AI Daily · 2026-09-26

AI tools are moving deeper into concrete business workflows rather than staying as general assistants. OpenAI’s Proaction case shows a non-technical founder using Codex to generate custom interactive demos, consolidate customer signals, and act on next steps, contributing to a 60% sales increase and more than 70 hours saved each month. LangChain reinforced production agents from two sides: integrating TypeSafe’s Jev for fast, cheap narrow decisions orchestrated by LangGraph, and launching LangSmith Custom Apps so teams can build shared interfaces for evaluation, labeling, and review around agent data. Replit’s acquisition of Atta applies the same logic to business analytics, embedding analysis directly into the working environment to let people closest to the problem investigate and act. Together, these moves lower the barrier for non-engineers and embed AI into sales, decision-making, evaluation, and analysis workflows.

North America · First-hand

OpenAI

⭐⭐ [Other] Proaction boosts sales 60% and saves 75+ hours with Codex

OpenAI published a customer story on how Proaction, a fleet-management software startup, uses Codex (along with GPT-Live-1 and GPT-6 Astra) to build, operate, and sell its product. Colin Knudsen, a non-technical co-founder and COO, feeds Codex the Granola recordings of sales calls, prospect email threads, and shared spreadsheets, and produces four to six customized interactive HTML demos a month, each taking 30 to 45 minutes to build. He estimates this spares 40 to 60 hours of engineering work per month, and the demos also serve as a visual reference for engineers once a prospect becomes a customer. Proaction says the custom demos raised the share of deals moving from first contact into solution development by 50% to 60%, increased sales 60%, and saved the founders 33 hours a month. Colin also uses Codex plugins for Granola, Gmail, Slack, Linear, GitHub, and HubSpot to gather customer context and act on it, plus a scheduled automation that reviews recent calls and prepares sales updates.

Ecosystem & Beyond (Products / Agents / Tools / Opinions)

Product Update

⭐⭐ [Product Update] Building Production Agents with Jev and LangGraph

A LangChain blog post explains how to build production agents using TypeSafe AI's Jev model together with LangGraph. Jev does not generate text; it returns typed answers with probabilities for narrow decision tasks such as routing and classification, and TypeSafe's benchmarks show it running up to 200x faster and 400x cheaper than leading LLMs on those tasks. The post describes an AI-powered software approach in which code owns the workflow and AI handles narrow, structured decisions, and lists Jev's properties: structured output, parallel queries, speed, and self-consistency across repeated runs. LangGraph handles orchestration, addressing context management and reliable execution.

⭐⭐ [Product Update] LangSmith Custom Apps: Build custom interfaces around your agent data

LangChain launched Custom Apps in LangSmith, letting teams build, publish, and run custom UIs on top of LangSmith data directly inside LangSmith. Users can describe in LangSmith Chat what data they want and how to visualize it to generate a working app, or start from provided templates and build against the LangSmith API with a coding agent. Once published, an app becomes a shared interface in the workspace, so teams avoid handling hosting, auth, and permissions, and can reuse it for recurring workflows. A highlighted use case is tailoring annotation interfaces to reviewers: engineers can see full traces including tool calls and intermediate steps, while subject matter experts see only the request, response, relevant context, and a rubric. LangChain also describes an internal Custom App for structured LLM evaluation and annotation with review queues and dashboards that write results into a dataset. Custom Apps are available on Plus and Enterprise plans.

Other

⭐⭐ [Other] Putting business analysis in everyone’s hands

Replit announced that it has acquired Atta, bringing Atta's approach to business analysis and purpose-built charting into Replit. Replit says its mission is to make useful intelligence accessible to everyone, which for businesses means letting the people closest to a problem investigate it, explain what they find, and act on their expertise. The post notes that product, marketing, sales and revenue operations teams each understand their customers, channels and pipeline, yet answering questions with data often requires turning that knowledge into an analysis request, waiting for results, and repeating the process for every follow-up. Atta shares Replit's belief, and the two are building toward a future in which analysis is part of the work itself rather than a separate destination.