimble AI-native disrupters who use streamlined, codified workflows to catch-up.So, how can a company maintain its competitive moat — one built on years of industry experience — in an era of rapid AI advancement? I believe Agentic AI presents a substantial opportunity for businesses seeking to boost their efficiency while preserving the uniqueness of their proprietary processes. By actively leveraging historical data available and encoding distinctive corporate intelligence into agentic workflows, businesses can hit the ground running and turn years of accumulated expertise into durable value.What are critical building blocks to consider to successfully configure agentic AI workflows?- Goal: What specific business question(s) are you trying to answer? Is it well defined? It is essential to have a clear vision not only of the objective, but also of what the ideal output should look like — including how the user should interact with your solution.- Data inputs: Ensure a sufficient volume of high-quality data is available and assess how it is organized across systems, and how you can connect your different data sources. Without this, any agent will struggle to deliver value.- Instructions/Actions: This is where an end-to-end mapping of the business workflows at hand is critical — including the sequence of actions (which step of the reflexion process should come first and next?), the decision tree(s) that should be followed (what should be priority data points to consider in the decision-making process?) and the stage(s) where decision(s) should be made. This “business logic” should be harmonized across all teams and functions that will interact with the AI agents ahead of any AI-workflow building.Most teams are aware of the foundational investment needed to consolidate and clean their data sources to enable AI use cases — and this is where we usually start with companies that are early in their journey –, but this “business logic” step is also likely to be one where a lot of time is spent. Ultimately, it will also influence the number of agents you should consider building for a given business process and how they should interact with one another.- One aspect that is easy to overlook — but is important in the context of Agentic AI — is the definition of constraints: in other words, what are the outcomes you explicitly want to avoid? These guardrails will define the boundaries your agent must operate within and are key to ensuring safe and useful outputs.Which core processes are good candidates to start with?- High-frequency trivial processes: Achieving widespread adoption of Agentic AI solutions across your organization will be facilitated if you start by addressing pain points that are top of mind for business users, especially those that occur frequently and were previously hard to automate.- Tasks relying on multiple data inputs sitting in different systems: If you can successfully connect the relevant data sources to solve a particular task when configurating an AI agent, you are likely to significantly enhance the user experience and demonstrate tangible value early on.Which core processes should be left aside when getting started with your first AI agents?- Error-sensitive use cases: Any agentic AI workflow will need extensive testing before you can guarantee a certain level of accuracy. Use cases where small mistakes carry big consequences should not be your entry point.- Tasks with subjective or inconsistent answers: If a single question can lead to different answers depending on who you ask or which data source you consult, it’s probably not the best starting point for an agentic workflow.Rather than chasing the perfect use case, the companies most likely to capture or maintain a lasting competitive edge thanks to AI are those that move early and fast. The ones that start leveraging their codified business processes today, even if not perfectly so, are the ones best positioned to trial, learn and evolve their Agentic AI capabilities over time. The sooner they begin, the faster their agents (human or not!) will be able to adapt and to continuously improve the very processes that have fuelled their success to date. In that regard, AI is no different from any previous technology coming to maturity — isn’t it Leonardo da Vinci who was already referring to “the urgency of doing”? Quoting his words, “knowing is not enough, we must apply. Being willing is not enough, we must do”.[1] This article does not consider any specific regulations that would apply in a given industry and impact the ability to leverage AI in day-to-day business