ADSGENCY AI X BUDGET FORECASTING

Enabling marketers to make confident, data backed budget decisions before spending by turning complex performance data into clear, scenario driven insights.

70%

Paid clients onboarded

faster budget decisions

↑ High

Planner confidence

Role

Lead Product Designer

Timeline

2025-Present

Team

Design, PM, Engineering

UX Designer, Sr Designer, Founder, PM

Deliverables

Flows, HiFi mockups, Component Specs, Handoff

CONTEXT

The interface lacked perceived credibility and system trust, making it feel insufficiently reliable for high stakes decision making.

How might we design a structured and inclusive support system that meets the diverse needs of students, enabling them to navigate the complexities of career decisions seamlessly and make timely, informed choices?

We realized users didn’t feel in control or confident in their inputs

I saw that the outputs weren’t helping users make clear decisions

We identified that unclear system logic was breaking user trust

We realized users didn’t feel in control or confident in their inputs

I saw that the outputs weren’t helping users make clear decisions

We identified that unclear system logic was breaking user trust

We realized users didn’t feel in control or confident in their inputs

The chart outputs weren’t helping users make clear decisions due to its interpretability constraints

We identified that unclear system logic was breaking user trust

OPPORTUNITY

AdsGency needed forecasting to drive decisions, not just display data. Only 4% active use among total paid plan users

Our goal wasn't to show numbers. It was to give marketing teams enough confidence to commit budget inside the product, and guide users with next steps. Without that, the tool had no real value in the planning workflow.

Performance marketers planning $10k–$100k+ monthly ad budgets across analytics platforms

Users

Core jobs to be done

Simulate budget allocation, see projected outcomes, and commit to a plan, without leaving the product.

USER FLOW

What if your forecasting tool didn't just show you what was coming- but told you what to do about it? That's the shift I designed, and I worked with Claude.ai to poke holes in the logic before a single line of code was written.

SOLUTION

Predictive clarity from first input to final commit, build with figma's MCP and Claude Code

FLOW 1

Focused budget input with system status visibility interaction

Our goal wasn't to show numbers. It was to give marketing teams enough confidence to commit budget inside the product, and guide users with next steps. This feature surfaces just enough of what goes into defining the budget, giving users visibility into the system without overwhelming them, so they can understand the logic, build trust, and feel confident committing spend within the product.

FLOW 2

Progressive disclosure through categorized insight cards

In conversations with the team, we aligned on one thing — the output needed structure. Not everything the AI surfaces carries the same weight, so we prioritized separating insights by category. I took that decision into the design, translating it into categorized cards that give each insight type its own space.

I took that decision into the design, translating it into categorized cards that give each insight type its own space.

FLOW 3

People wouldn't commit budget to a tool that could overwrite it

Budget locking freezes an approved baseline while everything unlocked stays explorable, so AI can re-optimize the rest without touching what you've committed. I built it from existing chart components, carrying the channel colors through.

FLOW 4

Skeleton screen and error states

Users used to land on a blank screen with no sense of what came next. I designed a skeleton state that previews the structure before data loads, so people know what they're about to interact with. And if the tool errors out, you pick up where you left off with nothing lost.

FLOW 5

Redesigning analytics charts for clearer multi-variable readability

Based on backend capacity, we designed the forecasting charts such that users can scenario-plan, drag to test budgets, see the per-platform breakdown, and get an AI nudge on where to shift spend

TESTING AND ITERATION

Testing and iteration cycles through prototyping and prompting, with human eye for design detail

Early passes surfaced two fixes. The design system purple was barely clearing contrast standards, so I built a Claude skill to run contrast checks and swapped in a compliant color and made standardized changes across every feature and back into the design system.

IMPACT

"This is giving our users the confidence to make decisions inside the product instead of taking the data elsewhere."

RETROSPECTIVE

Working on developer AI tools made me a sharper prompter, and that carries straight into how I design

Claude and Cursor let me wireframe fast, pressure-test flows, and generate prototype code for handoff. Speed isn't really the point. The work stays grounded in user needs, and that clarity is what makes it fast.