TAKANOBU SAITO / PORTFOLIO Back to the central hub ↗

03 / Capabilities

Think. Build. Put it to work. Verify.

I connect UX, business, AI, and implementation to move one problem forward, rather than treating them as separate specialties.

I show a connected practice grounded in real cases, rather than relying on a long list of titles.

01 / USERUX design
02 / BUSINESSMarketing
03 / CONTROLAI operations
04 / SYSTEMImplementation and data
DECISION

Scope, connected

I work across the full experience, before and after the screen.

From the initial conversation, I identify where people are uncertain, establish decision criteria, translate them into an interface or system, then run and validate it. When needed, I also design operational exceptions and points for human approval.

01

UX / Product design

Solve friction by rethinking the sequence.

Do not dismiss “I don’t understand” as a lack of information; treat it as being unable to choose the next step.

Connect user understanding, problem definition, information architecture, wireframes, prototypes, and validation in one process.

  • Turn uncertainty, hesitation, and comparison effort into a clear sequence of actions.
  • Decide what people need to understand first before adding more information.
  • Document the evidence behind decisions and what remains unverified, not just the finished screens.
Understanding users in a pet-care UX projectDesigning paths through a career assessment
02

Marketing / EC

Turn sales metrics back into choices people can make.

View acquisition, comparison, purchase, and retention through both user decisions and business KPIs.

Clarify product definitions and purchase conditions, then connect changes in sales, orders, CVR, ROAS, and CPA to the next action to investigate.

  • Standardize similar products for comparison and make pricing assumptions explicit.
  • Bring multiple KPIs together and make the priority of changes visible.
  • Avoid asserting a cause; point to inventory, ads, product pages, or other places for people to investigate.
Defining products for MTGCriteria for evaluating keyword performance
03

AI product / Operations

As AI does more work, people focus on exceptions.

Design decision points and evidence for human accountability before deciding how much to automate.

Manage project status, next tasks, pending approvals, blockers, and deliverables for multiple AI workflows in one operations view.

  • Define what AI may proceed with automatically and where a person must intervene.
  • Protect external messages, purchases, publishing, and production updates with approval gates.
  • Keep a continuous record from execution and evidence through learnings and the next task.
AI project boardDesigning human approval
04

Data / Implementation

Go beyond a proposal and build something people can use and verify.

Implement the necessary screens and calculations, connecting inputs, display, sharing, and ongoing operations.

Using HTML, CSS, and JavaScript, I build browser-based tools and dashboards with AI as an implementation partner.

  • Import CSV files and recalculate metrics with consistent definitions.
  • Build search, filtering, comparison, and timeline views into working screens.
  • Account for operational requirements in the interface, such as keeping user data on-device.
Designing Excel data inputsProduct search and filtering

One continuous responsibility

Connect disciplines to reduce rework.

When research, interface design, and implementation are separated, the reasons behind decisions get lost. I connect the initial question to validation and document the process so others can follow it.

  1. Observe the ambiguityDo not turn requests straight into features; find out who is getting stuck and where.
  2. Set decision criteriaBefore generating options, define what each needs to satisfy to move forward.
  3. Translate into information and interfacesOrder information by what users need to know next and make the options comparable.
  4. Build something that worksValidate interactions, inputs, exceptions, and responsive behavior that static images cannot show.
  5. Separate and document the resultsKeep verified findings, hypotheses, and unknowns distinct for the next decision.

What stays constant

Keep the same standards across disciplines.

The standards are clear user decisions, explainable implementation, and a clear distinction between verified facts and what remains untested.

01

Define the uncertainty before designing the interface.

Do not optimize appearance for its own sake; clarify why people are stuck and what they should do next.

02

Set the stop points before automating.

Prioritize the boundary of human responsibility over the speed of AI or computation.

03

Make it verifiable before calling it complete.

Build something people can operate, then document what was verified and what still needs checking.