ExpertsFMRebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch
Training Data · May 19, 2026
5 Selected Lessons:
heuristic · 2 reasoning_pattern · 2 case · 1
Why infinite-resource incumbents often won't enter a complex niche
Source
Training Data — Rebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch (May 19, 2026) · at 15:37 ▸
When this applies
when assessing threat of a much larger, better-resourced player entering your specific niche
The takeaway
Compare the incumbent's expected return-on-focus in your niche versus their other opportunities; if the ratio is unfavorable for them, the real threat is a simplified version, not full mastery of your complexity.
Why it works
A resource-rich player's opportunity cost of focus is the deciding factor, not capability: if a niche would take them years to master relative to the revenue available elsewhere in months, rational prioritization keeps them out of the complex version of that niche, even though they could build a simple version.
One level up
Competitive threat scales with the competitor's opportunity cost, not merely their capability.
Where it stops holding
This stops working when the large player decides your category IS strategically central to their core platform (e.g., becomes a land-grab priority despite opportunity cost), or when your niche is simple enough to master quickly.
Detail
approach: Weigh the complexity/time cost of mastering the niche against the revenue the incumbent could generate elsewhere in the same time.
signals:
- Niche requires years of focus to master
- Incumbent's other product lines generate comparable revenue in months
- History shows smaller companies (OpenAI, Anthropic themselves) beating entrenched incumbents with more resources
rationale: Anthropic added more ARR in recent months than ServiceNow added in 20 years, so devoting top talent to ITSM would be a poor use of their resources, though a simplified mid-market version is plausible.
In their words
I think in the past couple months, Anthropic has added more ARR than ServiceNow has in the past 20 years. Good point. And so, does it really make sense for them to take their best and brightest people to throw them at this problem
Individuals want autonomy, organizations want control — the AI adoption tension
Source
Training Data — Rebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch (May 19, 2026) · at 29:42 ▸
When this applies
when a new powerful tool creates a gap between what individual users want (more freedom) and what the organization overseeing them wants (more control)
The takeaway
When a powerful new tool creates divergent incentives between individual users and the organizations governing them, expect bottom-up shadow adoption to eventually force top-down policy change — bet on serving the signal rather than fighting it.
Why it works
Individual productivity incentives push toward maximal agent autonomy and access, while organizational risk incentives push toward restriction; this creates a structural tension that mirrors the consumer-vs-enterprise product design divide, and it typically resolves via 'shadow' adoption that eventually forces official policy change.
One level up
Tensions between individual incentive and institutional risk tolerance are a recurring structural pattern across technology adoption waves, not unique to any one tool.
Where it stops holding
This stops working when the downside risk of ungoverned individual adoption is catastrophic enough (e.g., regulatory, safety) that no amount of productivity pressure should override institutional control.
Detail
approach: Recognize the autonomy-vs-control tension as structurally similar to past technology waves (BYOD) and treat individual adoption signals as a leading indicator of what the organization will eventually need to embrace.
signals:
- Individuals adopting tools organically despite IT restriction
- Security teams defaulting to 'protect, protect, protect'
- Eventual realization that blocking the tool costs more than enabling it with guardrails
rationale: Companies that default to 'yes' will get ahead of competitors, accepting real security-incident risk as the cost of leading; companies defaulting to 'no' will fall behind.
In their words
And I think the companies that basically say yes as the default are going to be way ahead of the companies that say no as a default. And we're seeing this with the companies that just embrace it.
Automation must be easier than the manual workaround
Source
Training Data — Rebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch (May 19, 2026) · at 3:26 ▸
When this applies
when trying to get people inside an organization to adopt a new process instead of defaulting to the familiar manual shortcut
The takeaway
A new tool only gets adopted in-the-moment decisions if it is strictly easier than the existing workaround, not just more valuable in theory.
Why it works
People choose the lowest-friction path in the moment; if building the automation requires more steps than doing the task by hand, they will always do the task by hand, so adoption of automation is gated by relative ease, not by value.
One level up
Design for the moment of decision, not the ideal outcome — friction beats value in determining real-world behavior.
Where it stops holding
This stops working when the automation, once built, is reused so rarely that the upfront cost of building it never pays off even if building is easy.
Detail
rule: Make building the automation at least as easy as doing the manual task, or people will always choose manual.
threshold: Two-click manual reset vs multi-step workflow builder — if manual is fewer steps, manual wins every time.
rationale: Adoption decisions happen in the moment under time pressure, so relative friction dominates the choice.
In their words
And I think that it comes down to that. People in the moment have to make that decision. And you want that decision to be very, very easy to opt for the automation, not opt for the manual step.
Customer insight as the durable moat vs. copyable product features
Source
Training Data — Rebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch (May 19, 2026) · at 5:43 ▸
When this applies
when competing in a market where product features can be rapidly replicated by well-resourced competitors
The takeaway
When execution speed becomes commoditized across an industry, the remaining durable advantage is depth of customer understanding, not the artifact itself.
Why it works
As AI accelerates implementation speed for everyone, the thing that's hard to copy shifts from 'can you build it' to 'do you know exactly what to build' — which requires sustained, deep, first-hand exposure to customer problems.
One level up
Identify which layer of your value chain is becoming cheap and shift your moat-building effort to the layer that remains scarce.
Where it stops holding
This stops working when customer needs are simple/static enough that deep ongoing immersion yields no additional insight beyond a basic survey.
Detail
rule: Stay embedded in customer conversations continuously (not just periodic check-ins) to build an insight advantage that's harder to copy than product features.
threshold: 100 customers, daily Slack presence in every one.
rationale: Implementation work has gotten very fast, so product advantage is copyable overnight; customer empathy/insight is not.
In their words
More and more of the moat for any startup is the customer insight, the empathy of like actually understanding what they want. And if we can have a differentiated advantage around the customer insight, that's going to be much more valuable than having a product advantage, which is copyable overnight.
Slop automation from over-easy workflow creation
Source
Training Data — Rebuilding IT From the Ground Up for the AI Age: Serval's Jake Stauch (May 19, 2026) · at 3:26 ▸ · at 5:43 ▸
When this applies
when lowering the barrier to create any artifact (code, content, configs) leads to proliferation of near-duplicate, low-quality outputs
The takeaway
Lowering the cost of creating something removes the implicit curation that scarcity used to provide, so you must add explicit curation/governance to prevent duplication and confusion.
Why it works
Ease of creation removes the natural friction that used to force consolidation or reuse; without governance, many people independently solve the same problem, creating redundant artifacts that confuse downstream systems (e.g., which workflow should the AI run).
One level up
Every unlock of ease creates a new bottleneck somewhere else in the system — plan for where it moves to.
Where it stops holding
This stops working as a concern when usage volume is low enough that duplicates are rare or easily caught by a single owner.
Detail
situation: Making it very easy to build automations (natural-language-to-code) led users to create many near-duplicate workflows, e.g., the 20th password reset workflow nearly identical to 19 prior ones.
decision: Built a new agent with full contextual awareness of all prior workflows that proactively flags duplicates and recommends consolidation/categorization when a user tries to create a new one.
outcome: The agent acts as an expert assistant translating business requirements into product implementation while preventing workflow sprawl.
lesson: Ease-of-creation tools need a governance layer with memory of prior artifacts, or they generate 'slop' that confuses both humans and downstream AI.
In their words
We built a new agent basically on top of Serval that we're really excited about that has full contextual awareness of all the workflows you've ever built before, how they work, what they're going to do.