Udi Menkes
Product/engineering voice in the newsletter discussing agent UIs and component-based interfaces in chat apps. He recommends evaluating component choice and content accuracy with a flexible harness.
Key Highlights
- Udi Menkes is repeatedly cited for practical guidance on agent interfaces, especially component-based experiences inside chat products.
- He recommends evaluating both answer correctness and UI component choice with a flexible harness rather than treating agent output as text alone.
- His Intuit work centers on AI Business Advisor, including peer benchmarking and next-step recommendations based on actions and outcomes across many businesses.
- He argues AI PMs should validate simpler retrieval baselines and strengthen customer discovery before adding more complex architectures.
- His commentary also connects product design to monetization, including pricing and support for very high-spending AI customers.
At a glance
- Role
- Product/engineering commentator
Udi Menkes
Overview
Udi Menkes is a product and engineering voice cited across the newsletter for practical thinking on agent user interfaces, evaluation, AI product strategy, and business-advice systems. He appears most often in discussions about how agents should present answers inside chat surfaces like Slack and Teams, how teams should evaluate both the correctness of answers and the choice of UI components, and how AI products can move beyond text-only responses into structured, component-based experiences.For AI Product Managers, Menkes matters because his examples connect product design, model behavior, and operating metrics. His commentary spans flexible eval harnesses for agent UIs, retrieval baselines such as file-exploring agents versus semantic search, continuous discovery practices, pricing implications of power users, and Intuit’s AI Business Advisor, which benchmarks businesses and recommends next steps based on observed actions and outcomes at scale.
Key Developments
- 2026-08-02: Shared AI Engineer’s video of his talk on Intuit’s financial-advice system, describing a system that derives millions of business state-action-outcome trajectories, uses reinforcement learning to select better actions, and trains an LLM to generate advice.
- 2026-09-08: Recapped Amplitude’s Wave as a production system that identifies issues through analytics, opens PRs, deploys approved changes, and measures results, highlighting concrete product-improvement metrics.
- 2026-09-16: Argued that product teams need structured continuous discovery, including asking about past behavior, using an opportunity-solution tree, and involving engineers in customer interviews; he warned that faster building via agents does not replace understanding the problem.
- 2026-10-05: Said every product now has “2 customers,” suggesting his prior brand model is no longer sufficient in an AI environment.
- 2026-10-11: Recapped Noah Shinn and Instinct’s hospitality demo that generated multiple hotel-booking interfaces tailored to different user intents; Menkes argued that builders should define components, signals, assembly rules, and constraints so agents can assemble the right interface.
- 2026-10-31: Recapped Perplexity’s internal one-domain test in which a file-exploring agent outperformed semantic search on retrieval, answer accuracy, and speed; he recommended testing an `ls/grep/read` baseline before adding a vector index.
- 2026-11-03: Commented on Ramp data showing extreme concentration in AI spend among top firms and recommended designing pricing, limits, and support around high-spending accounts rather than relying only on per-seat pricing.
- 2026-11-07: Announced a new AI Business Advisor that he said draws on the actions and outcomes of millions of businesses, enabling benchmarking and specific advice.
- 2026-11-09: Shared that Intuit’s AI Business Advisor, which he leads, includes peer benchmarking similar to NADA 20 Groups, comparing a business with similar businesses, identifying strengths and weaknesses, and advising on next steps.
- 2026-11-10: Recapped an implementation for agents answering in chat apps such as Slack and Teams, where responses render as charts, customer tables, and export/share buttons; he recommended evaluating both component choice and content accuracy with a flexible harness, and noted Vercel’s json-render and Chat SDK as useful building blocks.
Relevance to AI PMs
1. Design agent experiences as interfaces, not just answers. Menkes repeatedly emphasizes that agent outputs should be assembled from preapproved components based on user intent. For PMs, this means defining a component library, decision signals, and guardrails before shipping agent experiences.2. Evaluate the UI decision and the content decision separately. His chat-agent examples suggest a tactical eval framework: measure whether the answer is correct, whether the selected component is appropriate, and whether the workflow supports actions like export, sharing, or drill-down.
3. Use simpler retrieval and stronger discovery before overcomplicating the stack. His recommendation to test `ls/grep/read` before vector search, along with his emphasis on continuous discovery, gives PMs a practical reminder to validate user problems and baseline architectures before adding sophistication.
Related
- Intuit: Menkes leads AI Business Advisor work there, especially around benchmarking and next-step recommendations for businesses.
- AI Business Advisor: Closely associated with his product work; the system compares businesses to peers and advises on improvements.
- Vercel, json-render, Chat SDK: Connected through his discussion of component-based agent interfaces rendered in chat apps.
