Rory O'Driscoll

Partner at Scale Venture Partners; recurring 20VC co-host.

Voice: SaaS veteran, fluent in multiples and financial math. Dry humour; sharp counterpoints to hype. Voice lives mostly in podcast appearances, less so on LinkedIn.

Frameworks

• Corporate layoffs attributed to AI fall into five distinct categories: (1) excuse for overhiring, (2) declining growth misattributed to productivity, (3) capex reallocation to AI infrastructure, (4) genuine productivity gains, and (5) skills mismatch requiring different talent profiles.
• Evaluate growth and burn together via a 2x2 matrix of growth rate vs. burn rate; companies in different quadrants require different strategies, with high-growth/high-burn being the hardest to fix because cutting burn is easier than preserving growth while lowering burn multiple.
• SaaS sales target the functional user performing work, while AI sales target the decision-maker seeking business outcomes—a fundamental shift in buyer persona and value proposition.
• Companies that flatten at their initial product hit the 'TAM trap' and trade at 2-4x revenue unless they add adjacent products; incremental growth points command exponential valuation premiums (e.g., 55% growth → 35x vs 28% growth → 20x).
• AI capex sustainability requires a productivity-per-worker ROI calculation, while game theory explains why overinvestment persists despite questionable returns—the foundation model companies can still win by being subsidized through hyperscaler commodity competition.
• When evaluating high-multiple acquisitions, compare quality-adjusted revenues by weighting each revenue stream by the underlying business quality (profitability, margins, durability) rather than accepting nominal multiples at face value.
• When facing a transformative opportunity, there are three strategic paths determined by your resources and time horizon: build enabling infrastructure (modest outcome), acquire and transform an incumbent (good outcome), or build full-stack from scratch (exceptional outcome). Your choice is constrained by capital and runway.
• When high-growth cloud valuations face inevitable slowdown, companies must choose one of three strategies: fight (compete for existing market), focus (narrow to defensible niche), or fly (build next-generation intelligent automation).
• In times of maximum uncertainty, planning should follow a structured sequence: acknowledge what you don't know, execute clear no-regrets moves, then address existential threats.
• In competitive battles, the default outcome is 'winning ugly'—the winner spends disproportionately more than the loser to achieve dominance. The exceptional inverse pattern is when a competitor catches up at dramatically lower cost and faster speed, signaling structural advantage.

Principles

• AI software defensibility only emerges at scale through winning market position, not through early technical novelty. Investors choose between early-stage product-market fit risk versus late-stage execution risk, trading valuation for information.
• For AI adoption in software companies, internal process improvements (GTM, engineering) are table stakes that create no advantage. Competitive differentiation requires either delivering a meaningfully different end product to customers or achieving a unique cost structure.
• When a technology triggers simultaneous universal demand (demand shock), initial hypergrowth masks an approaching saturation cliff that invalidates linear growth extrapolations—distinguish durable adoption curves from one-time market exhaustion events.
• Venture capital has bifurcated into two separate businesses: ultra-late-stage private/public-style investing (concentrated, large deals) and traditional early-stage venture (Series A/B, dispersed). These operate independently with different dynamics.
• When new-paradigm companies are valued far beyond the TAM they replace, valuation mismatches make M&A impossible—incumbents can't afford to buy, startups can't afford to sell—forcing a binary outcome of scale-to-IPO or failure.
• Integration depth is the primary predictor of retention in enterprise software; products with many integrations create switching costs that anchor customers regardless of price competition.
• The best companies start narrow in a defensible niche with strong margins, then expand TAM as they scale—rather than pursuing vast markets where small share implies commoditization.
• Expert predictions about technology displacement are systematically overconfident and should not drive major life decisions; predictive humility is warranted even from domain leaders.
• Surviving managers must stay close enough to the work to verify reality themselves; distance from execution creates catastrophic blind spots during downturns.
• SaaS thrives in domains where compliance risk and deterministic outcomes are mandatory; AI changes implementation but not willingness-to-pay when getting it wrong carries severe penalties.
• Infrastructure winners emerge independently from platform shifts rather than in direct correlation with them; winning infrastructure companies benefit from but are not determined by underlying platform timing.
• In venture capital, early success predicts future success not just through network effects (referrals) but by building risk tolerance—past wins give investors the psychological courage to make bold bets.
• When you've missed an early round, paying up for the next round is rational—the incremental data point between rounds dramatically reduces uncertainty and justifies the higher price.
• Public market multiples inversely validate the venture model: low multiples for slow-growth companies and absurdly high multiples for high-growth companies mean venture's bet on extreme growth still generates outlier returns despite overall market compression.
• When market direction becomes obvious, capital floods in and execution speed becomes the dominant variable; investors must accelerate their own process to match the new tempo.
• To get board approval on compensation, founders should proactively package all necessary information and rationale in advance, making it easy for investors to say yes.
• When the potential outcome is enormous, even a low-probability bet has high expected value due to the asymmetric upside.
• M&A strategy should be driven by existential threat to the core business model, not simply additive growth—acquire to replace a dying revenue stream, not to compound a healthy one.
• When an IPO pops far beyond typical 20%, long-term investors hit their exit targets immediately and sell rather than hold, undermining post-IPO stability that issuers want.
• AI startups exhibit a distinct GTM pattern: faster sales cycles paired with lower win rates and extended CAC payback periods compared to traditional SaaS.
• Major regulatory or economic disruptions create proportional opportunities for automation startups that reduce the friction those disruptions introduce.
• In-house lawyer enthusiasm is a rare and highly predictive signal of product-market fit in legal tech; when you get glowing references from a notoriously hard-to-please user group, it represents a meaningful market anomaly worth acting on.

