360 Finance / UX + Process + AI

From financial signals to confident human decisions.

A connected finance workspace where explainable AI helps teams investigate, and people approve, execute and verify every consequential action.

Read Case Study ↓Live Product ↗
360 Finance project cover
01 / Problem

Finance needs more than another dashboard.

Enterprise finance spans performance, cash, source records, reconciliation, reporting and close. A discrepancy discovered in a KPI can become an investigation, a correction, a review and a close dependency.

360 Finance connects those activities so a financial signal retains its path to source evidence, a human decision and a verified result.

How might we reduce investigation work without reducing human judgment?

02 / Research

Separate known evidence from hypotheses.

The work began with the supplied product brief, the Finance 360 reference and direct walkthroughs of the implemented prototype. These supported hypotheses about leadership needing orientation before detail, disconnected evidence creating investigation work, confidence being mistaken for certainty and completed jobs concealing unresolved dependencies.

Commissioned interviews and measured usability outcomes were not part of this project. The role models and problem framing establish what needs to be tested with finance practitioners.

02 / Research

One context, three responsibilities.

CFOs need a clear view of performance, liquidity and decisions requiring accountability. Controllers inspect discrepancies, supporting records and dependencies before certifying readiness. Accountants search records, trace sources and prepare an explicit handoff to a reviewer.

The role models come from the brief. The working demo illustrates distinct preparation and approval boundaries.

Three User Responsibilities

CFO

Direction & accountability

Are we on plan, and what needs my decision?

Understand performance and liquidity, then approve a defensible next step.

Headline KPIs with a useful comparison

Forecast assumptions and financial consequences

Evidence and reviewer context before approval

A summary can look reassuring while hiding a material exception. Switching to another report loses the decision context.

Overview → cash or performance → evidence → approval

A forecast is a projection. A recommendation is a proposal. Neither is a completed financial action.

Controller

Accuracy & readiness

What is blocking close, and can I trust the fix?

Resolve discrepancies, inspect supporting records and verify the outcome before certifying readiness.

Exceptions prioritized by impact and severity

Ledger and source evidence together

Visible dependencies, reviewer notes and status history

A job marked complete is not proof that every source record reconciled. Unresolved dependencies can still block the close.

Close blocker → exception → source records → review → verification

Approval and execution are separate states. A successful job still needs its result checked.

Accountant

Investigation & preparation

Which records explain this difference, and what should I prepare?

Find the relevant line items, establish a supported explanation and prepare a clear proposal for a reviewer.

Search, filters and record-level source lineage

Persistent paths back to the original finding

Ownership and an explicit review handoff

A variance without a source path creates another round of exports, manual lookups and questions to colleagues.

Finding → transaction → evidence → draft → submit for review

The accountant persona can investigate and prepare. The demo denies approval and execution for this role.

03 / Process

Organize around the decision.

Shared period, entity, currency and scenario context carries an investigation across 22 connected destinations. The experience moves from a prioritized signal to explanation, source records, a concrete proposal, human review, separate execution and verification.

One complete investigation follows Revenue → Europe → Germany → Enterprise → Product A → ABC Corporation → INV-4821, with breadcrumbs preserving the way back.

Summary first. Evidence stays close. States mean something. Leave a trace.

04 / Architecture

AI prepares. Humans decide. The result gets checked.

A representative bank-match journey connects invoice INV-4821 with bank record BNK-99821, both showing $125,420. The explanation exposes the matching amount, counterparty and settlement date alongside illustrative confidence.

A Controller or CFO reviews the evidence and records a note before approving. Execution is a separate simulated step; verification checks the result and retains the decision history.

The seven stages connect Detect → Explain → Investigate → Propose → Review → Execute → Verify.

The Seven-Stage Decision Journey

Make the exception visible before it becomes a task.

Detect

System signal

The bank reconciliation finding identifies an unmatched $125,420 entry. Severity, impact and ownership help the reviewer choose what to investigate first.

EX-003 · unmatched bank transaction · representative Q3 data

Surface the trigger and affected records. Detection must not silently change a financial record.

