Return the Controls — AI Operations Secretary Tasukiba's UI Without a Gatekeeper

The one sentence behind the design

Tasukiba’s UI is held together by one belief.

Humans are satisfied when they have agency.

A vendor playing gatekeeper sits frontally against that. Tasukiba walks the other way.

Concretely:

  • Plan changes, cancellation, storage add-on downgrades
  • Member adds, role changes, leaving
  • Data export (CSV / Markdown / Excel)

All of these are UI that completes without vendor round-trips. Today, the closing post of Chapter B, is about that.

👉 Tasukiba product page

The friction trap in typical business SaaS

Many business SaaS deliberately add friction to cancellation and shrink operations.

  • Bury the cancel button in a deep menu
  • “Please pick 10 reasons for cancellation” — a psychological wall
  • “Please contact support” — an intermediate step

The intent is to buy time to win the user back. Business-rationally, it’s not wrong.

From the user’s side, though, this is a design that takes away freedom. The moment you decide “I want to leave here,” the service catches your sleeve. It leaves a bad taste. Tasukiba doesn’t want to leave that taste.

This is essentially the same idea as The Courage to Be Disliked, Part 2 — separating tasks. Whether the user keeps using the service is the user’s task — not mine. My task is only: provide value that makes them want to keep using it.

”Don’t block who comes in, don’t block who leaves” — implemented

The screen-level reality:

Plan change

In the tenant settings panel, the “Plan” section:

  1. Current plan (e.g. Beginner)
  2. “Change to Expert” / “Change to Pro” buttons
  3. A “confirm change” dialog with pricing preview
  4. Immediate switch

No vendor approval. Tenant admin’s authority is enough.

Cancellation

At the bottom of the same panel:

  1. “Cancel tenant” button
  2. Confirmation dialog (“all of X’s data will be deleted”)
  3. Immediate cancellation

No “pick a reason” intermediate. An optional “send feedback” link sits beside it — sending feedback isn’t required.

Storage add-on downgrade

If you’ve expanded storage:

  1. Current capacity and usage shown
  2. Slider to pick new capacity
  3. “Next invoice reduced by ¥xx” in real time
  4. Confirm — immediate

No “please consult support before downgrading.” Just the slider.

Export — “your data is yours”

Tasukiba lets you take your data out, anytime.

EntityFormat
KnowledgeMarkdown, bulk download
Risks / IssuesCSV export
RetrospectivesCSV / Markdown
MemosMarkdown
ProjectsCSV (with tasks)

This preserves a state where you can leave Tasukiba at any time.

When your data feels like a hostage, you can’t relax into the service. When you can take it out at any time, you can relax — and keep using it.

“Earn trust by guaranteeing freedom” — straight line back to the “non-dependent distance” I wrote about in series part 2.

Operational rules for not playing gatekeeper

As the operator, I run things so that “please stop here” isn’t a thing I say.

  • Plan-change pre-approval → not required
  • Cancellation reason interview → optional
  • Member-add cap → configurable, but at the tenant admin’s discretion
  • Role-change vendor approval → not required

Exceptions: legal compliance / security violation response only. Those are exceptions for user safety, not for vendor convenience.

The one exception: blocking self-role downgrade

There’s exactly one place user freedom is restricted.

An admin cannot lower their own role.

This is the safety net for “what if the last admin leaves the tenant” (ADR-0014). Role changes get audit-logged, and only another admin can change a role.

This isn’t about “restricting users.” It’s about “preventing operational accidents.”

When such an exception exists, the UI states the reason in plain language.

“You cannot change your own role. Please ask another admin.”

The restriction stops feeling arbitrary. And: when Tasukiba says “no,” it always offers an alternative path alongside the no. That’s a rule.

”UI = API authorization equivalence”

Slightly technical, but Tasukiba holds the principle:

A button you can’t click in the UI is a request the API won’t accept.

ADR-0014. Concretely:

  • UI disables a button → simultaneously the API returns 403 Forbidden
  • UI hides a menu → simultaneously the API path is inaccessible

This is the structural enforcement of “what you see equals what you can do.”

The experience of “the button was there, I clicked it, and I got a 403” is the UI lying. Tasukiba refuses to let the UI lie.

One line

Tasukiba’s UI design compresses to one line: return autonomy to the user.

  • Plan change / cancellation / downgrade → self-service
  • Data export → anytime
  • Self-role-change block → exception, with the reason stated
  • UI = API authorization equivalence → the UI doesn’t lie

Some of this looks like “extra work for the operator with no upside.” It did take a lot of implementation time.

But what’s being handed to the user is freedom. And, as The Courage to Be Disliked, Part 3 — here and now puts it, freedom is the ground that trust grows on. Short-term it’s a sub-optimal choice. Long-term, I believe it’s the only shape that keeps going.

Chapter B done

Four Chapter B posts complete:

Chapter K next — community, hiring, culture. Tomorrow: “empathy as the hiring filter” — gathering people by values, not by technical skill.

About Tasukiba

Tasukiba Knowledge Relay’s UI is built on the conviction that autonomy belongs with the user. See the product page for actual screens and self-service flows.

Contact

Feel free to reach out with any questions or feedback.

Get in touch
Tasukiba — The AI Operations Secretary for Knowledge & Project Management