Game of Thrones both made and ruined fantasy TV. In its final seasons, it had offices blocking off lunches to bet on which characters would live. It captured the full attention of a nation. Every project that followed was forced by studios to deploy massive budgets. The Wheel of Time, The Witcher, The Rings of Power, George R.R. Martin's deeper IP spinoffs. Each had a huge marketing push, each had high expectations.
They all failed.
We're living through the Game of Thrones era of AI. And those days are likely already over.
Everyone in tech seems to think that automation will save the workforce from boring labor. In reality, the promise of full agentic automation is a lie.
It's seductive. On the surface, automation seems logical: work a little extra now, build a system and a process, and that system will bear the brunt of the labor in the future. It maps onto our intuition because this is the tradeoff many of us make every day: (Should I write a formula and copy this file into Excel to strip the quotes, or should I put on a podcast and sit here for 15 minutes and do it manually?)
But when you've worked at a company with lots of agents built during hackathons, weekends, and "personal dev time", what you find is that you've got a lot of overlap. Ambiguity emerges. Two agents from two departments, with two different pithy names, do functionally the same thing. One agent replicates data from another agent to customize it for another department. Five other downstream agents consume data from the replica. Competing initiatives try to standardize the "Bot Development Lifecycle". Multiple internal re-brands fight over what the "real front door agent to all other agents" should be named.
And suddenly, even though each commit has an owner attached to it, the ownership becomes obscured in the dependencies. We have more information about who did the work than we've ever had before, and yet there's still an accountability gap. We have very little information about who's actually accountable for the output of any single bot in that system.
A commenter paying expensive rent in San Francisco will say this is overly simplified. We're not naïve, they'll say, we all know that someone has to be accountable. It's about efficiency, after all, not complete automation. Only the grunt work will be automated.
Gartner thinks over 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls.
Do you know, right now, who's accountable at your company for the costs, value, or risks, including decision risk, of a specific agent? Is that person aware that they've drawn a short stick?
If you're really honest, is the stated owner accountable for anything but holding the bag in a crisis?
We live in a time where many processes will be at least partially automated by AI. The market demands it. Tech leaders are trying to recreate their exec's Game-of-Thrones-style honeymoon AI experience with their $1M/episode ultra-high-expectations retellings of SaaS' greatest hits.
Just like the over-funded successors to Game of Thrones, many of these initiatives will fail.
The reality that we're seeing on the front lines is that despite failures, parts of jobs will be hollowed out. Labor will look different. But as of right now, we're not seeing compelling evidence that there will be consistently less work for people. If GDP growth exceeds 3% while job growth declines, we can have that discussion. For now, my network in tech consistently reports pressure to work more than two years ago. Humans are in-demand, even if hiring rates are flagging.
A clear short-term play is to add accountability. Add transparency. Which agents exist? What data do they produce and consume? How is that governed? How are the human beings in the system impacted by these changes? Are processes still clunky? Is there an opportunity to make someone's job better?
The real work of the AI era is the opposite of automating humans into obsolescence: it's directly giving them accountability and governance of the changing territory. There aren't clean playbooks for this because we're writing them in 2026. Every business needs human owners of data, processes, and automations.
If you're struggling, we can help.
