
Does your process need AI… or just simplifying?
Artificial intelligence can improve a process. But first you need to understand what slows the work down. Depending on the problem, the right answer may be one step fewer, a “classic” automation or a targeted use of AI.
“We’d like to put AI into our process.”
The intention is understandable. Teams are short on time and repetitive tasks end up weighing on them. What AI can do makes people want to act.
Before choosing a tool, ask this question: what exactly is making you lose time?
It may be incomplete information, double data entry or an approval that always waits on the same person. Part of the problem then lies in how the work is organized.
Automating a useless step is also a way of keeping it.
Understand the work as it is actually done
The starting point is to follow a real file with the people who handle it. Where does the request arrive? What information has to be found? When does the file wait, and why?
This observation helps distinguish time spent working from time spent waiting. Speeding up the drafting of a document will have limited effect if it then sits for several days in an approval loop.
Take a simple example: preparing a quote. A request arrives by email, an employee looks for the missing details, copies the data into a file, then has the price approved. This same process can benefit from several improvements, in different places.
Simplify what can be simplified
If requests often arrive incomplete, a short, tailored form can limit back-and-forth clarification. If the same contact details are entered in several files, let’s keep them in one place and reuse them. Each approval should also play a clear role, proportionate to what is at stake.
For our quote, this means defining the required information, identifying the source of prices and specifying which situations require a particular approval.
These changes can be tested without waiting for a large technology project. Involve the people who do the work: they know the exceptions that procedures don’t show.
Automate what follows stable rules
When data is structured and rules are clear, a more classic automation can take over: creating a file when a form is received, transferring contact details between two tools or sending a reminder at a known deadline.
For our quote, it could fill a template with the client’s information, apply an approved price grid and route the file to the right manager based on the amount.
Before developing a new solution, check what your software can already do. Feasibility will depend in particular on its integration capabilities. Also assign someone to handle exceptions and keep the rules up to date.
Use generative AI to process text and documents
Generative AI can help read requests worded in different ways, spot requirements in documents or prepare a summary from several sources.
For our quote, it could extract the needs expressed in the client’s email and prepare a first draft from authorized references. This use must be tested on representative files.
You also need to check what information can be used and in which tools, especially when it is confidential. Prices and terms must come from approved sources. Before sending, the person responsible reviews the proposal and validates the commitments, with the information and time needed to do it properly.
The approaches can therefore be combined: simplify the flow, automate predictable actions and use AI for certain information-processing tasks.
Measure the gain after the checks
To choose, look at what changes across the whole of the work. Preparation time, lead times, corrections and the quality of the result provide useful benchmarks.
A response produced quickly may require a long review. An automation may shift the load toward handling exceptions. These efforts must be part of the comparison, along with operating and maintenance costs.
Time freed up is not automatically a financial saving. It can be used to clear backlogs, better support clients or absorb more requests. You need to specify the value sought and check that it materializes.
Choose a well-defined process, set a few baseline measures and test on a set of files. The results will let you continue, adjust or stop. You can learn and simplify as you go, without waiting for a perfect process.
Alexandre Cieux
Director, Innovation and Sustainability
Have a process in mind?
Not sure which approach is right? Discover Pulse30 to frame the need, test a targeted solution and measure its value before deciding what comes next.








