Designing the brain that runs a B2B marketing team's pipeline Abmatic AI

From early product foundation to designing the trust and control systems behind an agentic AI platform for B2B account-based marketing.

Abmatic AI replaces the patchwork of tools B2B marketing teams normally stitch together into one platform for accounts, contacts, campaigns, and attribution, built around a single autonomous agent that plans, launches, and optimizes an entire ABM motion on its own. I started working directly with the founding team to shape the product's earliest experience, and stayed on full-time through its evolution into an agentic AI company.

Over two years, my work spanned brand strategy, the design system, information architecture, and the core flows that let both technical and non-technical marketers set up, monitor, and trust an AI system acting on their behalf.

ROLE

COMPANY

DURATION

SCOPE

UX Contractor → AI Product Designer

Abmatic AI, San Francisco

2 years

Brand strategy, Design system, IA, Agentic UX, Conversational AI

View live abmatic.ai

The Industry problem + Competitive landscape

When I joined, the product was a pre-MVP SaaS tool with a handful of base flows such as accounts, contacts, a few campaign types, analytics and more. Before touching the interface, I spent time understanding the industry itself: who the target audience was, what the founder's vision was, and what was structurally broken in how B2B marketing teams already worked.

The core issue was fragmentation. Account-based marketing teams ran intent data, personalization, paid media, and outbound sequencing across separate, disconnected tools and every hand-off between those tools lost context. I distilled that into five concrete, benchmarkable problems:

Intent data, personalization, advertising, and outbound sequencing lived in separate, completely disconnected marketing platforms entirely.

Every single campaign still required manually assembling audience, creative, and personalization entirely by hand.

Teams simply had no single source of truth for real-time account status today.

By the time data was finally consolidated, the buying window had already shifted.

SDRs, demand gen, and ABM teams always worked from entirely different, siloed data sets.

Engagement remained reactive rather than predictive, as teams simply responded to incoming signals passively.

Competitive benchmarking was a continuous practice at kickoff, to inform the first design system, and throughout, to keep the product's interaction and visual language ahead of the category.

The existing ABM and intent-data platforms in the market were largely dashboard-led and rule-based: they surfaced signal, but still required a human to act on it.

That gap, tools that report versus a system that executes, is where I saw the clearest opening for design to differentiate the product.

The pivot

Exploring agentic AI started as an experiment and became a strategic shift. Instead of treating AI as a supporting feature, we first explored mirroring team roles technically unified on the backend, but still fragmented to users before redesigning the product around autonomous agents that could handle complex B2B marketing workflows end to end. That decision led to a complete rethink of the product experience, and a full rebrand into an AI-agent-focused company.

Platform Flow

The end-to-end path a new user takes, from the first visit to the website through to the Agentic AI Clara running their campaigns.

A new account starts genuinely blank. There's no data to show until the client installs a tracking snippet on their own site. That installation step is deliberately the very first thing a user does, before anything else is unlocked, because every downstream feature depends on it.

Flow 1

Sign in & walkthrough

A first look at the platform from the moment you log in, how it introduces itself before you touch a single feature.

Feature 2

Installation

How fast the platform goes from zero to live and set up and running in minutes just via a single snippet.

Flow 3

Meet Clara
The Agentic AI

Once installed, the fastest path into value is the agent itself surfacing what's already known about the account base.

Clara does two jobs — greets strangers on the public site and runs a marketer's campaigns, accounts, contacts, and analytics behind the scenes — and had to feel like one consistent personality doing both.

builds the account plan from intent + firmographics

ships personalization, ads & sequences in minutes

re-allocates spend toward what converts, hourly

Flow 4

Dashboard, accounts & contacts

The command center where every account, contact, and buying signal lives in one connected view.

Flow 5

Campaigns
Feature system

Every new client pulled the product in a new direction. Rather than bolt features on, I designed a system that could mainly make the platfrom self explanatory, optimized visiblity for accounts and contacts, absorb 12 campaign types across three categories without fragmenting the IA, and provide automatic analytics and more.

Inbound, Outbound, A/B testing, and On-site widgets

OpenAI, LinkedIn, Display and Retargeting,

Run by the agent's bidder

Sequences, Intent detection, Agentic flow, and Agentic Chat

Flow 6

Automations,
Conversions, Visitor Reveal, Analytics

Watch the agent work: WHEN → IF → THEN playbooks running live, with zero manual input.
Tracking the moment a visitor becomes a real, attributable pipeline outcome.
The numbers behind every campaign, account, and channel read in plain language.
De-anonymizing site traffic in real time turning an anonymous visit into a named account.

Designing trust: transparency & control

ABM workflows were already complex; introducing autonomous systems added new questions about trust, transparency and control. A large part of my role was translating agent behavior into something simple and predictable for non-technical users and defending that clarity in the room.

I worked closely with the founders and with frontend and backend engineers to define how AI behavior should surface in the interface. Most of the real design debate wasn't about what the agent could do but it was about how much of that should be visible to the user, and at exactly which moments a user should be able to step in and intervene. Those weren't visual decisions; they were shaped by real technical constraints, which pushed me to think beyond static screens and design for system behavior and edge cases.

The hard question was never "what can the agent do", it was how much of that to show, and when to let the user take the wheel.

Visuals, design system & marketing identity

The visual system evolved in three real stages and I owned creative direction through all of them.

My design
process and
how it evolved

The tools changed, but more importantly, how close I sat to actual shipped code changed and that shift is the part of my growth I'm most proud of.

I inherited early designs in Adobe XD and migrated the team to Figma for better component reuse and a tighter developer hand-off. From there, the process became collaborative in real time: sketching and iterating directly with developers on calls, and using FigJam to work through flows together rather than throwing finished screens over a wall.

Studying competitors was part of the ongoing process, not just the research phase. Benchmarking how other products solved similar interaction problems kept our visual language from going stale. When I noticed the product had grown inconsistent page to page as features shipped fast, I unified it into one coherent system, which was necessary groundwork for scaling.

The biggest shift was learning to deploy. I started using Claude to translate verified Figma designs into working UI changes, test them locally with the developer, and only then ship live. That took me from handing off static documentation, as I'd learned in an earlier internship, to pushing and verifying changes myself. Getting fluent in GitHub and deployment wasn't part of my job description, but it closed a gap I used to just document for someone else to solve.

Abmatic AI makes our marketing team smarter and our sales team faster. It gives us real-time intent signals and personalized messaging tools that have drastically improved our targeting capabilities and conversion rates. The results speak for themselves.

Stéphane Le Mentec

Head of Demand Generation · Ketch

Impact

Following the launch of the AI-agent-based platform, the company onboarded multiple enterprise clients in a short period of time.

Several enterprise clients were able to replace or significantly reduce their marketing teams using the product
Users consistently described the experience as simple and intuitive, especially compared to other tools in the market
The unified agent model replaced three siloed single-task bots with one coherent system

What this project taught me

Working on this product taught me how to design for emerging technology where established UX patterns don't yet exist. There was no reference pattern library for agentic software to lean on. I learned how important it is to balance automation with user control, and how design can make complex systems feel trustworthy rather than opaque.

Replacing manual marketing work also meant, in some cases, replacing marketing jobs and that wasn't a comfortable realization the first time I heard it had happened. It reinforced that interface decisions carry consequences beyond usability, and that designing for automation comes with a responsibility to the people whose work it changes, not only to the people buying the product.

Most importantly, it reinforced something I now treat as a working principle: good UX isn't about hiding complexity. It's about guiding users through it with clarity and confidence.

Ishita Neema

Product Designer · Abmatic AI