Give a reason the user can inspect.

Explain

AI assistance

The sample explanation connects the missing invoice prefix with matching amount, counterparty and settlement date. The confidence indicator sits beside the evidence rather than replacing it.

INV-4821 · BNK-99821 · same amount and counterparty

Show supporting facts, the scope checked and limitations. A confidence percentage is not approval.

Keep both sides of the match in view.

Investigate

Human judgment

The reviewer moves into reconciliation, compares the invoice with the bank entry and follows source lineage. The original finding remains connected to the proposed resolution.

Two-sided record comparison · $125,420 on each side

Expose the source path and preserve context. The reviewer needs a way to disagree with the recommendation.

Turn an explanation into a reviewable draft.

Propose

AI assistance

The proposed match records the affected identifiers, amount and intended change. Submission places it in the approval queue; preparing a draft does not execute it.

Draft proposal → pending approval · linked finding and records

Make the action scope concrete. Editing an approved proposal clears approval and requires another review.

Require an explicit decision and a reason.

Review

Human control

A Controller or CFO reviews the evidence, records a note, and approves or rejects the proposal. The reviewer identity and decision become part of the local history.

Reviewer persona · required note · approved or rejected state

No default approval. No execution from an unapproved proposal. Demo personas illustrate the policy, not authenticated identity.

Make the consequential step deliberate.

Execute

Separate action

A separate execution control becomes available only after approval. The prototype simulates the match and records the result. Repeat execution and changes to a locked period are denied.

Approved proposal → simulated result · repeat-execution guard

Execution needs its own control and visible feedback. The prototype never moves money or posts to an external ledger.

Close the loop with evidence of the outcome.

Verify

Result check

Verification checks the execution state and marks the finding resolved. The history preserves preparation, review, execution and verification so someone else can reconstruct the decision.

Verified action · resolved finding · connected local audit events

Done must mean the result was checked. The local history demonstrates traceability, not an immutable compliance log.

04 / Architecture

Trust lives in difficult states.

Unapproved proposals cannot execute. Rejection retains the reviewer’s reason. Editing an approved proposal clears that approval and requires resubmission. Accountant personas cannot approve or execute.

A partly recovered import still requires three manual source matches. A CFO close action checks verified dependencies and locks subsequent financial changes.

These states make authority and incomplete work visible. Preparation, approval, execution and verification remain separate, inspectable steps.

05 / Product

Make financial meaning legible.

Four headline KPIs establish the overview before deeper metrics and records. Comparison badges name their baselines. Cost overruns use unfavorable status even when their numeric variance is positive.

Planning keeps growth, margin and capex assumptions attached to future projections. The visual system uses Poppins, white surfaces, navy text, blue actions and semantic status colors.

Explore the actual screens below, then open the product library to try the connected workflows.

05 / Product

Preserve context when the screen gets smaller.

At 390px, the checked overview, planning and cost-center pages preserve context and contain wide tables within their cards. Navigation moves into a mobile sidebar while the reporting period, entity, currency and scenario remain available.

Working prototype and planned validation. No measured business impact or commissioned user interviews are claimed.

06 / Results

A working outcome with clear next steps.

The result is an inspectable prototype with 22 destinations, three responsibility models and seven decision stages. Recorded checks covered the build, TypeScript, domain and recovery behavior, desktop workflows and bounded mobile layouts.

The next step is moderated research with finance practitioners to test evidence inspection, state comprehension and appropriate skepticism toward AI before optimizing speed.

Production work would replace local personas and browser persistence with authenticated users, server authorization and durable records. No measured business outcomes or usability scores are claimed.

07 / Live Product

Explore the live product.

Explore the product library: financial overview, planning, AI investigation, human review, and verification. Sample financial data, deterministic AI responses, and simulated actions make the prototype safe to explore.

LET’S TALK

Let’s design incredible
work together.

I'm open to Senior Product Design roles in Dubai — AI platforms, enterprise products, and agentic experiences.

Book a Call ↗Email Me ↗