Skip to manuscript
Ardavan Mir

Intuit Enterprise Suite · Enterprise SaaS

Evidence folio E.01 · Public-safe edition

AI-native workflows · Financial systems

Defining an AI-native target state for enterprise finance

I helped define and communicate AI-native product direction for complex enterprise finance workflows — translating ambiguity into clearer information architecture, trust patterns, prototype concepts, and leadership-ready product narratives.

Role
Product design, information architecture, prototyping, product storytelling
Focus
AI-native workflows, enterprise finance, trust, explainability, human judgment
Output
Frameworks, sanitized patterns, prototype narratives, alignment artifacts
Status
Public-safe case study

Stage 02 · Structure

The challenge was not “add AI.” It was make AI legible.

Enterprise finance workflows are dense, high-stakes, and full of dependencies. People need to understand what changed, why it matters, what evidence supports it, who needs to act, and what happens after an action is taken.

Context

As AI becomes more present in business software, the design challenge is not simply making systems more automated. The challenge is making intelligent systems understandable, trustworthy, and useful where human judgment still matters.

Design problem

A smart recommendation is only useful if people can evaluate it. The experience needed to make reasoning visible, clarify risk, show supporting evidence, and preserve user control before meaningful action.

Make the system legible before making it smart.

Stage 03 · Prototype

Four moves made the direction discussable.

My role was to translate complex enterprise finance concepts into experience models, information architecture, prototype concepts, and leadership-ready narratives.

01

Structure the system

Map the entities, relationships, decisions, constraints, and failure points before designing the interface.

02

Make evidence visible

Show the signals and source context behind a recommendation so people can understand why it matters.

03

Preserve human judgment

Treat approval, escalation, and review as core parts of the experience — not friction to remove.

04

Close the loop

After a decision, show what changed, what was recorded, and what can be reviewed later.

Framework contribution

From AI-enabled features to a governed enterprise system

AI-native could not mean adding an assistant to existing screens. We began with a simple diagnostic: if the AI disappeared, would the product still behave essentially the same way?

I helped synthesize recurring interaction decisions across product research, strategy literature, and customer evidence, then translate them into a framework for enterprise finance. The work connected intent, durable work objects, evidence, context, role, and governance into a shared model the team could build on.

Exhibit E.01c · AI-native diagnostic

The diagnostic separated feature augmentation from a product whose operating model depends on intelligence.

“If the AI disappeared, would the product still behave the same way?”
  1. 01AI added to interface

    The existing workflow still defines the product.

  2. 02AI shapes workflow

    Intelligence begins to reorganize how work moves.

  3. 03AI-native operating model

    Intent, evidence, governance, and action become one system.

Exhibit E.01d · Research to framework

We examined patterns across enterprise products, strategy research, and customer evidence without exposing confidential counts or source material.

Stage 04 · Evidence

Patterns leave the design logic visible.

The public evidence is intentionally abstracted: generic labels, synthetic diagrams, and the decision architecture behind trust—never confidential product UI.

Reusable interaction framework

Five elements made the operating model concrete.

A governed enterprise system did not mean giving AI unrestricted control. It meant creating an environment where AI could work across complex financial contexts while keeping its evidence, permissions, consequences, and moments of human judgment visible.

F.01

Intent

Begin with the outcome and let the system compose a path through the work.

F.02

Environment

AI operates through a scoped product environment rather than living in a separate assistant.

F.03

Artifact

Consequential work persists in a governed object shared across roles—not in a transient conversation.

F.04

Evidence

Claims remain connected to their sources, freshness, policy context, and review state.

F.05

Gate

Authorization becomes available only after scope, evidence, policy, and expected impact are understood.

The proposal gave the team a shared language and reusable foundation for subsequent AI-native exploration.
Exhibit E.01e · Scoped environment and governed object

The shell is the persistent environment through which AI acts. It supplies scope, context, evidence, roles, and permissions; it is not the AI itself.

