Axon’s 2,200-Person AI Mandate Is the Management Test Most CEOs Dodge
Most CEOs say they want AI productivity. Axon has put a 10% target on 2,200 people and made the bluff measurable.
Most CEOs talking about AI are doing corporate theatre. They’ve bought licences, held a town hall, appointed somebody to “lead transformation” and called it progress.
Axon has done the harder thing: it has put a number on the table and made people accountable for hitting it.
On October 14, Axon chief product and technology officer Jeff Kunins said the company has set a mandatory 10% productivity-gain target for every team in its roughly 2,200-person research-and-development organisation. Teams that push beyond 15% can earn stock grants, provided the broader team clears that 10% baseline. By the end of 2026, Axon expects AI to compress the work required across product management, design and software engineering by 25% to 30%.
That is not another “we’re exploring AI” press release. It is a management decision.
And it exposes the question every founder, CEO and senior operator now needs to answer: are you actually redesigning work, or merely giving people a faster way to produce the same amount of nonsense?
Axon has made AI a performance issue
Axon is best known for Tasers, but it is no longer simply a hardware company. Founded by Rick Smith in 1993 as Taser International, the business now sells body cameras, drones, licence-plate-reader cameras, software and AI products to law enforcement, emergency services and private-sector security customers. It has sold more than one million Taser devices and has reported ten consecutive quarters of revenue growth above 30%.
Its AI push is happening inside a business where mistakes are not trivial. Axon’s tools can help turn body-camera audio into first drafts of police reports, translate more than 50 languages and dialects in the field, and redact sensitive information from footage. The company has also faced criticism over potential errors in AI-generated reports and broader concern about surveillance technology. Earlier this month, Los Angeles City Council delayed a pilot involving Axon’s AI-enabled licence-plate readers on streetlights.
That backdrop matters.
If you run a consumer app and your AI writes a dodgy product description, somebody gets annoyed. If you run a public-safety technology company and your AI produces biased language or gives police bad information, the consequences can be brutal. Axon has said its internal ethics council reviews product development, and Kunins acknowledged that early versions of its Draft One report-writing tool showed bias in language used for white and Black subjects. The company says it fine-tuned the models to make the outputs more consistent.
So Axon is not mandating AI use in a consequence-free playground. It is doing it where accuracy, oversight and trust are part of the product.
That makes the internal target more interesting, not less. Management is saying: we need to move much faster, but we cannot pretend quality and accountability no longer matter.
The 10% target is the point — not the 25% dream
The headline number is the hoped-for 25% to 30% compression in effort. Nice. Every executive loves a big shiny target.
The more important number is 10%.
Why? Because that is the number tied to operating behaviour now. It forces every team to stop treating AI as a side project for the enthusiastic nerds in the corner. Product managers, designers and engineers have to find work that can be done with fewer resources, then demonstrate the gain.
This is what proper leadership looks like: a clear baseline, a deadline, an incentive and consequences for doing nothing.
Too many companies are making a basic mistake. They measure adoption — seats purchased, prompts written, training modules completed — instead of output. That is like celebrating that everyone owns a hammer while the house is still half built.
Axon’s approach is not perfect just because it is numerical. A productivity target can easily become a joke if managers count quicker drafts as finished work, dump verification onto already overloaded staff, or quietly lower the quality bar. Anyone who has lived through a corporate efficiency programme knows the tricks.
But at least Axon has chosen the right battlefield: work design. Not vibes. Not usage statistics. Not a CEO LinkedIn post written by an intern and polished by ChatGPT.
Middle managers will either make this work or quietly kill it
Here is the bit the AI evangelists skip: the burden of this transition lands on middle management.
A recent Wharton Human-AI Research and GBK Collective survey found that 62% of managers and above use AI daily, up from 46% a year earlier. Yet the share who described themselves as excited fell eight points to 51%. Usage is climbing; enthusiasm is cooling. That is what happens when a tool moves from demo day to real work.
Managers are no longer just creating work. Seventy-two percent said their focus has shifted toward managing AI-generated output. In plain English: someone still has to check the machine’s homework.
That creates a nasty trap. A company tells employees to use AI, output increases, and managers become the human quality-control department for a torrent of average material. The business has not saved time. It has merely moved the work uphill and made the bottleneck more expensive.
The same survey found that teams reporting strongly positive AI returns were much more likely to measure growth outcomes — things like new revenue or sales conversion — rather than efficiency alone. They were also more likely to have changed management structures and workflows.
That is the real lesson for Axon and everyone copying Axon: you do not get the prize for making existing tasks 10% faster. You get the prize for deciding which tasks should disappear, which decisions should move closer to customers, and which people should be freed up to build something that earns money.
The overlooked risk: rewarding speed can punish judgement
I like that Axon is using stock grants. If people create durable value, let them share in it. That is far better than demanding more output and banking the upside entirely at head office.
But the incentive needs a second leg: quality.
If your reward system only pays for speed, don’t act surprised when people optimise for speed. You will get rushed customer replies, flimsy code, incomplete analysis and managers approving work they have not properly read because the queue is now too big.
For Axon, where its technology intersects with policing and surveillance, that risk is obvious. But it exists in every business. A finance team can close faster and produce worse decisions. A sales team can send more outreach and damage the brand. A software team can ship faster and load the company with technical debt that becomes next year’s fire.
The contrarian view is that most businesses should not start with an AI target across the whole company. They should start with three ugly, repetitive, measurable workflows where they already know the cost of delay, error or rework.
Then they should apply an Axon-style standard: a baseline, an owner, a target, a time frame, an audit trail and a quality metric.
If the workflow improves, expand it. If it creates more checking than it saves, kill it. No sacred cows, no endless pilot.
What this means for you
If you are a founder or operator, do this next week.
First, pick one workflow that consumes serious time every week. Not “use AI in marketing.” Be specific: preparing sales proposals, reconciling invoices, triaging customer tickets, drafting investment memos, turning calls into CRM notes, or reviewing contracts.
Second, measure the current state for two weeks. Track elapsed time, labour hours, error rate, rework and commercial outcome. If you cannot measure the before, your claimed AI gain will be fiction.
Third, set a modest hard target. Ten percent is a sensible place to start. Make one person accountable for the result, but do not make them accountable only for speed. Tie the target to a quality measure and, where possible, a growth measure.
Fourth, decide who verifies the output before the AI starts producing it at scale. Do not dump that job on a manager without removing something else from their plate. Verification is work. Budget for it.
Finally, share the upside. Axon’s use of stock grants is the right instinct. If AI lets a small team create more value, the people who changed the operating model should benefit. Otherwise employees will rightly conclude that “productivity” is just code for doing more with less until someone else gets a bonus.
AI will not reward the companies with the prettiest strategy decks. It will reward the ones willing to make decisions, measure results and change the way work actually gets done.
Axon has put a 10% number on the board. Most CEOs are still choosing fonts for the presentation.