- Ramp: Source of data he used to argue for AI pricing and support models tailored to high-spend customers.
- Perplexity: Referenced in his recap of retrieval experiments comparing file-exploring agents with semantic search.
- Amplitude / Wave: Featured in his examples of production AI systems that detect issues, ship fixes, and measure outcomes.
- Noah Shinn / Instinct: Used in his commentary on adaptive interfaces assembled from components, signals, and constraints.
- AI Engineer: Shared his talk on Intuit’s state-action-outcome approach to AI-generated financial advice.
Frequently asked questions
What is Udi Menkes?
Product/engineering voice in the newsletter discussing agent UIs and component-based interfaces in chat apps. He recommends evaluating component choice and content accuracy with a flexible harness.
What should AI product managers know about Udi Menkes?
Udi Menkes is repeatedly cited for practical guidance on agent interfaces, especially component-based experiences inside chat products. He recommends evaluating both answer correctness and UI component choice with a flexible harness rather than treating agent output as text alone. His Intuit work centers on AI Business Advisor, including peer benchmarking and next-step recommendations based on actions and outcomes across many businesses.
Explore this topic
Newsletter Mentions (38)
“in Udi Menkes recapped an implementation introduced “this month” for agents answering in chats such as Slack and Teams, with a video showing a Q3 revenue response rendered as charts, a customer table, and export and sharing buttons.”
#14 in Udi Menkes recapped an implementation introduced “this month” for agents answering in chats such as Slack and Teams, with a video showing a Q3 revenue response rendered as charts, a customer table, and export and sharing buttons. He recommends evaluating both component choice and content accuracy and using a flexible harness; Vercel’s json-render and Chat SDK let agents build interfaces from preapproved components.
“Udi Menkes 🚢 shared that Intuit’s AI Business Advisor, which he leads, includes peer benchmarking similar to NADA 20 Groups, where car dealers compare P&Ls and learn from one another’s numbers.”
#16 𝕏 Udi Menkes 🚢 shared that Intuit’s AI Business Advisor, which he leads, includes peer benchmarking similar to NADA 20 Groups, where car dealers compare P&Ls and learn from one another’s numbers. The product compares a business with similar businesses, highlights strengths and areas for improvement, and advises on next steps.
“in Udi Menkes announced a new AI Business Advisor that he says draws on the actions and outcomes of millions of businesses.”
#11 in Udi Menkes announced a new AI Business Advisor that he says draws on the actions and outcomes of millions of businesses. Business owners can enter information, benchmark across several areas, and receive specific advice through the linked tool.
“Udi Menkes 🚢 commented that, per Ramp data, the top 1% of AI-paying firms on Ramp spend $9,568 per employee a month on AI versus $15.70 for the median firm—about 609x more.”
#13 𝕏 Udi Menkes 🚢 commented that, per Ramp data, the top 1% of AI-paying firms on Ramp spend $9,568 per employee a month on AI versus $15.70 for the median firm—about 609x more. He recommends designing pricing, limits, and support for these high-spending accounts rather than relying only on per-seat plans.
“Udi Menkes 🚢 recapped Perplexity’s internal, one-domain test in which a file-exploring agent beat semantic search on correct-document retrieval (~87% vs ~39%), answer accuracy (63% vs ~37%), and speed (~25% faster). He recommended testing an ls/grep/read baseline on evals before adding a vector index.”
#6 𝕏 Udi Menkes 🚢 recapped Perplexity’s internal, one-domain test in which a file-exploring agent beat semantic search on correct-document retrieval (~87% vs ~39%), answer accuracy (63% vs ~37%), and speed (~25% faster). He recommended testing an ls/grep/read baseline on evals before adding a vector index.
“#10 in Udi Menkes recaps Noah Shinn, founder of Instinct, demonstrating Instinct’s hospitality experience on October 9 (year unspecified): it generates three different hotel-booking interfaces based on location, price, or family needs, then books the selected room.”
#10 in Udi Menkes recaps Noah Shinn, founder of Instinct, demonstrating Instinct’s hospitality experience on October 9 (year unspecified): it generates three different hotel-booking interfaces based on location, price, or family needs, then books the selected room. For AI product builders, Menkes argues for defining components, signals, assembly rules, and constraints so agents can tailor interfaces to each user’s intent.
“Udi Menkes 🚢 commented that every product now has “2 customers,” making his long-used brand model (becoming the name a person reaches for first) incomplete.”