Exhibit E.01f · Governed decision lifecycle

Authorization is not the starting point. It becomes available only after the system can show what is intended, supported, permitted, and expected to change.

  1. EstablishedIntent

    Outcome and scope recorded

  2. EstablishedEvidence

    Sources and freshness visible

  3. EstablishedPolicy

    Constraints checked

  4. EstablishedImpact

    Expected consequence reviewed

  5. ConditionedAuthorization

    Available after prior conditions

  6. After actionReceipt

    Decision and change recorded

  • Scope known
  • Evidence current
  • Policy satisfied
  • Impact reviewed
Exhibit E.01g · One object, role-adapted views

The underlying evidence, state, and history remain shared while density, available actions, and permissions adapt to the role.

Shared evidence
Sources, freshness, policy context
Shared state
Review ready · action pending
Shared history
Proposal, review, and decision record
Finance leader

Decision-level view

Material impact, exceptions, policy posture, and authorization.

Finance operator

Preparation-level view

Source detail, open evidence tasks, resolution options, and escalation.

Design evolution

Early concepts explored visual confidence tiers. As the framework matured, the emphasis moved toward explicit evidence, freshness, policy, review, and authorization states—signals people could interpret and act on more reliably.

P.01

Evidence rail

A supporting layer that gives users a clear trail of signals, assumptions, and source context.

P.02

Decision packet

A structured object that brings together the recommendation, rationale, impact, risks, and next step.

P.03

Human judgment checkpoint

A deliberate moment where the system asks for review, confirmation, or escalation before meaningful action.

P.04

Approval receipt

A closing artifact that records what was approved, what changed, and where the user can review it later.

Exhibit E.01a · Trust-pattern studies

Complexity → clarity system map

A simplified model for turning ambiguous financial state into inspectable action.

01State02Signal03Evidence04Decision05Action
Human judgment checkpoint

A deliberate pause before meaningful action, preserving review and control.

Evidence rail

Supporting context that helps users evaluate why a recommendation matters.

  • Source contextAttached
  • FreshnessCurrent
  • Policy checkSatisfied
  • Review stateHuman review due
Decision packet

A structured object that brings together recommendation, rationale, risk, and next step.

Recommendation
Review classification variance
Evidence
Supporting sources attached
Impact
Material change — review required
Human review
Pending approval
Approval state
Awaiting decision
Approval receipt

A closing artifact that records what changed and where it can be reviewed.

Decision recorded
Classification review approved
Reviewer
Finance lead
Time
Synthetic timestamp
Follow-up
Audit trail available

Withheld / NDA materialInternal screenshots, roadmap details, and confidential demo material are intentionally omitted.

Stage 05 · Story

Future direction became something teams could evaluate.

Rather than treating AI as a layer of automation, the exploration focused on how financial workflows could become more legible through structure, evidence, controls, and human judgment.

The public version focuses on the design approach and storytelling patterns rather than internal prototypes or roadmap details.

S.01

Framework transfer

Translated core AI-native experience concepts into reusable patterns that could be explored across related enterprise product contexts.

S.02

Prototype as alignment tool

Used functional prototypes and polished artifacts to make abstract product direction easier for teams to understand, critique, and refine.

S.03

Leadership-ready narrative

Turned complex capabilities into a clearer story: what the system understands, what evidence supports it, where human judgment matters, and what happens next.

S.04

Trust before automation

Kept the emphasis on explainability, review, authorization, and control before any meaningful action could be taken.

Exhibit E.01b · From ambiguity to alignment

Framework transfer map

A sanitized model for translating a core AI-native concept into adjacent enterprise contexts.

Prototype-to-storyboard arc

How ambiguous product direction becomes a prototype and then a clearer leadership narrative.

01Ambiguity02Structure03Prototype04Narrative05Alignment
Leadership alignment map

A simplified model for bringing evidence, prototype, story, and decision-making into one frame.

Future-state exploration model

A public-safe abstraction of how state, evidence, human judgment, and action can work together.

01System state02Signals03Evidence04Human review05Action path