PitchLens

What Conviction looks for in a pitch deck

Conviction is an AI-native venture firm purpose-built to back 'Software 3.0' companies — those that architect for AI rather than bolt chatbots onto legacy designs. Founder Sarah Guo argues we are 'living through the mother of those transitions now with AI,' a once-a-decade ground-shift, and that the winners will be the companies pursuing the most ambitious visions at the frontier. The firm invests early (often the first check), up and down the intelligence stack from chips and infrastructure to AI-native applications, and gravitates toward technical teams that build with velocity, taste, and pragmatism.

Stage focus: pre-seed, seed, Series A · Sectors: ai native applications, ai infrastructure, developer tools, inference compute, enterprise automation, legal tech, robotics, voice ai

What Conviction wants on each slide

Problem

A real, large problem in an industry where AI can now do work that was impossible before — ideally where the limiting factor is context/data rather than raw intelligence. Conviction favors founders attacking durable problems whose correct solution is expensive and private to establish (a real moat), not 'easy,' easily-measured tasks that commoditize.

“This matters because frontier models are already strong. Intelligence is no longer the bottleneck. Context is.” — CRM Was Built for Humans. Day was Built for AI.
“The valuable work is illegible by construction: anything you can put on a leaderboard, you can train against, so anything measurable is already on its way to commodity.” — Sarah Guo
“Systems that fail to capture and connect the full story of how a business operates cannot support real-time reasoning, no matter how capable the model on top appears to be.” — Sarah Guo
“The text-driven legal sector represents a huge opportunity for AI-driven change. Each generated token, if of sufficiently high quality, is worth its weight in gold.” — Sarah Guo

Solution

An AI-native architecture designed around the new capability, not a chatbot bolted onto a legacy system. Conviction wants founders who give the model the tools a human uses, decompose hard work into verifiable tasks, and obsess over 'Minimum Viable Quality' and taste — building products that feel intentional and coherent.

“Pick a problem, give the model the same tools humans use to get real work done, and optimize for the outcomes you care about.” — Conviction — Startup Ideas
“They thought about Minimum Viable Quality (MVQ), where to get the data (public and private) that would improve outputs, how to decompose sophisticated work into doable and verifiable tasks, and how to orchestrate those tasks into a seamless experience for a non-technical end user that just wants to file an S4.” — Sarah Guo
“This is taste. The relentless, almost painful ability to know what should exist, what shouldn’t, and where quality matters. It’s the difference between shipping a product and shipping a point of view.” — Sarah Guo
“Instead of asking humans to maintain records and assemble insight after the fact, Day is built so models can work directly over the live reality of the business.” — Sarah Guo

Market size

Markets large and malleable enough to birth a defining company of the decade. Conviction explicitly rewards founders who overshoot on ambition and capability predictions, and is wary of those who undershoot — the most ambitious version of an idea, attacked at the frontier, is the bet.

“We see too many startups undershoot in their capability predictions. We hope to see more startups overshoot.” — Not A Normal Market
“When they do, they tend to birth a cohort of companies that define the decades that follow.” — Sarah Guo
“There were plenty of people building online stores in 1994, but only Amazon had the audacity (and understanding) to think they could be “Earth’s biggest bookstore” and beyond.” — Sarah Guo

Traction

Evidence of velocity and real-world deployment over polished projections — 'just get live,' build the dataset and the customer trust as you ship. Conviction is comfortable backing pre-traction at first-check stage when the team's pace and technical depth predict fast iteration, but loves to see momentum (signed design partners, fast ARR ramps, production reliability).

“The most important step is to just get live.” — Conviction — Startup Ideas
“With over 250 clients today and on path to $100M of ARR, Harvey now has massive momentum.” — Sarah Guo
“yet in weeks, they could not only speak the language, but had a product vision and signed design partners.” — Sarah Guo
“real automation isn’t only the model getting better. It’s the product, the model, the workflow, and the firm moving together, and three of those four move at the speed of an organization.” — Sarah Guo

Team

Technical teams that build with velocity and pragmatism — people who are smart, have depth, move fast, ship, and have taste. Conviction weights pace, intellectual honesty, curiosity, agency, and (for leaders) the ability to move people, judging cofounder-grade character, motivation, alignment and ability.

“The most undervalued trait in startup hiring is pace.” — Pace
“As investors, we aspire to identify those rare companies who balance unreasonableness about what is possible with AI and clarity about how to get there. Both are required to reshape industries.” — Sarah Guo
“They finish things. Their history shows completion—not necessarily successful features and products, but finished ones. They often leave behind a trail of side projects.” — Sarah Guo
“It’s why, when we hire a CEO, the ability to deal with people weighs at least as much as the analytical horsepower, and a smarter model doesn’t change that weighting.” — Sarah Guo
“Each person we add to our partnership must raise the bar and make us as a partnership better. The bar for a partner is really the bar for a cofounder.” — Sarah Guo

Business model

Models that capture durable value rather than reselling commoditizing tokens — a token spent reasoning over a customer's proprietary data is worth far more than a generic one. Conviction favors focused applications that tune one workflow to a fraction of the token cost and keep the margin, and businesses that own the 'living ontology,' integrations, or operating layer of a domain.

