Synthesia advises deprioritizing token-usage metrics
Business Insider reports that Laura Gonzalez, head of people at Synthesia, argued against using token consumption as a primary AI productivity metric, saying 'It's not the metric for success.' Gonzalez told Business Insider that measuring token use risks rewarding activity over outcomes - analogous to judging a salesperson by call volume instead of deals closed - and recommended focusing on business impact instead. The piece joins broader industry coverage on 'tokenmaxxing,' a trend in which companies rank employees by AI token usage, which SHRM and CIO Magazine have also flagged as a misaligned adoption metric. Business Insider notes Synthesia, the London-based AI video company, expects all staff to use AI tools where relevant.
What happened
Business Insider published an interview with Laura Gonzalez, head of people at Synthesia, reporting that Gonzalez argued against token consumption as a measure of AI productivity. Business Insider quotes Gonzalez: "It's not the metric for success," and records her analogy that measuring token use would be like evaluating a salesperson by call volume rather than deals closed. Business Insider also quotes Gonzalez: "We expect absolutely everyone to leverage and use AI when necessary internally." Synthesia is a London-based company developing AI-generated video tools.
Industry context
The article enters a growing debate about "tokenmaxxing" - the practice of tracking employee AI token usage as a proxy for productivity, sometimes via internal leaderboards. SHRM reported in 2026 that tokenmaxxing offers HR a cautionary lesson about metrics. CIO Magazine noted that token counts are a reasonable adoption proxy but a poor productivity measure. Axios reported Salesforce is also pushing back on tokenmaxxing. Fortune published analysis in May 2026 that tokenmaxxing has not delivered the ROI companies expected.
Why it matters for practitioners
Measuring raw usage metrics (token counts, API calls, inference volume) skews incentives toward activity rather than outcomes. Practitioners selecting AI adoption KPIs should consider downstream business impact - time-to-ship, defect rates, customer satisfaction - alongside cost and attribution complexity. Gonzalez's framing aligns with the emerging practitioner consensus that burn-rate dashboards are diagnostic tools, not scorecards.
Scoring Rationale
A single-source Business Insider interview with one company's HR lead providing practical but opinion-level guidance on a real trend. The tokenmaxxing debate has genuine practitioner relevance and multiple outlets have covered it, but this specific item adds a niche data point rather than a strategic shift or new research. Score reflects minor-but-relevant practitioner advisory, not a technical breakthrough.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
