Mixer Mode in the public sector — a thought experiment, not policy advice
Disclaimer up top, and I'm going to weight it heavily: I'm not a public servant and I don't make policy. This post is structural — about how the State operates as a cognitive system — not political. It isn't left, it isn't right, and if you read it as partisan you're reading something I didn't write.
With that on the table: if I look at digitizing the State through the Mixer Mode lens, something appears that I think explains why so many of these projects fail. The senior public official sitting in front of the decision isn't running a workflow. They're running six simultaneous channels the BPM system was never built to see.
Disclaimer First (With Extra Weight)
I'm not a public servant, not a constitutional lawyer, not an expert in administrative reform. I've never sat in the chair I'm about to describe. What I have done is watched, from the technology side, several State modernization waves go in and come out — and noticed a pattern that keeps showing up.
What follows is pedagogical speculation. A demonstration of how the framework travels into a domain that isn't mine, and where the risk of misdiagnosis is structural. I'm writing it because the lens seems useful, and useful lenses should be tested against the hardest cases, not the easiest.
The double caution is this: a misdiagnosed policy costs more than a misdiagnosed deploy. A bad deploy you roll back. A bad policy you live with for years, sometimes decades, sometimes through changes of administration that can't fully undo it. So if I'm going to bring a framework into this room, I owe the reader an unusually clear note that the room is bigger than the framework.
The Senior Public Official's Channels
Let me name what I think is actually running when a senior public official makes a real decision — the kind that isn't routine, where their judgment is doing work.
There's a legality channel: will this survive a constitutional challenge, an audit by the Comptroller, a judicial review? There's a political channel: what does the minister above say, what does the rival mayor say, what does the opposition turn this into next week? There's a budget channel — and it's not one number, it's three: line item, accrued execution, fiscal year cycle. There's a territorial impact channel: who benefits, who loses, where geographically, and what does that mean for the relationships with the local actors who have to implement it.
There's an electoral cycle channel: does this get announced before or after March, before or after the next election, before or after the next ministerial change. And there's a citizenry channel: is this understandable to a regular person, or does it feel like the State is abusing them with paperwork they can't decode.
Six channels. None of them appears in a BPM diagram. None of them is captured by "approved / rejected / forwarded". And yet the senior official is modulating all six at once on any decision that matters.
Why Digitization Fails When It Reduces Decision to Workflow
Classic BPM, the kind that has been the dominant pattern for thirty years of State modernization, is forced hat mode. It collapses the decision into a finite state machine with three or four outputs. Approve, reject, forward, request additional information. That's the universe the system can represent.
The senior official was operating six channels. The system asks them to collapse those six into three states. The compression is enormous and most of what they actually knew about the case — the political read, the territorial nuance, the citizenry-facing concern — falls outside the schema.
What we then label as "resistance to change" is, in my reading, involuntary channel suppression. The official isn't resisting digitization out of laziness or fear of technology. They're refusing to put their signature on a representation of the decision that, in their experience, leaves out the parts that actually carry the consequence. They knew something the system wouldn't let them express, and they were right to flag it.
The typical result is the one everyone has seen: a parallel workflow appears in Excel, in WhatsApp groups, in unofficial meetings before the official meeting. That parallel workflow is where the six channels do fit. The BPM is the theater of formal compliance; the real decision happens elsewhere. And then we wonder why the digitization didn't deliver the productivity it promised.
What Changes With Agentic + Well-Done Meta-Software
Here's where the framework actually has something useful to contribute, and I want to be careful about it.
Agents can absorb a large fraction of the routine production: template resolutions, standard certificates, responses to repeated requests, drafts of official letters that follow established formulas. That's real, and it's where the immediate efficiency lives.
The mixer-fluent senior official — the one who has learned to direct agents rather than be replaced by them — then holds the six channels over the non-routine cases. The cases where judgment matters. The cases where the political channel and the citizenry channel are pulling in different directions and the resolution requires someone who can hold the tension.
For this to be safe at the scale of a State, Meta-Software has to come with the agents. Not as a nice-to-have. As a precondition. That means governance codified at the policy level. Traceability that survives the next administration. Structural validation against the legal and procedural framework. Contextual continuity across the political turnover that breaks institutional memory every four to six years.
If Meta-Software is missing and the agents are deployed anyway, what we get is: the agent decides, no one audits at the right granularity, and when something goes wrong the scale of potential harm is citizen-wide. The 19% slowdown that METR observed in software teams becomes, in this context, a political crisis. The cost is asymmetric.
What Doesn't Translate
I want to name where the framework explicitly does not reach, because that's part of using it honestly.
Democratic legitimacy isn't the output of an optimizable loop. You can make a public decision faster, cheaper, more auditable — and if the population doesn't recognize it as theirs, you've made something efficient and illegitimate, which is worse than slow and legitimate.
The judgment of "when not to apply the rule" — the discretionary judgment that every senior public official exercises in the cases where strict application would produce an obviously wrong outcome — is pure Pillar 1, and it's politically loaded in a way the framework doesn't address. Mixer Mode names the cognitive simultaneity. It doesn't answer the question of who is accountable when the agent decides wrong, or how the citizen affected by that decision contests it.
Those are questions for constitutional and administrative law, not for a framework about cognition under agentic conditions. I'd rather say that out loud than pretend the framework reaches further than it does.
A Note on Chile Specifically
I live and work in Chile, so I'll close with a note grounded here.
We've had roughly thirty years of State modernization initiatives — Reforma del Estado, Modernización del Estado, multiple waves of digital government, ChileAtiende, ClaveÚnica, and on. Each wave has pushed workflows and each wave has hit the same rock. The rock, in my reading, has a name now: institutional channel suppression. The senior public servant knew the six channels mattered. The system insisted on the three states. The mismatch was structural, not personal.
The next wave — agentic in government, which is already starting — will repeat the mistake unless Pillar 1 gets named first. That means designing the deployment around the senior official's six channels, not around the BPM's three states.
Do you work or have you worked in the public sector? Which specific channel did you feel the system never let you operate? I'm interested in the concrete case, not the general complaint — calibrating the lens against real experience is what makes it useful.
#MixerMode #DigitalTransformation #FutureOfWork #TechLeadership