The end of work — or how generative AI will change more than you think
AKA how Teamhouse is thinking about the future of work
Generative AI (GAI) was a novelty when we started building our first POCs for Teamhouse in January, but it’s now clear that software development has changed forever.
The current Wild West of experimentation is already accelerating into a ubiquitous landscape of transformative products and services, and GAI will soon be an every-day component of the user experience, a product consideration as familiar and necessary as design.
Lightspeed’s Alex Kayyal recently shared a venture capitalist’s perspective of GAI and its potential for SaaS products, declaring the dawn of AI-native companies as SaaS 4.0 — The Era of Cognition.
Central to this narrative, is the belief that AI will transform the relationship between the knowledge worker and their tools; AI will 10X the value proposition of SaaS by augmenting, supplementing and productising currently manual workflows. The efficiency of business processes will sky-rocket, the customer’s experience will become immersive and intuitive, and employees will ascend to more meaningful positions of purpose — or never be employed to begin with — and so on.
But what Lightspeed are describing isn’t so much where the hockey puck is going; while the tooling is still embryonic, this is what’s being built and deployed now. The well-worn adage that we overestimate what can happen in the short term and underestimate what can happen in the long term — it may just have met its exception.
Yes, many incumbents and well-funded challengers have spent years developing their own AI capabilities, but the playing field has been levelled and reset. Fresh-faced startups, without the burden of technical debt, can almost immediately meet their effort; a single Python developer in 2023 can deliver functionality in days, that took an engineering team weeks and months in 2022.
The direction of this current trend is irreversible. It’s not going to splutter and die like so many cycles of AR and VR because there’s immediate utility for countless use cases, and with APIs doing the heavy lifting, weekend hackathons and global corporations alike can experiment with ease.
So if the future of cognition is already here, in an embryonic but exponentially evolving state, what comes next?
The Lightspeed post (and other thought pieces regarding the application of GAI in SaaS) regards processes and workflows as foundational throughout this transformation — they’ll be automated, augmented, more efficient, but they’re still considered to be existing and recognisable functions of business.
We believe that owning workflows in businesses is the secret to building a successful SaaS product in the age of GAI. The models are irrelevant without deep, contextually relevant data, meaning that individual SaaS companies with significant usage can create and curate smaller sets of training data that are deeply relevant to their vertical.
But we believe that processes and workflows will become entirely irrelevant to the average knowledge worker. Further, we believe that workflows, as we understand them today, will cease to exist at all.
What do we mean by that?
For now, we’ll consider the first part of that statement — that process and workflows will become irrelevant.
Here’s IBM’s definition of a workflow:
A workflow can be defined as a system for managing repetitive processes and tasks which occur in a particular order, while a business process is considered more complex, consisting of multiple workflows, information systems, data, people and their activity patterns. A workflow is distinguished by its simplicity and repeatability, and it is generally visualized with diagram or checklist.
Workflows are generally considered to be structured; workflows are repeatable from end-to-end, and while there may be variables or permutations, these are considered and work won’t stray outside these parameters.
Today
Workflow automation and RPA already exists and is big business. What tends to slow down adoption and ROI in larger companies, is workflow creation; educating customers, mapping processes, and integrating tools.
And that’s something that Generative AI is already capable of in enterprise-grade software. Take a look at Tray’s Merlin AI:
Obviously, the team at Tray have spent years building a comprehensive framework that can support and manage the integrations suggested by Merlin, but GAI has already has the potential to slash time-to-value for new Enterprise customers, and improve productivity across business functions.
Tomorrow
The key to automating workflows is their repetitive nature. That’s also why humans want rid of them — people get bored and become inefficient when they have to do the same work over and over.
But you can’t always automate all of a workflow, because while we don’t want to do the work, we still need to do some of it — specifically, the decisions, approvals and discussions that need to occur throughout.
In workflow automation, we call these “human-in-the-loop” steps. Some decisions can be deferred to code because they’re trivial and based on logic, but what about more subjective decisions, especially when they impact the health of a workforce — an employee’s career progression, their prospects, their happiness?
Decisions deferred to machines has been a fundamental aspect of life for years — most of us have experienced these as a consumer. We struggle to accept (what we consider) subjective decisions meted out by machines (denied bank loans, cancelled services, unfair fees etc) and immediately seek intervention from a flesh-and-blood representative.
The similarly blunt-force application of GAI-powered determination to a workforce (a valued workforce, at least) would likely lead to resignations and a toxic, destructive culture; not because GAI doesn’t have the ability to be nuanced in its considerations, but because human nature requires us to point a finger at somebody for decisions we disagree with.
So if you accept that:
- Generative AI will continue improving in its ability to suggest, build, execute and correct operational workflows
- Consequential, subjective business decisions are likely to remain the purview of managers, not machines
What happens then?
You may reach the same conclusion as us — that not only will effort be abstracted from process, but that process will be abstracted from employees.
In fact, we believe the concepts of processes and workflows will become as irrelevant to the end-user as code, as tasks and tools become seamlessly orchestrated in AI-generated workflows that self-generate, organise and correct.
As a human-in-the-loop, your only contribution will be a daily list of approvals and determinations, discussions and suggestions — the work that matters, that deserves your attention, that you are uniquely qualified to consider.
Companies will no longer make hires purely for back office admin, because that type of work won’t exist. If we’re employed, it’ll be for our perspective and analytical skills. Our hands-on experience of utilising computers to perform laborious and endlessly repetitive sequences of administrative tasks will become a dim memory for the majority of the white-collar workforce, one we’ll endeavour to bore our disbelieving future children about.
This is the first in a series of posts about our thoughts the future of work, as written by Hanna, Jack and Paul.
Together, we’re building Teamhouse, a GAI-powered ticketing and workflow automation platform for People Operations (think next-gen Service Now).
If you’re interested in learning more, please get in touch.
