Reid Hoffman made a point recently that stuck with me. Most enterprises approach AI adoption like procurement: choose a vendor, buy licences, launch a pilot, collect usage reports, tick the strategy box. His argument (and I think he's right) is that the entire framing is backwards.
Most enterprise AI strategies were written for a world where AI generated text and humans decided what to do with it.
That world was three years ago.
Since then, AI got tools, persistent memory, and the ability to plan, act across systems, and run for hours without supervision.
The strategy didn't update, but the capability did.
Two types of AI are now happening in enterprise. Both are necessary. But most organisations are only running one, and treating it as the entire strategy.
Productivity AI makes individuals faster: copilots, writing assistants, slide generators. Harvard Business Review surfaced the problem recently: when individuals work 2x faster, expectations become 3x higher. The worker accelerates, but the system stays the same.
Productivity AI is necessary. The problem is that most enterprises treat it as the entire AI strategy.
Engineered AI changes the system itself. Agents embedded into workflows, decision engines, AI that redesigns how work moves through the organisation with guardrails and human oversight from day one. Processes get re-engineered, coordination costs collapse, and the P&L actually shifts.
The difference is the difference between giving everyone a calculator and redesigning the spreadsheet.
One makes the individual faster, the other eliminates the bottleneck entirely.
Until recently, Engineered AI was the domain of technical teams. That has since changed. Tools like Claude Code and Claude Co-work have made those capabilities accessible to marketing leads, operations managers, and finance teams, not just developers. The barrier between "use AI" and "build with AI" has collapsed, and that collapse is what makes the Two Lanes distinction urgent.
The evidence is already visible. Hoffman demonstrated the shift on his Possible podcast: a coding agent pointed at raw CSVs produced a full dashboard and McKinsey-style presentation in minutes. The feedback loop collapsed. Bob Sternfels (McKinsey) noted they saved 1.5 million hours through AI last year by "dividending" that time back into solving harder problems.
The reason most enterprises never get to Lane 2 is that they're bleeding time in places they've stopped noticing. Every enterprise pays three invisible taxes, and most don't measure any of them.
Search Tax: six people stop working to find the latest pricing deck. Thirty minutes later, someone finds it in an email from three weeks ago. Thirty minutes multiplied by six people, for one file.
Translation Tax: sales says "hot lead", marketing says "MQL", ops says "needs verification", finance says "show me the contract". Same opportunity, four interpretations, three days of back and forth.
Coordination Tax: the two-hour weekly planning meeting where the first 90 minutes is "what did everyone do last week?" Eight people in a room when two people in Slack would have decided in five minutes.
These three taxes sit in the layer where language models create the most immediate value. Lane 2 starts here, not with a transformation programme, but by turning the institutional memory buried in someone's head or a Slack thread into something structured, retrievable, and actionable.
Four conversations with NZ enterprise leaders in the past month showed exactly where the lane divide sits. One is treating AI like a competitive weapon, with a frontier team accessing Claude Code and Cursor, actively hunting edge.
The other three shut the conversation down. "We rolled out GitHub Copilot. That's agent-native engineering". "AI licences for the business are too pricey". "Locked to a homegrown GPT, and that's it".
All three are only running Lane 1, and all three think they're covered.
The cost of running one lane without the other compounds every quarter. The three taxes don't sit still: they multiply as teams grow, as projects layer, as coordination complexity increases. An organisation paying a 15% coordination tax today will pay 20% next year with the same headcount, because the volume of decisions, handoffs, and context-switching only moves in one direction. Meanwhile, Lane 2 organisations are compressing those same costs by 30-50%, which means the competitive gap between Lane 1 and Lane 2 widens every quarter without a single new hire on either side.
The talent problem accelerates this: the Head of Strategy I spoke with two weeks ago understood all of this. His employer's response was to deploy Copilot and walk away.
He's now starting at a competitor in April.
The people who understand the difference between the two lanes are exactly the people who will leave organisations that don't enable them. That's not a retention risk on a spreadsheet: it's the loss of the person who knew what to do, walking out the door with the playbook in their head.
Adding Lane 2 doesn't require ripping out Microsoft and starting over. It requires a Frontier Team: a small, cross-functional group of 5-8 people given access to frontier tools inside a governance framework.
Microsoft's own research calls the top 5% of companies by AI adoption "Frontier Firms", where 71% say their company is thriving vs 37% globally (Microsoft Work Trend Index, 2025). Not because they deployed a tool, but because they embedded AI into how work gets done across functions.
That's Lane 2.
The Frontier Team playbook for the first 90 days looks like this. Pick one of the three taxes as the target. Staff the team with people from the functions that feel the tax most: if it's the coordination tax, that's ops, project management, and one senior leader who owns the meeting cadence.
Give them access to frontier tools (not just Copilot) with clear governance boundaries. Set one metric: hours reclaimed per week across the affected workflow. Report back at 90 days with the number and the workflow redesign.
The organisations running Lane 2 will compound the gains.
Everyone else is paying three taxes they haven't measured, running a strategy written for a different era, while the people who understood the difference start somewhere else next quarter.
Passionate about all things AI, emerging tech and start-ups, Mike is the Founder of The AI Corner.
Subscribe to The AI Corner
