Ananthu S · Lead UX & Product Designer

Enterprise AI,designed to betrusted.

I design AI agents, assistants and data platforms that people can check, overrule and undo.

See the work
AvailableOpen to senior & lead rolesLet’s talk 3 wks → 2–3 daysClient data-source onboarding with Studio’s AI agents21 → 6Clicks on the platform’s most time-consuming routes after the v2.0 redesign
NowLead UX Designer, Prevalent AIDesigning enterprise products since 2019

About

I’m Ananthu, a lead product designer building enterprise AI that people can check, overrule and undo: agents, assistants and the data platforms beneath them.

Clarity first,
trust by design,
built to ship.

I lead UX at Prevalent AI, designing enterprise SaaS products with AI at the core: agents that do the heavy lifting while people keep the final say, an assistant that answers with its sources, and the data platform underneath them. The problem I keep coming back to: people making high-stakes decisions on systems they can’t fully see.

More about me

Focused on people.
Measured in outcomes.

✦ From ambiguity to decision.

Key facts

A snapshot of the work
so far.

Experience

  1. Nov 2020 – nowLead UX DesignerPrevalent AI
  2. Sep – Nov 2020UX DesignerTuna Software Solutions
  3. 2019 – Sep 2020UI/UX DesignerCrypt4Bits

Selected work
& case studies

Ask about my work

01Nov 2025 – Jan 2026

Studio

AI agents for data onboarding. Let AI agents do the work in a way people can check, overrule and undo.

3 wks → 2–3 days to onboard a new client's data

Explore project : Studio

02Sep – Oct 2025

Navigator

Agentic Search. Make an AI answer trustworthy enough to repeat to a board, and honest when it can't answer.

Live with customers, who now want it as the main way in

Explore project : Navigator

03Dec 2024 – Jan 2025

Exposure Management

Platform. Make a risk score explainable, and connected data readable, to people used to lists.

21 → 6 clicks on the most time-consuming routes

Explore project : Exposure Management

How I design AI

Show the reasoning along with the result.

Every AI output carries its why (the sources, relationships or rule behind it) at the point of decision, rather than in a help article.

Make certainty legible. Design the “I don’t know.”

Confidence is a number with consequences. Thresholds route uncertain work to people, and out-of-scope answers say so plainly.

Automation proposes. People decide.

Every automated action is attributable, reviewable before it lands and reversible after. Autonomy is granted step by step as it is earned.

Every fact knows where it came from.

Source, author (human or agent), time and confidence travel with data into every view, so people can judge information they did not produce.

Assistant · answer with sources

Why did data onboarding slow down this month?

Three new data sources1 needed manual field mapping2, and one source changed its format3 mid-month.

Based on 3 sources · point at one to see what it supports

Agent output · how sure it is

  • Customer name → Account97%
  • Region → Territory92%
  • Plan → Tier86%
  • Owner → Person74%
  • Cost centre → Team61%
  • Notes → ?38%

Agent suggestion · waiting for you

Matching agent suggests

Treat “Owner” as a person, not a team

Matches a work email for 1,092 of 1,204 records · 74% sure

Nothing changes until you decide.

    Record history

    Account tierEnterprise

    1. ImportedFrom the CRM sync · 09:12
    2. Matched by an agentMatching agent · 92% sure
    3. ApprovedBy you · 09:40 · can be undone
    4. Shown everywhere with this historyDashboards, assistant answers, exports