BCG: Token-Based Competition Has Arrived — Intelligence Is Now Abundant, Advantage Goes to Those Who Apply It Best
BCG today declared that companies have entered a new era of competition defined by token consumption, not model access.
The June 23, 2026 article “The Era of Token-Based Competition Is Here. Is Your AI Strategy Ready?” argues that AI models now deliver PhD-level intelligence on nearly every topic. When intelligence is abundant, winning no longer comes from having the smartest models. It comes from applying that intelligence to business problems more productively than competitors through the consumption of tokens — the measurable unit of AI work.
BCG explicitly links the concept to its 1988 “time-based competition” framework. Just as speed once became a new basis for advantage, token productivity is now the differentiator.
Key Data from BCG Analysis
- BCG examined token consumption across 107 public technology companies with more than $500 million in trailing twelve-month revenue.
- In software engineering workflows, companies in the highest token-use quintile posted 16.5% median year-over-year revenue growth, compared with just 5.1% for the lowest-usage group.
- Broader finding: 50% to 55% of jobs will be reshaped by AI, while only 10% to 15% are likely to be displaced.
- Fewer than 10% of employees currently use AI in an agentic way (delegating multi-step work to agents rather than simple prompting).
- Token spend is already creating measurable divergence in performance among tech firms.
The Four Critical Actions BCG Recommends
- Manage tokens like capital investment. Track ROInt (Return on Intelligence) — value created divided by combined labor + token costs — instead of narrow labor savings or raw token volume. Companies that only chase cost cuts or “tokenmaxxing” will leave value on the table or create perverse incentives.
- Move token accountability out of IT. Shift ownership of token strategy and spending to AI Centers of Excellence, business units, strategic planning, or finance so it sits at the core of how work gets done rather than as a cost-center afterthought.
- Treat tokens as talent enhancers, not pure labor substitutes. Over-cutting headcount risks losing the human judgment needed to convert AI output into trusted results. BCG notes Gartner’s prediction that half of companies cutting customer service staff due to AI will rehire by 2027.
- Make organizational change the enabler. Most adoption stalls for human reasons — lack of workflow redesign, unclear value, or resistance. Leaders must address culture and behavior change directly. Companies like JPMorgan Chase (with its phased Coach AI rollout) and Cloudflare (reducing “measuring” roles while accelerating engineering and customer-facing hiring) show the pattern.
Why This Matters Now
Tokens have become the practical currency of AI advantage. The firms redesigning entire workflows around AI — combining automation with human oversight — create compounding flywheels: systems that get smarter, faster, and cheaper over time while scaling capacity up or down with demand. Companies still treating AI as a bolt-on tool or pure cost play will fall behind on both performance and unit economics.
Actions to Take
- Enterprise leaders: Immediately map token consumption by workflow and calculate ROInt on current AI initiatives. Reallocate spend toward high-return, compounding use cases rather than broad experimentation.
- Founders building AI products or internal tools: Prioritize workflow redesign and human-AI team structures now. The revenue growth gap between high and low token users is already visible in software engineering.
- VCs & investors: Add token productivity and ROInt tracking to diligence questions. Portfolio companies that treat tokens as strategic capital and redesign work around them are pulling ahead on growth metrics.
- Talent strategy: The winners will need more people who can direct, evaluate, and improve AI systems — not fewer. Plan for role evolution rather than simple headcount reduction.
This report aligns directly with today’s McKinsey piece on data readiness as a scaling constraint and Gartner’s AI cloud market forecast. All three point to the same reality: raw compute and model access are no longer sufficient. Execution on data foundations, token economics, and organizational redesign is becoming the real moat.
