🧾 It's Not the Function That Disappears; It's the Segment That Adds No Judgment
Automation Risk Lives in Tasks
Within Colombian industrial companies there's talk of the human resources department, the quality department, or the purchasing department as if everyone working in those functions faced the same level of risk against automation. That way of thinking groups two types of work in a single label that function very differently within the same position.
Automation doesn't fall equally on all professionals in a transversal area; it falls on the proportion of the day devoted to tasks where there's no context to interpret, exception to solve, or direct responsibility to assume.
Think of the treasury team at a manufacturing company in the Coffee Region. There are two profiles sharing the same job title but not the same type of work. One extracts the bank position report from the financial portal, formats it in a spreadsheet, and sends it to management. Updates the accounts payable matrix, generates the weekly VAT report from accounting software, and does portfolio follow-up with the same template email. All those tasks are necessary, consume real time, and have consequence if not executed. They also produce predictable output from structured input, without anyone interpreting something outside expectations.
The other profile does something different with the same information. Detects that the formatted cash flow shows comfortable liquidity for the next thirty days, but a customer with high participation in invoicing has been paying fifteen to twenty days past contract deadline for three months. Crosses that pattern with the conversation from last week with the commercial director and with the payment behavior from the previous year. Writes an analysis for financial management recommending they review the credit limit before dispatching the next order. No system generated that analysis. The person who connected the three data points was someone with access to operational context.
It's not the function that disappears; it's the segment of the function that adds no judgment.
MIT economist David Autor developed a framework that has been a reference in labor economics for decades: the unit of analysis for automation isn't the job but the composition of tasks within the job. The Brookings institution applied that framework in a 2019 analysis on automation and artificial intelligence and concluded that office administration jobs are among those with highest relative risk because more than seventy percent of their tasks are potentially automatable. That proportion applies specifically to routine, transactional, and recording content within those jobs, not to the functions of context interpretation, exception handling, and decision-making with direct consequence.
The OECD published in July 2023 its Employment Outlook with a specific chapter on AI and labor market. A survey conducted in manufacturing and financial sectors of seven countries showed that AI use can generate positive results in job satisfaction and working conditions, but also revealed concrete risks in segments with higher routine content. The OECD estimated that twenty-seven percent of jobs are found in occupations with high automation risk when all available technologies are considered, including generative AI (systems that produce text, documents, or analysis from natural language instructions, available as a service the provider operates over the internet).
What that figure demands understanding is that within the same transversal area two profiles can coexist. Whoever designs controls for a management system and whoever records them mechanically share the same org chart but not the same exposure level. Whoever interprets why a quality indicator dropped in the quarter and whoever formats the indicator report produce different things even though the job title is the same. Automation doesn't affect them equally because, even though the business card matches, the work content is fundamentally different.
Consider the document management area of an agroindustrial company in Huila. Current tools can absorb tasks like classifying documents by type and destination area, alerts for contract expiration, extracting key information from scanned documents, and generating document indexes from templates. Those are activities that today consume hours of people with technical or professional formation, with high repetition, low variability, and predictable result. What those tools can't do belongs to a different segment: determine whether a contract modified by the supplier has implications for the quarter's audit, recommend whether the discrepancy is an administrative error or a condition change that needs escalation to the legal team, or assess which document version has evidentiary value in a dispute.
It's not the area that disappears; it's the segment of the function that adds no judgment.
When spreadsheets became standard in accounting departments of Colombian companies during the second half of the nineties, auxiliaries who spent most of their day manually calculating balances, depreciation, and amortization quotas lost that task as the job's reason for existing. Accountants who could interpret why a period balance didn't close, or how an adjustment to withholding affected month cash flow, became more necessary, not less. The tool absorbed calculation; the function reorganized around interpretation.
Recommended Resources
- Brookings Institution. Automation and Artificial Intelligence - How machines are affecting people and places (report)
- OECD. OECD Employment Outlook 2023 - Artificial Intelligence and the Labour Market (report)
- World Bank. World Development Report 2019 - The Changing Nature of Work (report)
First map the tasks of your position between those of structured production and those requiring context interpretation, then identify which of the first could be absorbed by available tools, then concentrate development of your profile on the second by assuming decisions with verifiable consequences, finally review the Brookings report to contrast your job composition with the segments that evidence shows as most exposed.
In your company, which transversal area has today the highest proportion of routine work and how are they measuring whether their professionals are migrating toward judgment functions or remain concentrated on tasks a system could absorb? 🧾