We have a problem with how we measure success. For decades, management has relied on proxies for productivity. We count hours sitting at desks. We count the number of reports filed. We count lines of code written. It was never a perfect system, but it was functional enough to run a business.
Generative AI has broken those proxies.
When a tool can draft a marketing email, summarise a meeting, or write a function in seconds, the old inputs no longer correlate with value. If you stick to your traditional Key Performance Indicators (KPIs), you risk incentivising the wrong behaviours. You might encourage your team to generate high volumes of mediocre work just because they can.
It is time to look at what we are actually measuring.
The end of the “hours worked” metric
The billable hour has been the standard for professional services and internal project management for a long time. It relies on the assumption that time equals effort, and effort equals value. AI severs that link.
Consider a junior consultant asked to research a specific market trend. Previously, this might have taken two days of reading and synthesis. Now, with the right AI tools, they might get to a solid first draft in twenty minutes.
If your KPI is purely based on utilisation or hours logged, that consultant has a problem. They have done the work efficiently, but they have “lost” two days of billable time. If you punish them for this efficiency, they will simply stop using the tools or fill the time with busywork. You need to measure the output’s quality and speed of delivery, not the time it took to bake the cake.
Shift from volume to impact
In a world where content generation is free, volume is meaningless. We are about to be flooded with average content, average code, and average emails. Measuring your marketing team on “number of blog posts written” is now dangerous. They could hit that target by lunchtime on Monday without engaging a single customer.
The metrics must shift to the outcome. Do not ask how many emails were sent. Ask about the conversion rate. Do not track the number of features released by your developers. Track the stability of the system or the reduction in customer support tickets.
You are looking for the “so what?” in the data. If the AI does the heavy lifting on production, the human KPI should focus on strategy and effectiveness.
Measuring the “human in the loop”
This brings us to a new kind of metric: quality control. When AI does the drafting, the human becomes the editor. This is a harder job than it sounds. It requires critical thinking, domain expertise, and the ability to spot hallucinations or generic advice.
We need to find ways to track the value added by that oversight. For a customer service team using AI chatbots, success isn’t just “tickets closed”. It is “tickets closed without requiring escalation”. That proves the human configured the AI correctly and stepped in only when necessary.
Practical takeaways
You cannot fix this overnight, but you can start auditing your current scorecard.
Review your input metrics. Look for any KPI that rewards time spent rather than value delivered. These are your biggest risks.
Add a quality gate. If you use AI for code or content, introduce a metric specifically for error rates or customer sentiment to counterbalance the speed.
Talk to your team. Ask them where the metrics feel unfair. They know exactly where AI has made their targets obsolete. They are likely just waiting for you to notice.
Auchenglen Consulting 