Is it acceptable to treat employees' AI usage spend as part of their measured cost?

Adam Foroughi · Apr 27, 2026

Token budgets and token-usage leaderboards are flawed and will drive companies to burn money on worthless output

If you hand people a budget and rank them on consumption they'll generate a bunch of crap with no revenue on the other side; companies should instead measure which token consumption aligns with their actual KPIs, at which point they'll want to invest in tokens rather than budget them

65:46 20VC: Applovin: $160BN Market Cap, $5.48BN Revenue, $10M EBITDA Per Head | Why the Best Do Not Need Mentorship | Why Founders Should Not Angel Invest | Why Kindness in Business Will Slow You Down with Adam Foroughi

Andrew Macdonald · Aug 17, 2026

Cost leaderboards are a bad target on their own — neither extreme is right because what matters is value created — but adoption leaderboards for the specific AI tool relevant to someone's role are genuinely useful

Blunt metrics get gamed, with people optimizing for the metric rather than outcomes; but every customer support agent should be using the AI assistant built for them, and non-adoption is worth questioning

Scope: adoption leaderboard should be domain-specific, not universal (not every employee needs a coding tool)

41:08 20VC: Uber President on The Untold Uber Stories: Travis, China and Self-Driving | Why Autonomy Is Existential | How to Beat DoorDash to #1 in Food with Andrew MacDonald

Aaron Levie · Apr 20, 2026

Token budgets should be allocated to wherever the most value is generated for the company, so aggressive 'token maxing' makes sense in software but is not obviously right in every industry

In software the value proposition correlates to how much software you can produce, so maximizing tokens and using leaderboards drives shipping and spreads best practices faster; that correlation doesn't hold everywhere

23:53 Token maxing makes sense only where value scales with output

Nikesh Arora · Jun 22, 2026

Their token policy is 'use judiciously and keep track' rather than a free-for-all: people using tokens well are not constrained

20:21 Track usage but do not constrain productive users

Alex Atallah · Aug 10, 2026

Making employee cost dynamic based on model usage is not demeaning because employees control how much they cost

Employees choose which tools and models to use, so they can influence their own cost and assess their own efficiency

55:54 Acceptable because employees control their own spend

Matan Grinberg · Jun 13, 2026

Companies should proactively set conscious token budgets and per-team limits rather than let usage run unbounded and react in shock, and orgs are heading toward highly nuanced per-team resource allocation

He has seen dozens of customers blow through usage without deciding which parts of the codebase deserved the tokens, then panic-install limits; awareness on the way up beats a sudden correction, and different teams warrant different limits

20:18 Set conscious per team token budgets in advance

Andrew Macdonald · Aug 17, 2026

Making AI usage and cost visible to employees — ideally a live cost counter in the tool itself — makes users more cost-conscious

Seeing the running equivalent cost of what you're doing as it scales changes user behaviour, like watching the total at a grocery store

40:46 Live cost visibility changes usage behavior