A summary of my article in Innovation & Tech Today.
In the first half of 2026, we learned that while AI is extraordinarily effective, it can also get pricey fast. Our budget ballooned, as did just about everyone else’s.
Rather than impose hard caps or red tape, we tried to make AI spending more self-regulating through a combination of individual autonomy, accountability, and a value-based investment framework.
The question I ask is: which of four jobs are we hiring AI to do? Drive the business, run the business, build the business, or discover what’s next?
Each warrants a different definition of success, investment rationale, and decision-maker.
Drive the business
AI that directly supports customers or revenue has the most obvious business case, measured through outcomes like conversions, revenue growth, and contribution margin. When the incremental economics are compelling, key deciders unlock additional spending quickly.
Run the business
AI can automate internal work in recruiting, analytics, advertising, content, and wherever repetitive tasks abound. The measure is simple: how much time or money are we saving per activity? Or, better yet, how much incremental value can that freed capacity unlock?
Budgeting for these tasks starts with department leaders, with Finance helping validate the economics.
Build the business
Coding is among the fastest-growing and perhaps largest areas of AI spending, but its impact is harder to measure.
Lines of code, pull requests, and feature counts measure engineering output. But does shipping more software faster actually move the business forward?
I am on the hook for demonstrating how that translates into company KPIs. I use executive peers’ skepticism as feedback to refine my narrative before high-stakes meetings that unlock this growing budget.
Discover what’s next
We carve out a dedicated budget and time for experimentation, giving employees the freedom to try new models, tools, and techniques.
Every quarter for the past 13 years at Pearl, we have held a two-day hackathon expressly for this purpose. Standout projects are recognized quarterly by our executive leadership team, celebrating technical improvements and customer experience innovations. Thirteen years in, we have so many ideas that curating them is its own problem (a good one to have).
The economics here are different. We are buying learning and future options. Many of today’s big revenue movers trace their origin to these early small bets.
Across all four, putting each type of AI spending in its place across our portfolio, along with its own definition of success, has made it easier for everyone to see the return on our AI budget. That has directly translated to better organic budget management. Our AI budget grew 3x so far this year, but the impact grew even more, so we never had to rein it in. AI usage keeps changing, and it may yet challenge this framework.