Appier Group has moved past asking whether employees use AI. It is now measuring how much AI actually delivers — and linking the answer directly to pay and performance targets. CEO Chih-Han Yu disclosed on August 13 that the company has formalized AI output in its OKR framework, making it roughly half of every employee's evaluation. Per-employee quarterly gross profit hit a record 10.25 million yen in the April-June quarter, up 38% year-on-year, the figure Yu is using to demonstrate the model works.
AI Outcomes Now Worth Half Of Employee Performance Scores
Appier tracks AI tool usage across workflows through internal systems — measuring not how many prompts are submitted, but what those tools actually produce: features built, model improvements made, concrete output delivered. The company has deliberately moved away from token-level accounting for individual engineers. The question it asks is simpler: did AI increase actual output?
That logic is now embedded in the OKR framework. Yu told analysts at the earnings briefing that performance assessments reflect a roughly 50-50 split — half tied to AI-driven outcomes, half to core business responsibilities. The policy covers the entire company, from engineers to operations staff. Yu said it applies to him personally as well.
"We want employees not just to use AI tools," he said, "but to actually convert AI into measurable productivity."
Yu clarified that the per-employee productivity metric is calculated as total company gross profit divided by total headcount — not restricted to engineers or the R&D function. Q2 per-employee gross profit of 10.25 million yen compares with roughly 7 million yen a year earlier.
Agentic Engineering Pushes Gross Margin Above 60% For The First Time
The most direct application of this framework is inside R&D. Appier has deployed what it calls Agentic Engineering — AI agents that serve as development assistants, helping engineers build features, run model tests and refine algorithms. The goal is to compress cycles that previously required more manual effort.
The results appeared in the margin line. Q2 gross margin reached 60.1%, crossing 60% for the first time in company history, rising to 61.3% on a constant-currency basis. Core business gross profit grew 43.5%, well ahead of overall gross profit growth of 33.5%. Yu acknowledged that deploying more agentic AI in R&D does raise compute and development costs, but said improvements in model accuracy and client return on investment more than offset the increase.
Appier describes the mechanism as an "AI Flywheel": faster R&D produces better models, which lift client ROI, which prompts clients to expand their spending on the platform — feeding more data and resources back into the next round of AI development, and around again.
AI Shifts Appier's Calculus On Building Versus Buying
The impact of agentic AI extends beyond internal operations to how Appier thinks about growth through acquisition. Yu said the equation is shifting: where software companies once needed to buy capabilities they lacked, internal teams can now build some of those capabilities faster with AI assistance. "You don't necessarily have to merge with another company," he said.
Appier has historically acquired products rather than revenue, and Yu said any future M&A would still prioritize targets that integrate tightly with the core business and generate product synergies rather than simply adding top-line scale.
The company outlined four capital allocation priorities going forward: vertical AI and Agentic engineering R&D; expansion in key markets and enterprise accounts; selective M&A with strong core-business fit; and progressively higher shareholder returns through dividends or buybacks as core free cash flow improves.
Yu Sees Each Employee Eventually Supervising A Fleet Of AI Agents
Yu offered a longer-term framing for where this trajectory leads. The next stage of value from AI agents, he argued, is not about replacing workers but about multiplying what each one can accomplish. He described a model in which employees act as supervisors — assigning development and execution work to AI agents, which can keep running through off-hours when human staff are unavailable.
"Humans need to sleep," he said. "AI doesn't."
In this model, the fundamental unit of production at a software company shifts from "one employee" to "one employee plus a team of AI agents." Appier's 38% year-on-year gain in per-employee quarterly gross profit is, in Yu's framing, the first financial data point in support of that argument. （Related： Appier Delivers Record-High Revenue, Profitability and Core Free Cash Flow ｜ Latest ）
