No PriorsNo Priors Ep. 73 | With Airtable co-founder and CEO Howie Liu
CHAPTERS
- 0:00 – 0:40
Airtable today: from “spreadsheet on steroids” to app platform (and why AI now)
Sarah introduces Howie Liu and frames Airtable’s scale and recent launch of AI features. Howie positions Airtable not just as productivity software, but as an easy-to-use application platform that enables real apps on structured data.
- •Airtable serves ~500k organizations and is increasingly enterprise-relevant
- •Reframing Airtable: not a spreadsheet tool, but an app platform under the hood
- •The episode’s focus: low/no-code, enterprise transformation, and enterprise AI
- 0:40 – 2:19
Origin story: learning the platform playbook at Salesforce
Howie explains the core inspiration for Airtable: Salesforce’s success came from being a customizable platform, not just a CRM feature set. Airtable aimed to democratize that platform model for “citizen developers” with far less technical friction.
- •Salesforce taught the power of the platform + customization model
- •Goal: make app-building accessible beyond traditional developers
- •Airtable’s ambition from the start was broader than a single “killer app”
- 2:19 – 5:34
Beating “build an app, not a platform”: spreadsheets as the adoption wedge
Sarah and Elad challenge the conventional wisdom of starting with a platform. Howie argues Airtable succeeded by anchoring itself in a familiar paradigm—spreadsheets—lowering the learning curve while quietly providing relational structure and app-building power.
- •Spreadsheets are the most prolific ‘app-building platform’ people already know
- •Design choices intentionally preserved spreadsheet familiarity (grid UX, shortcuts, copy/paste)
- •Most spreadsheet workflows are really database/workflow problems in disguise
- 5:34 – 6:03
From horizontal platform to repeatable use cases: templates and “many apps”
Howie describes how Airtable still needed “apps,” but in a scalable way—via templates and use-case guidance rather than single-purpose software. This balanced broad platform capability with practical starting points for users.
- •Templates served as an early mechanism to package use cases
- •Airtable effectively ships many lightweight apps rather than one vertical product
- •Guidance helps users cross the gap from generic tool to concrete workflow
- 6:03 – 8:26
Enterprise transition: low floor, rising ceiling (scale + extensibility)
Sarah asks how Airtable evolved into an enterprise-facing company. Howie explains this was always the plan: start with a low floor (easy adoption) and progressively raise the ceiling via scale, robustness, and extensibility needed for demanding enterprise workflows.
- •Strategic arc: simple first, then increasingly powerful/customizable
- •Scaling: from thousands of rows to hundreds of thousands and millions
- •Extensibility: scripting, integrations, serverless execution, and automations
- •Enterprise push was gated by platform robustness—early versions would have broken
- 8:26 – 9:35
Product-led growth as an enterprise “cheat sheet”: spotting hotspots
Howie explains how organic adoption inside large companies revealed where Airtable delivered high value. Airtable then doubled down on the emerging hotspots (e.g., content production workflows) to make solutions more repeatable in go-to-market and product.
- •PLG created internal enterprise footholds before formal enterprise sales focus
- •Usage patterns revealed which workflows to prioritize
- •Airtable leaned into repeatable enterprise processes discovered in the customer base
- 9:35 – 13:57
Howie’s evolving product management model: three hats, not one
Prompted by Sarah, Howie lays out why PM is hard and how his view shifted: PM often conflates distinct responsibilities. He highlights three critical “hats”: program management, product marketing/market understanding, and complex UX/information architecture.
- •PM role often over-indexes on execution/program management
- •Product marketing hat: market, competition, JTBD, and customer framing
- •Third hat: complex UX and information architecture for dense products like Airtable
- •Teams can split hats across leaders (design/eng/PM) as long as all are covered
- 13:57 – 16:31
Solving uniquely hard problems as culture—and applying it to AI productization
Howie connects Airtable’s differentiation to choosing hard, non-commoditized problems, citing an NVIDIA/Jensen Huang perspective on hard-problem selection. He argues Airtable’s success depends on turning model capabilities into usable product primitives and workflows—not just “bolting on” AI.
