Digital Manual Labor, Email And AI

Last month Jacob Wenger, co-founder of Shortwave, was a guest on The Future Of Email podcast series, hosted by yours truly. Shortwave provides a mobile and web email client — “Intelligent email, powered by AI.”  Jacob and his colleagues have been working on this for several years — before ChatGPT blew up in the media.

Jacob remarked that an individual’s email is probably the largest set of writing and personal knowledge of any app they use, which makes it a powerful candidate for the application of AI technologies. Shortwave summarizes message threads intelligently and clearly. It pulls together “like things” (travel, calendar appointments, newsletters). Some additional feature highlights:

• Automatic language translation of summaries
• Group emails for batch handling
• Auto-apply labels
• Batch email delivery to receive on your schedule

To me, this is the kind of improvement-of-use application of AI that Bill Gates hinted at when he said that AI is (now) the most exciting thing he’s observed since the graphical user interface. The email clients in broad use — Outlook, Gmail, Yahoo, etc — have highly-evolved GUIs. Just about everything that can be done with UX design has been tried, tested and refined in those GUIs — and, generally speaking, the individual inbox is still a hot mess. Apparently, the basic structures of email, laid down like so many Roman roads by the basic protocols of email, plus a healthy dose of human behavior, can’t quite be tamed by any graphical interface. If AI can reduce the workload without adding its own workload…great!

I’ll go out on a limb and suggest a macro-level prediction about AI from that: there’s a tremendous amount of work (and business) to be done, applying AI to reduce mundane tasks.

Most of the people I’d expect to read this post probably do “knowledge work” of some sort. It sounds impressive, and it is, but honestly, there is an awful lot of “digital manual labor” involved in knowledge work right now. If you have more than one browser tab open at a time, or more than one application running at a time, you are likely playing “human API”, moving digital stuff between different systems. It’s normal, but it also sucks; a huge amount of human attention gets burned up keeping track of this-GUI-versus-that. The number of applications and tabs we juggle in the course of a workday — and I’m one of the worst offenders I know — is just stupid.

It’s a far cry from Senate hearings about the risks of AI, and while there are serious things to consider about AI that justify that kind of attention, there’s also quite a bit of useful stuff to be done about all that digital manual labor.

There are market structures and market forces to all of those open tabs, though. None of the far-too-many SaaS companies behind each of them wants to be obsoleted. So we’ll see more of what we’re already seeing — AI-powered features-and-functions coming to existing products and services. That’s fine — done well, less manual labor a la Shortwave. That’s not a recipe for market revolution or market disruption.

The API economy that’s already in place is a fine accelerant to this market evolution. OpenAI — the company behind ChatGPT — brought the ChatGPT API to market along with the web-based chat interface that brought them 100M users in 2 months or so. While the chat interface is fun and fascinating and all of that, jumping back and forth from the ChatGPT tab to the some-other-tab is more digital manual labor. We don’t want more of that, right?  Tapping AI engines — ChatGPT and others — to do things within existing interfaces and systems makes sense.

OI member Paul Christmann — Chief Innovation Officer of Rasa.io (note: The Future Of Email newsletter is powered by Rasa) — shared a perfect example of this kind of API-driven improvement. The newest beta feature in Rasa is a “Generate with ChatGPT” to help editors draft newsletter intros and conclusions. It’s better and easier to have a contextual feature that handles “state” (it’s a newsletter, draft, target text length, English, etc, etc) to frame the AI response than to draft all of that into a prompt in another tab.

AI will spawn some revolutionary products and services, no doubt. I’d love to see Shortwave win their bold shot at an email-client revolution, for one!  But I think that the dynamics at play make AI look more like what Clayton Christensen — author of The Innovator’s Dilemma and the theory of disruptive innovation — called “sustaining innovation.” 

In the domain of email and email marketing, specifically, reducing “digital manual labor” is long overdue. Email marketers handle an almost absurd variety of digital tasks — data tasks, technical tasks, visual/creative work, writing, market analysis, audience segmentation, and the list goes on as the tabs open up. If our AI overlords take over even a fraction of the manual copy-and-paste stuff involved today, it’s tough to see a downside to that.

There’s a funny and slightly strange collision there, if you think about it, in the necessary collision of the graphic interface with artificial intelligence. GUIs enabled people to do a far broader range of digital jobs than character-mode interfaces, but GUIs presume a human intelligence making meaning of their graphical cues. Artificial intelligences may eventually learn how to ‘comprehend’ GUIs, so they can do useful things in context to replace that “digital manual labor”. It’s going to take some serious work; as evolved as GUIs are, they’re inconsistent and highly context-dependent because the GUI user has always been assumed to be human.

Ancient quote, probably apocryphal: ‘May you be doomed to live in interesting times.’  Apt here!

DALL·E 2023-05-22 11.28.50 - A man digging a hole with a laptop computer at the end of a shovel handle, photorealistic.