HFS Highlight: Hive brings cloud-enabled AI to the battle against hate speech – and bags $50m in Series D funding

A cloud-enabled AI start-up, with a crowdsourced two-million-strong distributed workforce and a timely solution to tackle content moderation in social media, has raised $50m of Series D on a $2b valuation.

San Francisco-based Hive has established a strong presence for its models in automated content moderation – including a ‘hate model’ which came to market just as the death of George Floyd triggered an explosion in Black Lives Matter activism – and with it the need for social media platforms to tackle hate speech.

It offers access to hosted machine learning models via APIs and charges for them on a usage basis. Its pre-trained models can automate previously manual or non-scalable tasks of interpreting image, video, text, and audio data. While its predominant use case is content moderation automation, Hive is also being used for contextual advertising, advertising and sponsorship measurement, and document parsing.

On top of a previously unannounced $35m in Series C, San Francisco-based Hive plans to invest its $50m war chest added in April, responding to the disruption of COVID-19. The pandemic has made automation mission-critical for business (exhibit 1).

Solutions target the pandemic-accelerated demand for process transformation

CEO Kevin Guo shares the HFS view that the COVID-19 pandemic not only pushed companies to automate legacy processes with Robotic Process Automation, it also forced them to rethink how work is done – accelerating the roadmap for the use of AI in new processes such as jump-starting AI models with tagged and vetted training data or making content moderation less manual. Hive is intent on providing enterprise AI solutions to power that wave. That ambition aligns with the findings in exhibit 1 – where automation as the catalyst to modernizing legacy business practices is a strategy now being aggressively pursued by 65% of the 400 global 2000 leaders, surveyed.

Exhibit 1: The pandemic has made automation the catalyst to modernizing legacy practices

Source: HFS Research, 2021. Sample 400 executives from global 2000

Enterprise solutions for data in the cloud help meet OneOffice aspirations

HFS believes that a new roadmap – towards a OneOffice approach in which processes run end-to-end across the organizational value chain – is powered by an integrated stack of emerging tech (exhibit 2) that complements the core, and natively automates your processes – enabling people and powering decisions.  You must design processes to achieve the outcomes and data you need in the cloud, and you must automate them to make that possible (whether through AI, APIs, customized code, or off-the-shelf software). We see a role for vendors such as Hive in creating enterprise solutions that meet the need for data in the cloud (see Data Optimization in Exhibit 2) that supports this next wave of intelligent automation.

Exhibit 2: HFS OneOffice Emerging Tech Platform

Source: HFS Research, 2021. Examples are representative

Business built on the founder’s frustration with existing out-of-box AI models

Hive was created four years ago in response to the need for accurate out-of-the-box AI models. It was built on a frustration the founder experienced himself when developing a consumer app and finding nothing on the market accurate enough to handle use cases such as content moderation.

The solutions Guo and his team have come up with mean data scientists can skip the critical and time-hungry (and soul-crushing) step of labeling large volumes of data to train models, enabling enterprises to embrace AI faster.

Revenues rise 300 percent in a year with 100-plus enterprise customers

In the last 12 months, Hive revenues are up by more than 300% – and the customer base now numbers 100-plus. Those customers include Reddit, ComScore, Walmart, Tango, Bain, and Cognizant.

Hive’s cloud-based, pre-trained, AI models are accessible via APIs. Those models are built on the belief that vast amounts of quality training data are the most important factor in building machine learning models and they have become one of the larger training data labeling start-ups as a result – with a global ‘workforce’ of crowdsourced contributors, increasingly paid in Bitcoin, now numbering more than 2 million.

The Series D investment in Hive – not to be confused with productivity software company, was led by Glynn Capital. Its recent Series C was led by Tomales Bay Capital. The firm has raised $121M to date.

The Bottom Line: Hive drives AI into process automation and out of the back office

Process automation has been dominated by RPA for the past few years, with its deterministic rules-based automation often incorrectly touted as intelligent. Hive is the real deal, offering vetted training data and pre-trained models to aid in cognitive automation of overly manual business processes. Often process automation arises as a reaction to making business processes tied to aging monolithic enterprise applications a little less manual and painful. While Hive is targeting manual processes, it’s mission is not propping up legacy systems – rather it is unleashing the power of AI to help enterprises make applied use of it with manual processes.   Hive is well-placed to support the drive to Native Automation. Its out-of-the-box models have already done the hard yards of labeling large volumes of data to train models – a big bonus for data scientists which gives enterprises a valuable shortcut to embrace AI faster.