2026-05-22-intro-dustin-jeremy-ross-jeremy-jawish

Debrief: Intro Dustin Jeremy Ross Jeremy Jawish — 2026-05-22

Summary

Dustin spoke with Jeremy Jawish, co-founder and CEO of Shift Technology, a fraud detection and AI platform for business insurance. The conversation covered Jeremy’s entrepreneurial philosophy, Shift’s vertical product strategy, the impact of AI agents on their go-to-market, and Dustin’s early semiconductor supply chain financialization thesis. Jeremy offered pointed feedback on the compliance wedge approach and strongly recommended a financial or warranty product structure over a logistics or compliance entry point.

Key Themes

Jeremy’s entrepreneurial journey and business-building philosophy: Jeremy was candid that Shift Technology began without a grand vision — he met his co-founder, did an insurance internship, and saw a problem worth solving. The first five years were improvisational (‘go with the flow, try to do bigger, faster’). The shift to deliberate planning came not for his own benefit but to give his team direction and comfort. He framed plans as psychological infrastructure rather than execution blueprints: ‘You need to have plans for people to know where they’re going. But if you need to change plans overnight, just change plans overnight.’ This is a useful foil for Project TBD’s own tendency to start from grand vision and work backward.

Vertical vs. horizontal product strategy: Jeremy argued strongly that companies should pick one lane — vertical or horizontal — and commit fully. For Shift, vertical (insurance) with multiple products was the right call. He noted that fraud detection alone is a small niche, insufficient to reach billion-dollar scale. Their customers now average 2.3 products each across a five-product suite, with two of those products not working. His advice to Dustin: ‘In vertical, you need to have multiple products,’ and you must be ‘willing to take risks on products that might not work at all.’ He also acknowledged the inverse: some niches are worth several billion dollars on their own, making multi-product expansion unnecessary.

AI agents restructuring Shift’s product and GTM: Shift is decomposing its five existing products into discrete agents, currently at seven products with a target of ten within three months. This isn’t new product creation so much as granular modularization — one fraud solution becomes a detection agent and a separate investigation/memo-generation agent. Jeremy noted that insurers are actively pivoting away from buying ‘big solutions’ toward buying individual agents, but the appetite distribution across specific agents is still unclear: ‘We’re just bombarding the market.’ This creates both a commercial opportunity and real strategic fog.

Liability and accountability gaps in fully automated AI agents: Jeremy gave a detailed and concrete example — an AI agent autonomously negotiating injury claims with third-party attorneys. If the agent makes an illegal or coercive move, it’s unclear who bears liability: the insurer, Shift, or the model provider. ‘Imagine you get a 3 million penalty and it’s because the agent screwed up. Is it the insurance company’s fault? Is it Shift’s fault?’ This liability vacuum is preventing full automation and forcing manual human checkpoints to remain in place, creating a structural bottleneck in the very market they’re trying to modernize.

Project TBD’s semiconductor supply chain financialization thesis: Dustin walked Jeremy through the core hypothesis: build a proprietary data asset from semiconductor supply chain activity and use it to underwrite financial products (price indices, derivatives, insurance) that others can’t offer, drawing comparisons to Coalition Cyber in the cyber insurance space. Jeremy engaged with this constructively and didn’t dismiss it, but redirected toward the financial/warranty product structure as the faster path to validation rather than logistics or compliance.

Compliance-based entry and incentive misalignment: Dustin described the founders’ early compliance wedge hypothesis — that automating trade compliance for semiconductor companies could serve as a data-generating wedge. He noted the hypothesis broke down because companies sourcing from or selling to restricted entities often prefer not to surface those violations. Jeremy challenged the magnitude of the deterrent: ‘Is there any fine they could be fined for it?’ When told the fines exist but aren’t sufficient, he agreed: ‘Okay. It’s not that big.’ This is the sharpest outside validation yet that the compliance wedge has an incentive alignment problem, though it sits in tension with the AI synthesis memos’ enforcement severity data.

Parametric insurance market limitations: Jeremy was direct that parametric insurance remains a small and underperforming niche. Customers prefer traditional models and prioritize price over simplicity. He pointed out that claims processing represents only 15% of premium cost, meaning that innovations in claim efficiency have a limited addressable base. He noted that even Coalition’s parametric cyber market is small. This is a meaningful constraint on one of Project TBD’s financial product hypotheses.

Defense tech opportunity with structural constraints: Jeremy noted defense as a capital-abundant sector with high investor appetite, but flagged sovereignty requirements as a hard structural limit — US government contracts require American-only companies. He suggested that defense could be a beachhead that unlocks other sectors once trust is established, but the entry filter is real.

Notable Quotations

“You need to have plans for people to know where they’re going. But if you need to change plans overnight, just change plans overnight.” — Jeremy Jawish. Context: Describing his philosophy on planning as team infrastructure rather than rigid strategy; relevant to Project TBD’s own early-stage planning instincts.

“In vertical, if you want to be in a vertical, you need to have multiple products.” — Jeremy Jawish. Context: Core strategic advice on vertical SaaS economics; directly applicable to how Project TBD should think about product sequencing beyond a compliance wedge.

“The parametric market is still small and it’s not worth it. People are just not comfortable with parametric triggers for insurance policies.” — Jeremy Jawish. Context: Challenging a key financial product hypothesis in Project TBD’s semiconductor supply chain thesis.

Themes & Contradictions

This conversation creates meaningful friction with the two AI synthesis memos (GEMINI and CLAUDE) and partial confirmation from the Richard Dasher meeting.

