The Turn: Superintelligence or Superunknown?
A renewed look at chat-as-an-interface for HR software
Eddie Vedder flew to Seattle in 1990 to rehearsal for a band that wasn’t even called Pearl Jam yet.
He ended up at a Temple of the Dog rehearsal instead, more or less by accident. Somewhere in that session, Chris Cornell needed him on “Hunger Strike,” a song that became one of the defining tracks of a decade of Pacific Northwest grunge.
The rest is history.
Serendipity probably isn’t a good way to justify the mechanics behind a killer song or a user interface for SaaS. But it’s what kept going through my head when I saw the announcement from Workday about Sana’s “Superintelligence” offering, which is part connector, part agent orchestrator, and part deep AI chat capabilities for the enterprise.
The last part follows a trend which I’m not sure I like.
Visual interfaces have always been full of accidental on-ramps. You hover over something you didn’t mean to click and discover a feature you didn’t know existed. You scroll past a setting, you flip through a menu, you stumble, and you might find something interesting.
A visual interface has texture. It has information without asking. It has those moments of friction and serendipity that people have come to expect.
Chat has none of that. If you don’t know that you don’t know, well, then you don’t know. Even the stuff that it’s really good at, like pulling information from all over the place and making some semblance of sense, requires some level of knowledge to even ask.
It’s easy to make fun of a dashboard that looks like it was created when the Seattle Supersonics still existed. But those usually existed because someone was making a decision based on it.
I will say that for users with motor impairments or low vision, chat with voice enabled can be a genuine step forward. Have you ever had an accessibility feature read you a data chart? Don’t (unless you need it). A chat interface that can be descriptive instead of visual is a godsend.
For regular users of ChatGPT, there’s probably a level of comfort (especially at the beginner level). But for power users, for those who really need the full ugly functionality of these enterprise tools, it’s going to be a sidestep at best.
Natural language chat interfaces are built for ambiguity. That’s why they are so good for beginners. But strip all that away and the supposedly frictionless interface requires you interact with a different level of precision. You train yourself to issue commands so that you get expected results. So your brain and fingers become a different type of UI.
Obviously Workday isn’t shedding its whole UI but other startups in HR have sort of embraced this chat interface as the default or, in some cases, only option. The move to position Sana as chat first might embolden them to push further.
I don’t think that’s wise.
Vedder wasn’t supposed to be in that studio. But the scene worked the way it did because people were in the same room, stumbling into each other’s orbits, and discovering things sideways.
The best interface might not be one that simply goes from taps or clicks to chats because it feels easier for everyone. Sometimes it’s the one that puts something you didn’t explicitly ask for in your path.
What else is happening
Workers Are Using AI to Sneak Out for Spin Classes and Skip Lunch Meetings. A Zoom-sponsored study conveniently finds that AI saves workers 30 minutes a day, which sounds pretty sweet. Bosses paying for AI productivity tools may be surprised to learn they’re also funding spin class memberships.
3 Charts on What CHROs Think About AI and the Future of Their Departments. Nice story from HR Dive showing that every CHRO says AI is their top priority, yet half haven’t figured out how to measure whether it’s actually doing anything. Apparently “AI strategy” for most organizations means “we’re aware AI exists and are very worried about it.”
HR’s Imperative: Invest in the Skills AI Can’t Touch. HR thought leaders are urging companies to finally invest in “human skills” like emotional intelligence and critical thinking, the same skills we’ve been systematically devaluing in favor of quantifiable metrics for the past two decades. Better late than never?
JPMorgan’s Jamie Dimon Says Remote Work Breeds ‘Rope-a-Dope Politics’ and Stunts Young Workers’ Growth A billionaire CEO insists young workers can only learn by watching him make mistakes in person, despite research showing remote workers are more productive and more engaged. The 1,200 JPMorgan employees who petitioned against the five-day mandate still seem unconvinced.
Report Reveals Widening AI Guidance Gap. Thomson Reuters finds that companies are enthusiastically pushing employees to adopt AI while having no coherent strategy for what that’s supposed to achieve. That’s fine, because only 22% of companies have any AI strategy at all. The bar for “ahead of the curve” remains comfortably low.
Major League Baseball’s Robot Umpires Are Here—and They Tell Us Something Uncomfortable About AI and Work. MLB introduced an AI challenge system for ball/strike calls, framed as a thoughtful human-AI collaboration. Given that nearly two-thirds of player ejections were over ball/strike disputes last year, human umpires weren’t exactly making a compelling case for their own indispensability.
Welcome to the Weird World of AI Agent Teams. Researchers discover that when you put a bunch of AI agents together without organizational structure, they spiral into chaos, a finding that will not surprise anyone who has never attended a corporate meeting. Coordination problems, it turns out, are not a uniquely human flaw.
Who Owns AI Agent Access? At Most Companies, Nobody Knows. 85% of companies have AI agents running in production, but nobody has agreed on who’s responsible for securing them. That seems like detail worth sorting out before something goes catastrophically wrong. With agents routinely over-privileged and borrowing human credentials, the next major breach has essentially been pre-scheduled.
Why AI Is a Massive Job-Creation Technology, Despite What You Think Josh Bersin argues AI creates more jobs than it destroys, pointing to rising software engineering salaries as proof. The “don’t worry, new jobs will appear… eventually” argument has a long and storied history of being technically correct and creatign short-term pain.
I just launched Beacon Turn
A little promotional news: I just officially launched Beacon Turn. If you’re reading this on my Substack, you know this.
You can read the full announcement on my LinkedIn and if you have any research, media, or advisory work, I’d love to chat about it.
Thanks and have a great rest of your week!