(Listen to the podcast version of this blog here).
From an overheard conversation:
"You are done with your appraisals? Already? How'd you manage that?"
"Ah, ChatGPT is a game-changer! This year, I'm not sweating this stuff."
It’s the season of performance conversations and many of us will turn to new tools hoping to make a daunting task a little easier. Some might approach the use of artificial intelligence sheepishly - it seems like cutting corners. Others will see AI as an ally that can get them through a bureaucratic “to do”.
AI certainly excels in synthesizing data. It can shift through vast amounts of information, highlight themes and communicate issues. It has the potential to support managers prepare for a focused, development-oriented conversation. Studies confirm that using AI in performance management reduces bias, positively impacting diversity, equity & inclusion. For our part, we have no issue with AI as a tool to support the drafting process. But make no mistake. AI can’t assess performance for you. Yet…
Wisdom. It’s your job
At the heart of performance management is our capacity as leaders to observe, assess, diagnose performance, and communicate meaningful feedback, that improves future results. This task remains firmly in the realm of the manager. And noone can do the thinking for you.
Performance assessment is simple in many respects. First and foremost, you need to form a point of view on whether expectations were exceeded, met or not fulfilled. You need to to review the expectations you set and how clear and achievable these were. In approaching a point of view, the best leaders will want to take a structured approach.
For example, leaders will want the opinion of stakeholders. Many will also consider getting inputs directly from the employee.
They will want to have a structured thought process, similar to the one we discussed in the article "FairTalk: Moving Beyond the Conversation in Search of Increased and Better Feedback". For example, they’ll comparing the performance of an employee against the performance of their functional peers. It may be appropriate to compare performance with employees in different functions, but of a similar level or grade. You could compare performance with industry “best in class” for a role if you think you know what that looks like. And you can try other comparisons too. For example, comparing the employee with their best performing subordinate for insight into how their strengths and development needs.
There may be other considerations. Did priorities change during the year? Were there head winds or tail winds which impacted performance? Was the employee new in role?
One of the key challenges in performance appraisal is the risk of unconscious biases and assumptions that may cloud our judgment. To minimize that risk, you need to test your assumptions, in a series of ways, that make evalation and diagnoses a conscious process. So you might use data too. How does the distribution of performance ratings compare to other managers? Are you too strict? Too lenient? Are there patterns in the population. For example, do remote employees do less well than those in the office?
It is this process of subjecting our diagnosis to rigorous review, using our own structured thought process, inputs of others and if possible, data, that minimizes the risk of unconscious bias.
Intelligence, artifical or otherwise, may be the capability to think, but it our wisdom that ensures we apply our knowledge and experience in a meaningful and ethical way.
If you can’t name it, AI can’t describe it.
Once you have formed a fair point of point of view regarding an employee’s outcomes against expectations, you need to ask yourself “why”. What behaviors did you see, or not see, that resulted in these outcomes and what different behavior would have resulted in even better outcomes. You will ask, "What is the observable behavior I want to see, but don’t?" This clarity on the behavior is crucial. We often encounter leaders who express concerns about their direct reports' performance but struggle to precisely articulate the issue.
It is the process of asking why that enables you to name the specific opportunity or gap which would, when addressed, have significantly elevated performance. This process of naming the gap is what makes the issue tangible. And once it's tangible, it becomes something you can discuss, document, measure, track, build, coach … And only once it is named can hold an employee accountable for improving it.
If you cannot name the behavioral gap, you cannot expect it to be addressed. Feedback based on vague generalizations, sweeping statements or hyperboles won’t help some develop. And if you can’t describe it, AI certainly can’t (well, it can, but that's what we call "hallucinations").
Even here, leaders need to apply wisdom if they are to help employees break down broad, complex skills into observable, actionable behaviors As leaders, we add value when we can name the gap we see, and the learnable behaviors within a bigger competency (microskills) which employees can deliberately practice.
AI can professionally summarize your feedback. Right or wrong.
Once you can speak to the why (1. Tell me why it matters), the what (2. Tell me how I'm doing), and the development opportunity (3. Tell me what to do), you’re ready to give feedback using our 3-step formula. If you have done your due diligence, the feedback will not only be fair, it will be focused and credible, because you invested time in understanding performance relative to expectations, and behaviors relative to others. You will be able to detail the highlights what’s working and what’s missing. And you will be able to describe gaps in terms of learnable skills, making your feedback helpful and actionable.
Here, AI can help. With the right inputs, the drafting process may be a little easier. And we’re all for that - especially if the time saved is redirected .. to helping the employee perform better.
But only after wisdom and judgment is applied. Without it, your feedback may be wrong. And AI will summarize it, beautifully.
A recent study in the Strategic Management Journal found that AI-generated feedback was significantly more accurate and consistent with the performance diagnosis, as well as more relevant to the employee. Great news for both managers and employees! However, when the employees found out that the feedback was written by AI, they reacted negatively – even though the result was objectively better than the one created by their managers. A mindset shift is needed in how we view technology in our day-to-day.
Conclusion
While AI can streamline the process of summarizing performance data, the executive function of observing performance, diagnosing behaviors, assessing outcomes, calibrating assumptions and providing feedback cannot be outsourced. By investing time in this managerial imperative, you pave the way for genuine development and improvement.
My boss. Thinking about me, my performance and my needs. It’s how most of us would want to be evaluated.
The views expressed here do not express the opinions of the organizations and institutions with which we are affiliated.
Comments