When a company decides to adopt automation powered by Artificial Intelligence, the most common mistake is not a technical one: it is starting with the wrong process. A good starting point produces visible results quickly and builds internal confidence; a bad starting point burns budget and time without proving any value. This is the four-criteria method we use at AutentIA during the Assessment phase with every client.
1. Frequency and volume
How many times a month does this task repeat? A process that happens 200 times a month with small variations is a far better candidate than one that happens 3 times a year, even if the latter looks "more important". Automation pays off at volume: every repetition multiplies the savings.
2. Time each run takes
Frequency alone is not enough: the task also has to take up meaningful time from a skilled person. Processing 200 one-line emails does not justify the same effort as putting together 200 quotes that take 40 minutes each. Multiply frequency × time per run to estimate the monthly hours at stake. That number is what justifies the investment — or doesn't.
3. Clear rules vs. human judgment
Processes with explicit, consistent rules (extracting data from a CAD drawing, checking a site log against a budget, reviewing expiry clauses in a contract, confirming appointments for a medical practice, or classifying incoming requests) are ideal for a first project: AI can run them with high accuracy from the very beginning. Processes that depend heavily on expert judgment and ambiguous context (negotiating a contract, deciding the final design of a facade) still benefit from AI as an assistant, but not as a first automation case — in those cases it's better to iterate carefully.
4. Risk of error and cost of failure
A process where an error is easy to catch and correct (a draft report that someone reviews before sending) is safer to start with than one where an error has costly or hard-to-reverse consequences (an automatic payment, a structural calculation with no review). For the first project, prioritize processes with a "safety net" — where a person validates the result before it takes effect — and leave the higher-risk processes until the team has built up trust in the system.
Putting it into practice
Make a list of 5 to 8 candidate processes and score each one from 1 to 4 on the four criteria above. The process with the highest total score — not necessarily the "most important" one for the business, but the one best suited to being automated quickly and well — is your starting point. What a good example looks like varies a lot by industry: at an architecture or engineering firm, extracting quantities from CAD/BIM drawings, checking site logs against the budget, and generating recurring reports; at a law firm, reviewing expiry clauses in contracts and triaging incoming case files; at a medical practice, confirming appointments automatically and following up on pending lab results; and at almost any small business, routing incoming requests to the right person and chasing quotes that have gone cold.
This prioritization exercise is exactly what our free assessment covers: in a single session we work with your team to identify which of your processes is the best starting point.