Early this year, we let every employee at ClickUp build their own AI agents and skills. Within months, we were running five agents for every employee, which felt like a transformation until we looked at the actual output and efficiency. In my own finance org, only 30% of employee-built agents and skills actually moved the needle.
When we looked into which AIs brought the most value, there was an obvious pattern. The successful agents were created by employees who understood where a process in their work typically tends to break, and could turn that knowledge into clear rules and processes for artificial intelligence.
We went from treating AI skill-building as an open invitation and instead focused on the people who were building AI agents that moved the needle.
That fixed the quality problem, but handed us a harder problem to solve. When you take your best talent’s craft and turn it into an AI agent, what does that employee get in return for their expertise and labor? As more companies have the best employees train their AI, leaders must have a response to this question.
Define the bar, then pay against it
When we changed our AI skill-building process, eligibility depended on a manager’s recommendation and a strong performance review. But identifying the right people is half the work. Companies also need to establish what qualifies as a high-value skill, and what it is worth, before anyone starts building one.
We created a framework that categorizes skills, assigns value, and determines the reward. It’s simple but makes it clear what employees should aim for and gives us a consistent way to reward them.
At ClickUp, every skill falls into one of four categories, and each category has its own measure of success. One, revenue acceleration skills are measured by the pipeline or bookings they add, and two, cost elimination skills are measured by the spend they remove, like canceling a software we no longer use. Time compression skills, three, are measured by how much faster a core process runs, like closing the books or responding to a customer. And executive bounties, four, are specific problems our leadership team has identified, each with a reward attached.
The category tells employees what type of skills to build with AI, but the proof of impact is what determines how much the AI skill is worth. Before creating a new AI skill, employees must record where the measure stands today, like how many sales meetings aren’t rebooked each month or how long it takes to close the books.
Then they can build, and once the AI skill has been running long enough to show a consistent pattern of value, finance compares the new numbers against the starting point. Finance then subtracts the cost of running the AI skill, including tools and maintenance, and what remains is the value the employee created.
We then look at what we would have paid to get the same result a different way. Hiring even a few experienced sellers is a major expense, so when one person’s AI skills deliver that same capacity, a standard raise no longer reflects what their work is worth to the business.
To determine the reward, we also consider how long the value will last. A skill that solves a single, contained problem earns a bonus. A skill or system the employee will keep running and improving earns a permanent raise, since the company will keep benefiting from that work every year.
Together, these steps give us a dollar amount for each AI skill created, which we can use to determine the compensation ranging from a one-time bonus to becoming part of our “unicorn” group. Unicorns are employees whose AI work changes what an entire team can do, and they receive a merit increase of more than 30% along with eligibility for compensation up to $1 million at the top of the band.
For example, one growth operations manager identified 17 places where we were regularly losing potential sales and built 11 AI skills to address them. Those skills recovered hundreds of thousands of dollars in pipeline and increased the number of sales prospects by 25%. They also saved every sales rep about 90 minutes in their day, which is equivalent to seven sellers’ worth of capacity. Because the results were large, verified, and directly tied to revenue, this employee joined our “unicorn” group.
Reward your talent for making their expertise reusable
Your best people are already training your AI. They are handing over the part of themselves that made them hard to replace, and the company can keep using that knowledge long after they have moved on. Hoarding what only one person knows feels like job security, but it is ultimately a ceiling. The people thriving in this model build the skill, hand over their knowledge, and move to the next problem. Each handoff allows them to take on a bigger, more interesting challenge.
This is the model companies should build toward as AI moves deeper into knowledge work—rewarding people for making their expertise reusable. That judgment is intellectual property, and paying for it is how you build AI that works while retaining top talent.
