Business process optimisation: should you start with AI?

In summary

• Business processes become more complex over time as organisations accumulate tools, data flows, rules and manual operations.
• Before automating a process or introducing AI, you need to understand the data, rules, responsibilities and controls that underpin it.
• The rise of AI and agents makes mapping and documenting existing processes even more important.
• Optimisation is not simply about automation: it means identifying what should be retained, simplified, removed, connected or automated.
• Improvements can be introduced progressively: removing manual data entry, connecting two systems or changing an approval step can simplify a process without overhauling the entire information system.
• Data quality, interoperability and process knowledge are essential if both people and agents are to act reliably.

Over time, organisations accumulate tools, data flows, manual operations and working habits. Processes continue to work, but they can become so complex that no one really understands them in their entirety. The arrival of AI and agents makes understanding and documenting them even more important.

Automating a process you do not fully understand does not necessarily make it better. Neither does adding AI.

Business processes age too

Machines age, operating systems become obsolete, software evolves, and we spend part of our time maintaining our solutions.

But business processes age too. An organisation grows, diversifies and changes its tools. Interfaces, imports, exports, Excel files and a few manual operations are added to make everything work together. People join, others leave. Eventually, data follows paths that no one fully understands from end to end.

And yet, it works. Until the day you want to change something.

Optimisation starts with understanding

When an organisation wants to optimise its business processes, the first instinct is often to look for a solution: a new tool or a new automation. But before changing a workflow, you first need to understand where the data is stored, how it flows, which rules apply and why certain operations are performed.

This is an important question: some operations are essential, while others exist because of constraints that are no longer relevant. Optimisation is not simply about making existing processes faster. It also means asking whether every step still has a reason to exist.

The more AI does, the more context it needs

Vibe coding now makes it possible to develop very quickly by describing to an AI what you want to achieve. I see this myself: something that would have taken a great deal of time not so long ago can sometimes now be completed in just a few hours.

But as a project grows, AI needs to understand more about what already exists: its architecture, components, conventions and rules. The more AI is capable of producing, the more important the context and documentation we provide become.

I believe we are going to see exactly the same phenomenon with business processes. An organisation may function because everyone understands their own part of the process. But how can we expect an agent to take action within that process tomorrow if no one can explain the whole thing to it?

AI forces us to make our processes explicit

For an agent to act, we need to know where the right data is, what it means, which system is the source of truth, which rules apply and which actions can be carried out without human approval.

AI therefore creates an interesting paradox: while it makes automation much easier, it also requires us to formalise what we want to automate more clearly.

Process mapping takes on a new dimension. It is no longer simply about drawing boxes and arrows, but about describing each step: Data → Rule → Action → Responsibility → Control. This documentation becomes useful not only to people, but also to machines.

From process mapping to process improvement

Mapping a process is not an end in itself. The aim is to understand the existing process well enough to identify what genuinely needs to be improved.

It starts with some fairly simple questions: where does the data come from? Which system is the source of truth? Who is involved? Which rule is being applied? Why does this step exist? Where do manual data entry, delays or controls occur?

This work often reveals very different situations. A step may still be essential but carried out in an unnecessarily complex way. Another may exist because of a historical constraint that is no longer relevant. Manual data entry can be replaced by a connection between two systems . An approval can be moved earlier in the process. A repetitive transformation can be automated.

This is not simply about looking at tools and technical data flows. It also requires talking to the people who actually use the process. This approach can, for example, involve workshops bringing together the different stakeholders . Some rules, constraints and working practices may not be documented anywhere: the knowledge sits with the teams who keep the process running every day. Reconstructing that knowledge provides a more complete picture of the existing process before attempting to transform it.

The objective is therefore not to automate the process exactly as it exists, but to determine, step by step, what should be retained, simplified, removed, connected or automated.

Only then does the question of which technology to use arise. Some operations can be handled through rules and conventional automation. Others can benefit from AI. Some may eventually be entrusted to agents capable of carrying out a sequence of actions, provided that the data they can access, the rules they must follow and the necessary controls are clearly defined.

This approach also makes it possible to move forward progressively. It is not always necessary to replace a system or rebuild an entire process. Removing manual data entry, connecting two data sources or changing an approval step can already deliver tangible benefits.

AI changes the tools, not the question

Does this mean everything needs to be redesigned from scratch? Fortunately, no. This is one of the principles that guided us when developing Simple Workspace: intervening progressively in one part of a process, removing an export, connecting two systems, automating a transformation or changing an approval step.

Technologies change. We used files, then APIs; now we are using AI and agents. But the fundamental question remains the same: what do we want to achieve, with which data, according to which rules and with which responsibilities?

AI does not make the issues surrounding business processes, data flows, interoperability and documentation obsolete. On the contrary, it gives us another reason to pay even more attention to them.

Have your processes become more complex over the years, and are you considering the role that automation, AI or agents could play within them?

We can help you analyse your existing processes, identify points of friction and determine the most relevant improvements for your organisation.

Would you like to simplify your publishing workflows?

Let's discuss your current processes and explore how to make better use of your data and streamline production.

Let's talk

JY. Jourdain

Cofounder of J2S