Companies tend to measure AI productivity by how much faster an individual can complete a task, but this approach can miss where AI is actually creating value. As AI takes on more managerial and administrative work, organisations should Chief Product Officer at Quickbase, look at whether teams are making better decisions, catching risks earlier, communicating more clearly, and giving managers more time to coach
In an interview with Digital Journal, Torres outlines why companies may need to rethink how they measure AI productivity.
Digital Journal: What should teams be looking at to understand whether managerial AI is delivering meaningful business results?
Marcus Torres: Managerial AI is making it easier for teams to do more – to generate more summaries, complete more administrative tasks, surface more data, and/or accelerate individual work. But more output doesn’t automatically mean better decisions, stronger collaboration, lower risk, or better outcomes.
For this reason, a lot of organizations have been moving away from task completion metrics and toward measuring value: Did we make a better decision and eliminate friction? How much risk can we eliminate? Do managers have more time to coach, and is the team operating better as a result?
It’s less about understanding how well the overall system of work is orchestrated, and instead whether or not that orchestration is improving:
Time to outcome: How long does it take the team to make and approve a decision? How quickly does the team get from starting the work to producing an actual useful business result?
Quality: Are we catching risks earlier? Is there less rework? Do employees have greater clarity about what they need to do?
Trust and control: Is AI respecting the same permissions as the people using it? Is AI operating within the guardrails we’ve established? Is the team curating the output of AI to drive the outcome for the business more effectively?
DJ: How can companies assess whether AI is improving communication, decision-making and coaching, rather than just increasing output?
Torres: Driving better efficiency in an obsolete way of working increases (AI) costs, not value. We need to reimagine the orchestration of people, processes, systems and now agents so the team can optimize for outcomes and exercise judgment.
For communication, are you seeing fewer miscommunications and missed steps in customer-facing work — and are your user satisfaction scores rising as a result? For decision-making, are your decisions holding up better over time? Are potential risks getting flagged earlier? For coaching, are more of your direct reports getting promoted? Is your internal engagement score trending up over time?
That’s one of the most important mindset shifts with AI. Instead of asking, “How do I make this existing step happen faster?” ask, “Does this step need to exist at all?” The organizations that get the most value from AI will reimagine the workflow rather than simply automate the workflow they already have.
DJ: How can organizations get a clearer sense of the role AI is playing in overall team performance?
Torres: It’s actually very difficult to separate AI’s contribution from everything else affecting team performance. Teams are dynamic; leadership, staffing, incentives, workload, process changes, data quality and organizational culture all play a role. In this sense, we should be careful about making broad claims like, “AI increased productivity by 30%.”
The goal is not to prove that AI caused 100% of an improvement. It’s to build enough evidence to know that AI is materially contributing to a better way of working, and whether that new way of working is actually sticking.
DJ: What are some warning signs that greater AI-driven productivity could be creating challenges around collaboration, ownership or employee wellbeing?
Torres: When output increases but collaboration starts to weaken, employees may feel less ownership over their work or become less able to explain, defend and take responsibility for what is being produced. This is what I call the “uncanny valley of work,” where AI-generated work resembles your own but feels hollow because you didn’t actually create it.
Successful AI integration preserves authorship: people establish goals and judgment points while AI handles the background work. Organizations should watch for signs of fragmentation, such as employees creating their own agents, apps and automations outside established architecture and governance, because that can lead to duplicated workflows, disconnected data and systems that IT cannot adequately monitor or support. Skill atrophy is another warning sign, especially when people become less capable of exercising judgment without AI support.
DJ: What should companies consider when piloting AI to understand its effect on leadership and management before a broader rollout?
Torres: Start with identifying a leadership problem. By establishing a baseline, you can measure your goals more accurately whether they’re measuring whether managers make decisions faster, spend more time coaching, give employees more clarity on the mission or gain better context on their work. Pilots should also explicitly define what AI can do autonomously, where human judgment is required and who ultimately owns each decision.
