Cached at:
04/20/26, 02:53 PM
# Unify engineers growth by using the right model for every task
Source: [https://openai.com/index/unify/](https://openai.com/index/unify/)
[Unify\(opens in a new window\)](http://unifygtm.com/)is a system of action built to reach the right customer, at the right time, with the right message, at scale\. The platform enables this through intelligent prospecting, hyper\-personalized messaging, and a repeatable, always\-on workflow that automates what’s working\.
Unify built its system with AI at its foundation\. The team believes sales growth should be treated as an engineering problem: one that is observable, measurable, and fast to iterate\. Unify used OpenAI o3, GPT‑4\.1, and the Computer\-Using Agent \(CUA\), along with continuous model benchmarking, to design a system that helps go\-to\-market \(GTM\) teams generate millions in pipeline with a fraction of the manual work, and which now generates 30% of its pipeline\.
“There’s something really special about human\-to\-human interactions that isn’t going away,” says co\-founder Connor Heggie\. “We’re using AI to automate the busywork and give teams the leverage to spend more time where it matters most: talking to customers and making timely decisions\.”
The job of GTM teams is to find the companies and people who have a problem your product can uniquely solve and get in front of them\. While many consider this an acquisition problem, Unify thinks of this as a search problem over huge amounts of unstructured, semantically rich data\.
Historically, solving this problem meant hiring large teams of sellers to manually research prospects, comb through websites, and write outreach from scratch\. With OpenAI’s models and Unify’s agentic architecture, teams can now run this search at scale with consistency, context, and speed\.
Unify’s AI product stack includes three core components:
- **The Observation Model**runs in the background, continuously researching companies in a customer’s total addressable market\. Powered by OpenAI o3, it detects high\-signal events from publicly available sources, like new hires or stack changes and uses multi\-agent workflows to surface strategic insights\.
- **The research agent**helps users answer open\-ended questions and generate outbound copy, like flagging if a company has recently made an expansion into a new region\. It uses GPT‑4\.1 for planning, CUA for dynamic browsing, and GPT‑4o for synthesis\.
- **Copywriting**, powered by agentic research and GPT‑4o, takes relevant data surfaced by the platform and turns it into drafts of hyper\-personalized emails, with details tailored to each lead\.
Unify built its platform expecting models to improve dramatically in reasoning, tool use, and interaction\. That foresight allowed them to invest ahead, scoping their Observation Model before the models to power it were released\.
When OpenAI o3 became available, Unify had already built and tested evaluations to determine whether it could reason through complex upstream decisions\. Once validated, the team immediately deployed o3 into its workflows, improving signal detection and enabling a more agentic system overall\.
Similarly, GPT‑4\.1 and CUA unlocked new research and planning tasks that had not been possible before, and were integrated directly into the product’s decisioning layer\.
Rather than evaluating models by accuracy or latency alone, Unify built structured tests for reasoning quality in real\-world GTM scenarios\. Those evaluations are critical at upstream stages like signal classification and directive planning, where model output shapes everything that follows\.
“One of the hardest parts of the search problem is identifying the right signals,” Kunal Rai, Software Engineer at Unify, notes\. “Reasoning quality really matters in those early steps, and we needed models that could perform near human\-level reasoning, understand nuance, and adapt across diverse use cases\.”
The team evaluated OpenAI o3 against other open\-source models\. o3 stood out for its two\- to three\-turn reasoning, powering early stage logic in Unify’s Observation Model\. GPT‑4o became the default for synthesis and reply classification thanks to its structured outputs and fluency\. CUA unlocked complex research tasks that static scraping could not support, including UI\-level work like navigating review sites or Trust and Safety pages\.
Unify’s mission is to help the best products win\. By designing their platform around AI agents and task\-specific models, Unify turns growth into a science, surfacing high\-intent signals to drive timely, relevant outreach\.
Today, that system powers 30% of Unify’s own pipeline and supports hundreds of millions in pipeline for its customers each year\. As model capabilities continue to evolve, Unify plans to push even more growth workflows into agentic, always\-on systems, giving GTM teams more visibility, more leverage, and more time to focus on meaningful conversations\.
“We’re trying to create a world where success is determined by product quality and customer fit,” says Heggie\. “We built this system so humans could focus on what matters most\. And OpenAI’s models help make that possible\.”