CRED is a platform where every enrichment action — CRM syncs, “Add to list,” automated runs — consumes credits. Users frequently faced unexpected spend, struggled to trace where costs came from, and lost confidence in the produ

Short summary

At CRED, I redesigned workflows around credit spending after discovering users were frequently surprised by hidden enrichment charges when adding to lists or syncing with CRM. I investigated through 6–8 stakeholder interviews, support ticket mining, funnel analysis, and session replays, which revealed visibility gaps, unclear triggers, and lack of traceability. I introduced a live credit estimator tied directly to CTAs, guardrails with warnings/approvals/daily caps, and a conditional review step for high-risk runs, along with clearer copy and accessible UI patterns. Post-action, I designed a job-centric Credit Activity view showing Estimate → Spent (Δ%), actor, approvals, timestamps, and error counts. This made every credit attributable, explainable, and exportable, reducing “unexpected spend” tickets, improving user prediction of cost, and restoring trust and budget control in high-stakes workflows.

The problem

• No visibility into when spend was triggered.

• No clear link between action and cost.

• Finance teams couldn’t answer “What did we pay for?” or “Who ran this?”.

• Support was flooded with tickets mentioning unexpected credits.

My role

I led the end-to-end redesign of credit workflows — from research and mapping mental models to designing new UI patterns and states. The goal: make credit spend predictable, explainable, and auditable.

Approach

Research & investigation

• User interviews with power users, admins, FinOps, new users.

• Session replays (FullStory/Hotjar) to identify hesitation and backtracks.

• Funnel and job log analysis to quantify fields × rows and outliers.

• Support ticket mining around surprise spend and tracking gaps.

• Flow review for clarity, error prevention, and accessibility.

Design principles

• Transparency — always show a numeric estimate before commit.

• Proximity — cost chips placed directly on CTAs, no hidden side effects.

• Trust through language — plain words (“Enrich”) instead of jargon (“Run waterfall”).

• Guardrails — approvals, daily caps, and hard stops built into the flow.

• Ownership — actor names, timestamps, approval trails, job IDs.

• AI UX clarity — translate complex AI enrichment logic into user-friendly options: Add without enrichment, Request approval, Schedule run.

Solution

Pre-action

• Live estimator (fields × records ≈ credits) shown before commit.

• Guardrails: warnings, approval request, hard stop if limits exceeded.

• Option to Add without enrichment for zero-risk flow.

Post-action

• Credit Activity redesigned as a job-centric audit view:

  • Estimate → Actual (Δ%)

  • Actor, timestamps, skipped/failed counts

  • Approvals trail and exportable reports

• Every credit now traceable, explainable, and attributable.

Results & impact

• Users can predict spend upfront with confidence.

• Finance teams gain a single source of truth for credit activity.

• Expected outcomes (validated with usability tests + support data):

  • −60% tickets tagged “unexpected spend.”

  • +90% of users correctly predict cost before running.

  • ≥90% of high-cost runs routed through approval.

Why this matters

Money is at stake. By transforming AI enrichment from a black box into a transparent, trustworthy interface, I helped CRED reduce friction, prevent errors, and restore user confidence.