ExpertsFMEP 145: Rick Smith (CEO, Axon): The Wild Ride From Near-Bankruptcy to $50B+
The Logan Bartlett Show · May 30, 2025
5 Selected Lessons:
heuristic · 2 case · 1 priority · 1 reasoning_pattern · 1
Choosing which legacy systems to displace based on retraining cost and time-pressure sensitivity
Source
The Logan Bartlett Show — EP 145: Rick Smith (CEO, Axon): The Wild Ride From Near-Bankruptcy to $50B+ (May 30, 2025) · at 56:47 ▸ · at 59:25 ▸
When this applies
when deciding which entrenched workflow/system to attempt to displace with a new product
The takeaway
Rank candidate displacement targets by (a) how much users already dislike the status quo and (b) how costly/risky retraining would be under operational pressure — pursue low-retraining-cost, high-dissatisfaction targets, abandon high-cost ones.
Why it works
Systems used under high time pressure by highly trained operators (dispatch, every keystroke matters in a crisis) have a prohibitively high retraining cost that outweighs incremental efficiency gains, whereas systems used without extreme time pressure, where users already dislike the legacy tool, have low resistance to replacement.
One level up
The size of the opportunity matters less than the switching cost borne by the entrenched user under their specific operating constraints.
Where it stops holding
This stops working when a segment's resistance to retraining can be offset by sufficiently large efficiency gains or by a greenfield customer with no legacy habits to unlearn.
Detail
hierarchy:
- 1. Customer already dissatisfied with the legacy tool (records: 'our systems kind of suck already') — pursue
- 2. High-pressure, highly trained workflows where retraining cost is prohibitive (dispatch: 'every keystroke matters... there's a ton of time pressure') — abandon or avoid
- 3. When a smaller competitor (FUSIS) proves faster adoption on an adjacent sub-problem (the map, not the console), acquire rather than keep building in-house
rationale: Axon spent tens of millions on CAD and shut it down after customers refused to retrain, akin to the QWERTY keyboard problem, but succeeded in displacing legacy records management because officers welcomed AI-driven improvement.
In their words
Dispatchers did not invite us to innovate. And so that was a case where we looked at it and said, you know what? Let's shut it down.
Making high-risk bets early because technology diffusion moves faster than people expect
Source
The Logan Bartlett Show — EP 145: Rick Smith (CEO, Axon): The Wild Ride From Near-Bankruptcy to $50B+ (May 30, 2025) · at 54:04 ▸
When this applies
when deciding whether to invest in a capability that isn't yet technically feasible but trend lines suggest will become feasible
The takeaway
If you believe a capability is on an exponential improvement curve, make the seemingly-too-early bet now, because being ready when the curve crosses the feasibility threshold is the advantage.
Why it works
If technological capability truly advances exponentially, the gap between 'impossible today' and 'feasible' shrinks faster than linear intuition predicts; starting the foundational work early (even while it looks like 'smoke and mirrors') means being ready the moment the enabling tech (e.g., GPT-4) arrives.
One level up
Position ahead of inflection points you can't precisely time, because the cost of being early is lower than the cost of being late when the curve is exponential.
Where it stops holding
This stops working when the predicted capability curve is not actually exponential, or when the organization can't sustain the multi-year cost of being 'early' without near-term payoff.
Detail
approach: Attend futurist/technology trend sources, form a belief about where exponential tech is headed over ~10 years, and start building the foundational product even while customers view it as unrealistic.
signals:
- Customers/market treat it as 'complete smoke and mirrors'
- A directional belief grounded in an exponential-technology thesis (Kurzweil)
- A clear enabling-technology milestone (e.g., GPT-4) that, once it arrives, unlocks the previously-impossible feature
rationale: Axon began building toward AI police-report extraction ~10 years before GPT-4 made it possible in 2024, and Smith says they 'turned out to be approximately right' about the timeline.
In their words
If you are thoughtful about that, you'll make bets while they're still high risk. You'll make the riskier bets, but the advantages you'll be there when the tech shift happens.