“a token spent answering a generic question is worth almost nothing, since anyone’s model can answer it, while a token spent reasoning over your company’s data is worth much more, because it does the thing you actually want, not just the plausible thing.” — The Untrainable
“a focused application can tune one workflow until it runs on a fraction of the token spend, and unlike the lab selling those tokens, it keeps the difference.” — Sarah Guo
“The company that defines the standard may become the path of least resistance to market.” — Sarah Guo

Competition

A defensible answer to absorption from below (tasks saturating and commoditizing) and above (labs pulling scaffolding into the weights). Conviction wants moats built on private, expensive-to-establish correctness, accumulated context/data, taste, and execution velocity — not thin wrappers around a model whose value the lab can reclaim.

“Is its correctness private and expensive to establish, the 'untrainable' kind that can’t be read off a leaderboard?” — The Untrainable
“From above, the labs are trying to get the models to swallow their own scaffolding... all the apparatus that used to wrap a model is being pulled into the weights, until the wrapper is the model.” — Sarah Guo
“Correctness like that isn’t only private, it’s the slow kind of moat capital can’t collapse.” — Sarah Guo
“In competitive evaluations, this coherence becomes your edge. Polish beats features. The product that looks finished gets tagged as “enter­prise-ready”—even when it’s not.” — Sarah Guo

Why now

A credible thesis on why AI capability has crossed a threshold that makes this newly possible, and why the founder is reasoning as if capability progress continues. Conviction sees AI as a once-a-decade paradigm shift and wants founders aiming at the most ambitious, frontier version of the opportunity in this narrow window.

“We are living through the mother of those transitions now with AI.” — Not A Normal Market
“It is challenging to really reason as if you believe in continued capability progress, even when you do.” — Sarah Guo
“Computer use and vision language models are changing the equation. For the first time, we can build robotic systems that reason and act like humans.” — Sarah Guo
“But now we’re on the cusp of a step-change, where AI models can take robotics from brittle, single-purpose systems to autonomous, general-purpose machines.” — Sarah Guo

The ask

Conviction is a high-conviction lead and often the first check, writing $1M-$25M and re-investing across rounds when belief is high. Founders should be ready for a long-term, hands-on partnership rather than a transactional round; the firm deliberately does not scale and trades short-term economics for long-term potential.

“We are early-stage focused, and we invest early, and are often the first investor.” — Conviction — Conviction Partners
“I’ve invested in every round of this company, most recently in the $150M Series D, through our new venture fund, Conviction.” — Sarah Guo
“We value our reputations and our performance above fund scale. To achieve this, we expect to do a lot of work for a very long time, and continually trade off short-term economics for long term potential.” — Sarah Guo

What Conviction rewards and penalizes

What excites Conviction

  • Technical founders who ship fast — demonstrable pace, finished side projects, and short thought-to-deployment cycles
  • An AI-native architecture built around the new capability (e.g. a context graph / data model) rather than a chatbot bolted onto a legacy system
  • Clear, evident taste — a coherent point of view expressed in error messages, UX, and what was deliberately killed
  • Overshooting on ambition and capability predictions; framing the largest possible version of the opportunity
  • A defensible moat of private, expensive-to-establish correctness, accumulated context/data, or owned integrations — not a thin wrapper
  • Founders who balance unreasonableness about what AI makes possible with grounded, customer-centric clarity on how to get there
  • Decisiveness with incomplete information and comfort being wrong ~40% of the time to be right faster

Watch-outs for Conviction

  • Undershooting on AI capability — building for today's models instead of next year's
  • Bolting AI/chatbots onto an incomplete legacy data model where intelligence, not context, is assumed to be the bottleneck
  • Building on easily-measured, benchmark-able tasks that will commoditize from below or be absorbed into the model from above
  • 'Taste' claimed as a pitch buzzword but not backed by costly product choices ('preference,' not taste)
  • Slow, 'manager-mode' leaders who need excessive data or consensus before acting and recreate big-company structures
  • Founders who confuse taste with surface aesthetics (nice font, design agency) rather than depth in error states and core flows

Who decides at Conviction

Sarah Guo — Founder & Managing Partner
“As software becomes abundant, the ability to make intent clear becomes scarce.”
Mike Vernal — General Partner
“What do you want in a cofounder? Character, motivation, alignment and ability.”

What Conviction actually backs

Where Conviction is thinking now

Conviction's published guidance

Benchmark your deck against Conviction →

PitchLens grades your deck slide-by-slide against Conviction's own published playbook, with a citation on every point.

← Browse all 124 VC firms · PitchLens home