This $1.3 Billion Startup Records Employees’ Work To Train AI

追加: 2026-05-14

[00:00:00]Today on Forbes, this $1.3 billion startup records employees work to train AI.

[00:00:10]Every company these days wants to figure out how to automate people's work with AI.

[00:00:15]Turns out AI can also help with that.

[00:00:19]Founded in 2019, San Francisco based startup Scribe makes a browser extension that sits on employees laptops, recording their screens and silently watching them work.

[00:00:30]Along with giving businesses insight into the steps involved in repetitive tasks, Scribe's AI software can then automatically generate step-by-step guides and tutorials that clearly explain how different teams operate, complete with annotated screenshots and click instructions.

[00:00:48]That's also perfect for teaching AI agents how people work, what to do, which tools to use, and how to handle different tasks on their own.

[00:00:57]CEO and co-founder Jennifer Smith says, quote, "Companies are realizing we need to make our organizations legible to humans and agents."

[00:01:08]Today, 80,000 customers, including LinkedIn, HubSpot, and T-Mobile, use Scribe's guides to train new employees on complex workflows and zero in on inefficiencies, helping them save time and money.

[00:01:23]Teaching agents, rather than humans, is still nascent. [music] Thanks to Scribe, $1.2 billion in 2025 revenue marketing software firm Clavio learned its sales reps were spending hours switching between different tools to find information about prospective customers, a process that can be automated in the future.

[00:01:45]At another company, Scribe found that customer service representatives had to go through 20 different tools like email and Teams to answer the customer's simple question of, quote, "Where is my order?"

[00:01:57]As a third, Scribe discovered that support reps were spending more than 400 hours copying and pasting between different systems.

[00:02:06]The AI wave has served as a tailwind for Scribe's [music] growth.

[00:02:10]The company announced last week that it crossed $100 million in annualized recurring revenue in April.

[00:02:16]About $8.3 million in revenue that month.

[00:02:20]More than 6 million employees have Scribe's app on their laptop, and businesses pay anywhere from a $20 subscription to five to seven-figure contracts.

[00:02:29]Some 600,000 organizations use the app's free version, which captures work people do in a browser, whereas the paid plans can record desktop applications, too.

[00:02:41]The startup's AI models, built on top of OpenAI, Anthropic, and Google systems, have recorded and analyzed 15 million different workflows, like onboarding new hires, managing tickets, or responding to customers.

[00:02:56]And observed how people are working across 40,000 business applications.

[00:03:00]The startup hit a $1.3 billion valuation after raising a $75 million Series C round in November.

[00:03:09]The startup's roots go back to Smith's consultancy days.

[00:03:12]Working as a business analyst at McKinsey more than 15 years ago, Smith realized most companies had a limited understanding of the actual work. The clicking, typing, and copy-pasting across different applications employees spent hours on.

[00:03:28]The way to learn what people were doing so their bosses could make them more efficient >> [music] >> was even more painstaking.

[00:03:34]Smith would pull up a chair behind an employee and peer over their shoulder, taking notes on exactly what they did, >> [music] >> and interviewed dozens of employees who'd found better ways to use tools.

[00:03:46]She says, quote, "Institutional know-how lives in people's heads. It's arguably a business's most important asset and it's not something you actually own or can see or use.

[00:03:58]Smith was convinced there was a better way than hovering over an employee's shoulder, awkward as it was.

[00:04:04]The answer?

[00:04:05]AI.

[00:04:07]In the summer of 2019, she teamed up with co-founder Aaron Podolny to start Scribe.

[00:04:13]The company offers two products: Scribe Capture, which records workflows and converts them into guides, and Scribe Optimize, which analyzes all that data to find inefficiencies to improve.

[00:04:25]Smith claims the app only captures data from work applications like Slack and Salesforce.

[00:04:31]It anonymizes the data by identifying trends at a team level, rather than monitoring individual activity, like where employees spend most of their time.

[00:04:41]Smith says, quote, "It's built to measure the work, not the workers."

[00:04:47]That means Scribe isn't watching if people are watching YouTube during their lunch break, for instance. Though, you'll have to take their word [music] for it.

[00:04:56]Tools that track and record employees' work to increase productivity have been around for decades, but so-called boss ware exploded during the pandemic, to workers' great chagrin.

[00:05:08]Thanks to AI, such tools have only become more prominent and sophisticated, tracking everything from keystrokes and mouse movements to clicks.

[00:05:18]In late April, Meta told thousands of employees that it would track everything they do on their computers to train AI models on their day-to-day tasks.

[00:05:26]Angry employees called it a privacy violation and a quote "callous decision," according to the New York Times.

[00:05:34]Scribe customers insist that the application isn't to surveil workers, rather to deeply answer the question, "What do we do here?" and figure out how it can be improved.

[00:05:45]For full coverage, check out Rashi Shrivastava's piece on forbes.com.

[00:05:49][music] This is Kieran Meadows from Forbes.

[00:05:54]Thanks for tuning in.