My AI versus The Company AI: How Knowledge Workers Conceptualize Forms of AI Assistance in the Workplace

Share Share Share Share Share
DOWNLOAD PDF

GSuite is changing the nature of Knowledge Work across 5 million businesses through AI-powered assistance. To ensure that this evolution reflects the aspirations and priorities of workers, Google and Stripe Partners conducted a multi-national ethnography of Knowledge Workers covering a range of industries. We identified that workers distinguish between ‘Core’ and ‘Peripheral’ work: the work they are paid to do and identify with, and the work that does not contribute to their success or happiness. Workers want assistance to enhance Core work and remove Peripheral work, nuanced across a spectrum of support. This framework and taxonomy has been adopted by teams at Google to inform strategic decisions on how AI is integrated by GSuite. New features are being implemented within Gmail, Slides, Docs and Sheets that bring these principles to life in the user experience.

INTRODUCTION

AI and automation are often spoken of as threats to human agency due to their potential to take over activities that humans are currently doing at work. In mainstream media narratives (e.g. Forbes, 2018) AI-based technologies are presented as something that is either present (and takes over) or absent (leaving humans in charge). This creates a false dichotomy and unhelpful distinction between the two states.

This paper is based on joint research conducted by Google and Stripe Partners in 2018. The objective of the research was to investigate the role of assistance, as idea and practice, in professional knowledge work. Data for this paper is derived from ethnographic interviews and workplace participant observation in several European countries.

Our research revealed the relationship between AI and workers is more nuanced than is often portrayed. We found that knowledge workers do not fear AI in of itself, but have a fine-tuned sense of how they want to perceive and experience its role in their work. These distinctions can vary between workers, driven by personal ideas of status, identity and professional responsibility.

The recommendations and insights informed both Google's short term product strategy for G-Suite, as well as providing a number of foundational frameworks and common taxonomies that have been adopted across the organisation from leadership to different product teams.

OVERVIEW OF PROJECT

Research context

Recent reports (e.g. Davis et al., 2018) illustrate how the world of work is changing because of AI. Some jobs are being automated, while others are evolving. There are many technologies and services that are driving this shift. GSuite has been adopted by over 5 million businesses around the world, and the AI-driven features it integrates make it an important actor in this context.

Strategically, GSuite is focused on supporting the evolution of human knowledge work rather than automating it. GSuite's stated mission is to elevate human accomplishment through machine learning augmented tools in the workplace. The objective is to help people to focus on their most important tasks, and, in doing so, enable companies to thrive.

GSuite is poised for the next wave of change in collaborative work. Individual contribution is almost always just one piece of a puzzle within complex knowledge workflows. The Google research team were looking to enable this collaboration not just within GSuite's products, but across the products they use everyday.

Google believes that people should be able to collaborate in context, with Machine Learning and AI features built-in. Consequently, these capabilities should augment how people at work collaborate. This must be done responsibly and to the benefit of workers and businesses. Hence a focus on such tools is an opportunity of investment in Google's customer's employees and their company's culture. Google also is aware that great care need to be taken when designing with AI Principles (https://ai.google/principles/) and Responsible AI Practices (https://ai.google/responsibilities/responsible-ai-practices/)

RESEARCH OBJECTIVES

As researchers we realised we needed to dig deeper than these strategic principles to translate this vision of AI-powered work from the perspective of workers.

So we embarked on a program to understand the world of knowledge workers, exploring questions such as:

  • what tasks in their everyday work do they value, which ones do they loath?
  • which activities in their roles do they believe they give most value to their employers?
  • what are the opportunities for G-Suite to provide Creative Assistance during the process of content creation: what types of work would people most appreciate having replaced or helped by AI?

Importantly, by taking a ‘bottom-up’ perspective the project sought to provide the team with an understanding of what assistance workers need today. This focus meant that resultant outcomes are designed to support existing working practices rather than replace them. Our research focus was therefore on incremental improvements to existing working practices, rather than analysing workers systematically to identify opportunities to fundamentally change or remove roles.

KEY OUTCOMES

The main contribution of the project within Google has been twofold (see more detail in ‘Implications and Impact’ section below)

  1. Embedding a new set of taxonomies and frameworks that inform AI-related decision making throughout the GSuite organization
    • The frameworks outlined in this case study have been socialised across both the executive and product layers of the organisation, helping teams prioritise and develop strategies for integrating AI into their products
  2. Driving product innovation within specific GSuite teams
    • Many product teams at GSuite (Gmail, Calender, Sheets, Docs) have now adopted these frameworks to inspire and guide how they integrate AI into their products, with many examples of new features already live

METHODOLOGY

Challenges to address

With a research brief to ‘explore attitudes to AI-assistance in professional knowledge work’ there was a significant methodological challenge for the research team in how to cover this topic that moved beyond existing tropes (both positive and negative) driven by the public discourse on AI and its potential role for work in the future. Researching technology that is not yet in (widespread) use is always a challenge as there is often no obvious existing behaviour to look at or existing preferences to discuss and explore. How is this possible to explore ethnographically? The problem is exacerbated because research participants could struggle to distinguish between prominent media-driven perceptions and the reality of their own behaviour.

Furthermore, knowledge work is a nebulous concept with ambiguous boundaries (Cross, Taylor & Zehner, 2018). Attempting to cover it in one research project is exceedingly difficult. It is broad in the range of people who do it (from secretaries to lawyers to nuclear scientists), in the range of activities it describes, in the range of (types of) organizations it takes place in and in the range of meanings attached to it. Academic research into knowledge work is typically either very abstract, looking to draw out general principles of knowledge work (Davenport & Prusak. 1998) or more narrow and not even attempting to say anything about the topic of knowledge work as a whole, but rather say something relevant about a specific type of work, workers or places. The challenge for this research was in doing ethnographically grounded research that would lead to insights with implications across the entire spectrum of knowledge work.

OUR RESEARCH APPROACH

Terms of reference

‘Knowledge Work’ was a term coined by Peter Drucker (Drucker, 1969). As commonly understood, it describes the growing cohort of workers who “think for a living”. Knowledge Work is therefore a broad category! Our study encompassed a range of knowledge workers: from designers to accountants to administrators to engineers to brand strategists. Nearly all our participants worked for large organizations and were primarily based in corporate HQs rather than remote working (although some remote working practices were observed). Within this, there was a mix of levels. We spoke to everyone from senior leaders to support staff. Everyone we spoke to existed within a wider team with whom they produced work collaboratively, although the frequency and intensity of collaboration with co-workers did vary across our sample.

‘Creative Assistance’ is a term used within Google to describe forms of AI that support knowledge workers within the GSuite product experience. This includes technologies that have been launched in the last 24 months such as Smart Compose in GMail (https://support.google.com/mail/answer/9116836?co=GENIE.Platform%3DDesktop&hl=en) and Suggested Layouts in Slides (https://support.google.com/docs/answer/7130307?visit_id=637038105256693940-2801069891&p=suggest_layouts&hl=en&rd=1)

Researching knowledge work

Highly skilled knowledge work is a complex process constituted by small tasks executed by individuals. These add up to larger tasks and workflows executed by multiple individuals, which lead toward desired outcomes. Researching such work requires mixed approaches in order to explore its complexity. For this research it entailed a combination of research with individuals and organizations.

Pages: 1 2 3

Leave a Reply