Major tech/behavior shifts take about seven years to sink in
Source
The Logan Bartlett Show — EP 145: Rick Smith (CEO, Axon): The Wild Ride From Near-Bankruptcy to $50B+ (May 30, 2025) · at 4:04 ▸ · at 54:04 ▸
When this applies
when introducing a genuinely new technology or behavior change to a conservative, safety-critical customer base
The takeaway
Expect multi-year latency between a technology's technical readiness and its market adoption inflection, and plan funding/patience accordingly.
Why it works
Adoption of unfamiliar technology by risk-averse institutional buyers requires a slow build of trust, infrastructure, and a triggering event before it tips into mainstream use.
One level up
Timing asymmetry between invention and adoption is itself a strategic resource — being early and surviving long enough is the edge.
Where it stops holding
This stops working when the market has no structural resistance (regulatory, cultural, or trust-based) to the new technology, in which case adoption can be much faster.
Detail
rule: Budget roughly seven years for a major new technology to go from introduction to real traction in a given industry.
threshold: ~7 years
rationale: Both the original TASER and body cameras each took about seven years to 'really take off' per Smith's direct experience.
In their words
It seems to me like major tech changes in this industry seem to take about seven years, to sink in.
Focus on the problem, not the monetization, and the business model follows
Source
The Logan Bartlett Show — EP 145: Rick Smith (CEO, Axon): The Wild Ride From Near-Bankruptcy to $50B+ (May 30, 2025) · at 20:57 ▸ · at 37:14 ▸
When this applies
when deciding what to build next in a company with multiple possible directions
The takeaway
Sequence product strategy as: identify a valuable unsolved problem → solve it → then design the business model around the solution, not the reverse.
Why it works
Starting from 'what valuable problem can we solve' orients the team toward genuine customer value; a monetization model is comparatively easy to design once real value is proven, whereas starting from monetization risks chasing superficial revenue without real value.
One level up
Value creation precedes value capture in the order of strategic thinking, even though capture is necessary eventually.
Where it stops holding
This stops working when the organization has no slack/capital to survive the period before monetization is figured out — in genuinely resource-constrained situations, the revenue model may need to be decided earlier.
Detail
rule: Don't start by asking how to make money; start by asking what valuable problem to solve, then build a financial model around the solution.
threshold: n/a
rationale: Every time Axon did this (body cameras, AI report writing), it created a huge business; solving valuable problems get-s you into 'the right headspace.'
In their words
I actually don't spend any time really thinking about how we can go make money. What I focus on is what are interesting, valuable problems we can go solve?
Simplifying the TASER to one-dart-per-trigger despite expert objection
Source
The Logan Bartlett Show — EP 145: Rick Smith (CEO, Axon): The Wild Ride From Near-Bankruptcy to $50B+ (May 30, 2025) · at 31:22 ▸
When this applies
when every expert/customer intellectually rejects a proposed design simplification, but the team hasn't actually tested it experientially
The takeaway
When intellectual/expert consensus says a radical simplification 'can't work,' build a prototype and let people experience it before trusting the a priori reasoning — direct experiential feedback can overturn seemingly obvious expert judgment.
Why it works
Experts reasoned from the old mental model (both darts needed simultaneously) without experiencing the new interaction pattern directly; actual hands-on use revealed the simplification removed a hard cognitive task (trigonometric aiming) that the old model had normalized as necessary complexity.
One level up
Deeply internalized complexity (treated as essential) is often removable, but only experiential testing — not consultation — will reveal that.
Where it stops holding
This stops working when there's no safe/cheap way to prototype and test the change experientially, or when the complexity really is load-bearing and removing it causes failure, not just when experts predict it wrong.
Detail
situation: Engineers debated how to use 10 available dart slots; 'every police expert' insisted two darts per trigger pull were necessary and that firing one was a 'terrible idea.'
decision: The team built a prototype firing one dart per trigger pull and had people physically test it rather than relying on expert opinion.
outcome: Users immediately found it far easier to aim and control, eliminating the need for 'angle trigonometry,' and the simplified design became the new standard, extending effective range from 0-45 feet.
lesson: Non-obvious design wins can be invisible to experts reasoning abstractly, but obvious once experienced directly.
In their words
Only when they experienced firing it did we all have this realization, my gosh, we should have simplified this thing years ago.