Opinions

• Legacy SaaS companies facing AI disruption must prove they can transition existing customers to AI-native versions of their products; if they can't succeed with a perfect setup (large customer base, obvious use case), the second act transition model fails for the entire sector.
• Public market SaaS valuations remain distorted until AI-native high-growth companies go public, providing investors with the actual reference point to rationally price legacy 10% growth/30% FCF companies against mythical 300% growth/negative cash flow alternatives.
• AI-era companies will achieve strong free cash flow through lower headcount efficiency (e.g., $1B revenue with 100 employees at 60% gross margins) rather than replicating SaaS-era 70%+ gross margins. Market share in new categories goes through a 'chrysalis period' of fluidity before congealing into durable positions once enterprises standardize.
• Wall Street's embrace of AI required abandoning SaaS love—capital flows as zero-sum between narratives. When extreme growth outliers emerge (10x GAAP revenue YoY for 3 years), capital rushes to 'singularity bets' rather than slower picks-and-shovels plays.
• Capital alone provides no competitive advantage or differentiation for startups.
• Your fund size determines your strategy (LP type, check size, diversification); calculate minimum viable fund size by working backwards from target portfolio construction, then set a deliberately conservative public target to create fundraising momentum.
• At massive scale, sheer capital deployment and market presence become a structural competitive advantage that forces competitors to expend disproportionate effort to achieve visibility.
• AI revenue growth depends on expanding total addressable market rather than reallocating existing enterprise software budgets; companies must build simple products for mass adoption to achieve meaningful market expansion.
• LLMs will automate routine work in white-collar roles including security analysis, pushing value toward high-skill work and making those skills premium.
• Post-Covid SaaS growth deceleration requires deliberate, company-wide effort to reaccelerate ARR; default trajectory won't naturally restore prior growth rates.

Predictions

• When growth slows and capital becomes expensive, every SaaS company must execute a 'get profitable while preserving upside' playbook—a transition Box successfully demonstrated in 2020 that will become universal in 2023.

Voice samples

• > "AI has become the justification for every layoff. It's the perfect excuse card, but there is a lot of spin involved."
• > "When you sold SaaS, you sold to the department that did the work. When you sell AI, you sell to the person who wants the outcome."
• > "Revenues that come from profitably sending rockets into space are worth 5X revenues that come from reselling tokens at 0 GM."
• > "If you're a software company today, there is only one question that determines success or failure: how does AI change the end product you deliver your customers?"
• > "Anthropic is doing the opposite. They've caught OpenAI in half the time at a quarter of the training cost. If you're OpenAI looking at these numbers, that's a bad fact pattern."
• > "Don't let AI predictions ruin your life. Remember, nobody knows shit."
• > "Capital is a crappy differentiator."
• > "You're not interested in vibe coding your payroll. You want to outsource this to someone wildly competent, have them take the responsibility."

Generated from 1605 items, 51 kept after dedup. Full attribution: `logs/rory-odriscoll.jsonl`.