Conflict — compliance wedge viability: Both AI memos rank the compliance wedge as the top near-term thesis, with CLAUDE scoring it 3.95/5.00 and citing ‘enforcement-driven pain highest of any thesis’ backed by the Applied Materials $252M and Cadence $140M+ settlements. Jeremy’s direct reaction was skepticism: he questioned whether fines are large enough to change behavior, and Dustin confirmed from customer conversations that fines haven’t been sufficient deterrents. This is the second time this incentive misalignment has surfaced — Dustin raised it as a finding in the meeting itself (‘these companies actually don’t want to have better compliance’). The AI memos appear to have modeled demand from the regulator’s perspective (fines are large) rather than the buyer’s perspective (fines are manageable relative to business benefit of non-compliance). Jeremy’s pushback sharpens this gap but doesn’t resolve it — the fines ARE large in absolute terms, and the ECCN/BIS regime has escalated materially.

Partial confirmation — financial product structure: Jeremy’s recommendation to pursue a financial or warranty product structure rather than logistics aligns with Thesis II (Insurance and Structured Risk Products) in both AI memos, which noted that battery warranty reinsurance is ‘commercially real’ (Munich Re/Hithium). Jeremy didn’t endorse any specific product but his directional guidance — faster feedback loops, capital availability from reinsurers — maps closely to what the memos flag as Thesis II’s structural advantage.

Confirmation — parametric market limitations: Jeremy’s skepticism about parametric insurance echoes the CLAUDE memo’s note that the EU Battery Passport Regulation 2023/1542 ‘restricts data reuse — fatally weakens data-asset-first models.’ Both arrive at caution about parametric from different directions.

Alignment with Dasher: Richard Dasher flagged critical materials dependency and defense supply chain vulnerabilities as macro tailwinds. Jeremy’s defense tech commentary (capital-abundant, sovereignty-constrained) is directionally consistent with Dasher’s framing, though neither went deep on semiconductors specifically.

Business Problems & Painpoints

Jeremy did not describe pain as a buyer or operator in the semiconductor space — he is an outsider to that domain. His pain signals are therefore analogical rather than direct.

Indirect pain signal — compliance incentive misalignment: Jeremy’s quick acceptance of Dustin’s finding that companies don’t want compliance surfaced (‘Okay, it’s not that big’) suggests he has seen similar dynamics in insurance fraud — where surfacing fraud can create internal political problems for the insurer. The pain of NOT knowing may be easier to sell than the pain of knowing. This is a product positioning implication.

Structural pain — agent liability void: Jeremy described a live operational problem at Shift: no one — not the insurer, not Shift, not the model provider — is willing to accept liability for fully automated agent decisions, particularly in claim negotiation. The result is a forced human checkpoint that defeats much of the automation value. This is not Project TBD’s problem directly, but it signals a category of liability infrastructure that doesn’t yet exist in AI-native vertical SaaS — potentially relevant if Project TBD builds compliance or underwriting agents.

Strategic pain — niche sizing: Jeremy’s observation that fraud detection is too small a niche for billion-dollar outcomes without multi-product expansion is a cautionary signal for Project TBD. If the semiconductor compliance niche is similarly narrow, single-product revenue may be insufficient. He framed this not as a problem to solve but as a strategic reality to plan around: ‘Depends when you want to build. If you want to build a company that is a niche company…’

Implicit pain — parametric market immaturity: Jeremy’s comment that parametric insurance customers prefer price over simplicity (‘they want best price over simplicity’) suggests that financial product innovation in insurance faces a demand-side conservatism problem, not just a supply-side data problem. Buyers may not value novelty enough to pay for it.

Emotional Signals

Jeremy was relaxed, direct, and generous with his time. He engaged most energetically when discussing the agent liability problem — this is clearly live and unresolved at Shift, and his tone shifted from reflective to animated when walking through the scenario of an agent making an illegal negotiating move. He seemed genuinely uncertain about the path forward there, which is unusual for someone who otherwise spoke with confidence. He was mildly dismissive of parametric insurance (‘it’s not worth it’) with the flat affect of someone who has already investigated and closed the door. His China commentary carried a subtle wistfulness — ‘I used to be fascinated by China’ — followed by deliberate redirection to focus, suggesting he’s made peace with deprioritizing it. No resistance or guardedness was detected around the Project TBD questions; he engaged the thesis as a peer, not a skeptic.

For Founders

1. Jeremy accepted Dustin’s compliance incentive misalignment finding quickly and without challenge — but the AI synthesis memos cite $252M+ fines as evidence of existential enforcement severity. Are these two views actually in conflict, or are they describing different buyer segments (e.g., large fabs vs. mid-tier distributors who calculate differently)? What would it take to know which dynamic dominates?

2. Jeremy’s strongest directional advice was to pursue a financial or warranty product structure because of faster feedback loops and reinsurer capital availability — but he also said parametric insurance markets are small because customers prioritize price over simplicity. How do Bliss and Dustin reconcile the ‘go financial’ advice with the ‘financial products face demand-side conservatism’ observation, and does the semiconductor context change the calculus?

3. Jeremy described planning primarily as a team management tool — ‘the planning is not for me, it’s for the team’ — and Shift spent its first five years going with the flow before becoming more deliberate. Project TBD is doing the opposite: starting from a grand vision and working backward to a wedge. Is that top-down structure an asset that speeds validation, or a constraint that might cause the founders to force-fit early evidence to a pre-formed thesis?