Salency
Our story

Four people. Four months. One bet about how sales actually works.

We started Salency because we watched the same thing fail in different rooms — reps inheriting accounts with no memory of what was said, customers repeating themselves, deals losing trust at the handover. Every behavioral fix made it worse. This is the story of what we did instead.

The moment

Late 2024. Howard was helping out on customer success at Sequence. First onboarding call for a new web3 game-studio customer. By minute 15 the customer had said some version of “I already told your colleague this” three times. The rep who closed the deal was two desks away. Nobody had asked him to do a handover. The CRM had a stage field and a one-line note.

Howard stopped thinking of this as a behavior problem. Behavior doesn’t fail at that scale. Infrastructure does.

Who’s building it

Five founding-AE / BD seats in four years — Dora, Sequence, Treasure, Nijta, Viggle (a16z-backed). Ran the HubSpot→Monday CRM migration at Sequence and came out knowing flat fields can’t hold what the customer said. Before sales, he led MUFG Hong Kong's first Panda Bond issuance. MBET, Waterloo.

In January 2026 he interviewed a senior relationship manager at a fintech who said 40–50% of her customers get visibly annoyed when asked to repeat themselves. Her verbatim: “I avoid recording my calls because people tense up.” That was the input constraint named by the person Salency was being built for.

Data analyst at Adaptavist Group for 3+ years, running the operational reporting stack across HubSpot, Snowflake, and DBT. She’s lived the exact pain Salency fixes — revenue data scattered across platforms, no single layer where the story survives.

When Howard showed her the thesis, her first reaction was “I’ve been doing this by hand for five years.” She joined as co-founder two weeks later. Before Adaptavist: Hong Kong commercial banking at Nanyang Commercial Bank (credit for HKD 4B+ listed clients) and operations at a Toronto crypto exchange.

Scaled MakersValley from 0 to $2M ARR (6.5y, NYC) as an applied-AI operator — not a research hire. Built Salency’s extraction and pain-to-product mapping engine. Trained 2,000+ engineers on AI-assisted workflows along the way.

The core technical risk in Salency isn’t “can an LLM summarise a call.” It’s “can extractions stay cited, scoped to a specific catalog, and trustworthy at volume.” He has shipped that class of system before — provenance-tracked AI context systems in production.

Shristi was the designer who shipped Salency’s V1. “Sales intelligence platform’s MVP application using AI-augmented design skills” on her portfolio is this one. She joined the way the others did — already in the work before the decision was formal.

Five years designing enterprise complex-systems tools before Salency. Sole designer at Index Exchange for a 10-engineer ad-tech platform, shipping workflows for blacklisting and account administration at ad-tech scale. Redesigned legacy shipping applications at StatysTech with user research driving the journey rewrites. Fortune 500 HR interfaces at Myplanet. Her process opens with a user, not a wireframe. AI-augmented design is her stated top skill — not a pivot.

What we bet on: institutional memory isn’t another feature. It’s a layer. And the bottleneck is about to shift.

Read the full thesis →