#12 𝕏 Udi Menkes 🚢 commented that every product now has “2 customers,” making his long-used brand model (becoming the name a person reaches for first) incomplete. Also covered by: @Udi Menkes 🚢 , @Udi Menkes
“Udi Menkes argues that product teams need structured continuous discovery: ask about past behavior, use an opportunity-solution tree, and include engineers in customer interviews. As agents increase building capabilities, he says effective discovery is becoming rarer—and building faster without understanding the problem only produces more products nobody wants.”
#13 in Udi Menkes argues that product teams need structured continuous discovery: ask about past behavior, use an opportunity-solution tree, and include engineers in customer interviews. As agents increase building capabilities, he says effective discovery is becoming rarer—and building faster without understanding the problem only produces more products nobody wants. #14 in Guillermo Rauch announced Vercel Labs, Vercel’s public research and experimentation arm for sharing what the company supports and researches, along with experiments that did not pan out.
“#2 in Udi Menkes recapped Amplitude’s Wave, a production product that identifies issues through analytics, opens PRs, deploys approved changes, and measures results—in one example, opening a PR within 15 minutes.”
#2 in Udi Menkes recapped Amplitude’s Wave, a production product that identifies issues through analytics, opens PRs, deploys approved changes, and measures results—in one example, opening a PR within 15 minutes. Personalization increased suggestion-card engagement by 160% in an A/B test, while other changes raised successful searches from 92% to 98%, onboarding completion from 6% to 9%, and blog sign-ups by about 45 per week.
“𝕏 Udi Menkes 🚢 shared AI Engineer’s video of his talk on Intuit’s financial-advice system, which derives millions of business state–action–outcome trajectories, uses reinforcement learning to select better actions, and trains an LLM to generate advice.”
#1 𝕏 Udi Menkes 🚢 shared AI Engineer’s video of his talk on Intuit’s financial-advice system, which derives millions of business state–action–outcome trajectories, uses reinforcement learning to select better actions, and trains an LLM to generate advice.
Related
AI company that publishes Claude models and safety guidance.
The AI company referenced in relation to research on chain-of-thought. It is relevant as the source of the research being discussed.
A coding assistant used for execution and implementation work in the tutorial referenced by Peter Yang. It is presented as part of a workflow paired with different models for different tasks.
Anthropic’s AI assistant used in the newsletter for coding and knowledge-base workflows. It appears both as a general assistant and as Claude Code for household documentation tasks.
CEO of Vercel and a prominent builder in AI infrastructure and workflows.
Cursor is an AI coding tool that introduced Team Projects in beta for shared agent context across teammates. The update emphasizes coordination, conflict handling, and team-level task unblocking.
Frontend and AI app infrastructure company behind Workflow SDK and Workflows.
Creator and newsletter writer focused on product, growth, and hiring trends. In this issue he is referenced for a recurring hiring review highlighting AI-era labor shifts.
Investor and builder credited here with an evaluation framework for comparing agent harnesses.
Technology company whose open-ontology repository is highlighted for agent query reliability and semantic context generation. For PMs, it is relevant to structured data and agent grounding.
A partner company listed in the Sign in with ChatGPT launch set. The newsletter provides no additional context beyond the integration mention.
An AI search company referenced in a note about growing usage of its hybrid cloud-orchestrator/local-subagent setup in the Mac app. It is relevant as a product organization experimenting with agent architectures.
Autonomous or semi-autonomous software assistants that can be configured, automated, and secured with tools and integrations. The newsletter discusses memory, safety, and Home Assistant automation patterns around agents.
A payments company referenced for internal agent use and operational safeguards around agentic workflows.
Linear is a product-management and issue-tracking platform that is integrating AI agents into workflow assignment. Here it is highlighted as the place where Amp can be assigned issues directly.
Ramp is a fintech/spend-management platform used here as the source of firm-level AI spending data. It provides the benchmark for comparing high-spending and median customers.
Cloud Code appears to be a coding agent or coding workflow used to generate launch videos from websites. The newsletter describes it as working with Fable 5 and HyperFrames.
Jensen Huang is the CEO of NVIDIA and a prominent advocate for AI infrastructure and open ecosystems. In this newsletter he is referenced via an NVIDIA letter about open models and defense harnesses.
An AI company or model provider mentioned for releasing a cost-efficient open-weights multimodal model. For PMs, it signals open-weight multimodal competition and cost pressure.
Reusable Claude-based skill modules that package agentic workflows into portable components. The newsletter frames them as a way to avoid building AI agents from scratch.
A script-like design artifact or workflow described as being executed by coding agents. The newsletter frames it as part of a shift toward autonomous, personalized design capabilities.
A travel and lodging platform increasingly associated with AI-driven experiences and services. The newsletter mentions it in the context of a new hire from Meta.
Stay updated on Udi Menkes
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free