- •Hard problems avoid commoditization and demand distinctive product thinking
- •Airtable’s platform approach differs from simply building vertical SaaS clones
- •AI success hinges on productizing models into workflows and UX users can trust
- 16:31 – 20:57
Howie’s AI journey: from neural nets to LLM reasoning as the real unlock
Howie traces a long-running interest in neural networks from the ‘AI winter’ through ImageNet-era breakthroughs to today’s LLMs. He emphasizes that reasoning (not novelty text generation) is the underexploited capability and that current models already enable far more value than realized.
- •Early exposure to neural nets and the idea of ‘learning rules from data’
- •Image breakthroughs (CNNs) showed scalability and real-world impact
- •Transformer/LLM era: reasoning and synthesis are the biggest step-change
- •Belief: even with model progress frozen today, massive untapped value remains
- 20:57 – 22:27
Airtable AI strategy: not just AI in the product—AI for customer-built AI apps
Elad asks about early AI integration and product choices. Howie explains Airtable’s dual mandate: improve Airtable’s own UX with AI and, more importantly, enable customers to build AI-powered workflows atop their data, automations, and interfaces.
- •Two layers: AI-enhanced Airtable UX and enabling customers’ AI apps
- •Initial beta focused on ‘runtime capability’: easy LLM calls embedded in workflows
- •Value comes from combining AI with first-party data, repeatable processes, and automation
- 22:27 – 26:19
The real blocker is imagination and know-how: hands-on customer design + templates
Howie describes what they learned from a large AI beta: enterprises are intimidated and lack intuition for what models can do beyond chat experimentation. Airtable addressed this with deep customer partnership plus prompt templates and guided patterns to make AI feel concrete and repeatable.
- •Market immaturity: fear/unknowns and limited understanding of model capabilities
- •ChatGPT-style play doesn’t translate automatically into business workflow redesign
- •Approach 1: hands-on design partnerships with customers (e.g., legal workflows)
- •Approach 2: templates/prompt libraries tied to common Airtable use cases
- 26:19 – 29:12
Beyond chat: structured, recurring workflows and human-in-the-loop reliability
Elad asks what’s missing to unlock more product value. Howie argues the industry over-focuses on chat UX; the bigger opportunity is deliberate automation inside structured processes with resilient outputs, human review, and chaining AI + human + automation steps.
- •Chat/RAG is useful but only a small slice of AI’s enterprise potential
- •Big wins come from process decomposition: which steps can be automated?
- •Need error-resilient designs: approvals, edits, and human-in-the-loop checkpoints
- •A platform with data + workflows is essential to address the long-tail of enterprise processes
- 29:12 – 35:39
Scaling AI intuition: workshops, “AI design patterns,” and productized primitives
Sarah asks how Airtable codifies AI intuition for customers. Howie outlines Airtable’s AI workshops (lightweight vs. Palantir-style bootcamps), emerging workflow “design patterns” (e.g., self-critique chains for translation), and new product primitives like many-shot prompting and AI-driven recommendations.
- •AI workshops teach foundations, prompting techniques, and real customer patterns
- •Design patterns for workflows (e.g., generate → critique → revise/review) improve quality
- •Product roadmap: many-shot prompting, learning from examples, and feedback loops
- •Future onboarding: start from content sources (Drive/Box/Gong) → extract to structured tables + workflows
- 35:39 – 41:25
Code generation vs. no-code: why no-code remains the editable interface for humans
Sarah asks whether code generation changes Airtable’s long-term relevance. Howie argues code-gen helps developers, but non-developers still need outputs they can inspect and iteratively modify—data schemas, logic, and interfaces in a human-readable no-code format—especially for complex business apps.
- •Code-gen is powerful for snippets and developer augmentation, but complexity is the barrier
- •Sophisticated enterprise apps (ERP-like workflows) are too nuanced to generate end-to-end reliably
- •Non-technical users need inspectable, directly manipulable representations (schemas, flows, UI)
- •Without AGI, AI-generated apps still need human steering—no-code is the control surface
