Breaking
Personal AI builds the memory layer for self-improving AI Founded 2020 in San Francisco as Human AI Labs $7.8M seed to build the Personal Language Model Models ~120M parameters - small by design Enterprise names cited: AT&T, T-Mobile, Comcast, NVIDIA Will Hurd joins the board (2024) Live on Google Cloud Marketplace Personal AI builds the memory layer for self-improving AI Founded 2020 in San Francisco as Human AI Labs $7.8M seed to build the Personal Language Model Models ~120M parameters - small by design Enterprise names cited: AT&T, T-Mobile, Comcast, NVIDIA Will Hurd joins the board (2024) Live on Google Cloud Marketplace
Profile / Artificial Intelligence / Est. 2020

Personal AI.

The company that looked at the race to build one giant brain for everyone - and decided to build a small one that only knows you.

Personal Language Models Private by design San Francisco ~53 people
Personal AI - Unified memory and context for self-improving AI
EXHIBIT A. The pitch in five words: "Unified memory & context." Personal AI's own house style - mono type, a lot of white space, three small dots standing in for a very large idea.
Share the story

Everyone else built a library. Personal AI built a memory.

Type a question into the usual chatbot and you get the average of the internet - confident, fluent, and faintly nobody's. Personal AI starts from the opposite end of the telescope. Its bet is that the most useful model in your life is not the largest one ever trained, but the smallest one that actually knows you: your messages, your meetings, your decisions, the way you phrase a "no." Out of that raw material it trains a Personal Language Model - a compact AI, roughly 120 million parameters, that answers from your data instead of the crowd's.

It is an unfashionable idea in an industry addicted to scale, which is exactly why it is interesting. While the headlines chase trillion-parameter frontier models, this San Francisco company - legally Human AI Labs - has spent since 2020 arguing that personal beats general. Not artificial general intelligence. Artificial personal intelligence.

2020
Founded (as Luther AI)
$7.8M
Seed capital raised
~120M
Parameters per model
3
Co-founders

A second act, written by a two-time founder

Before Personal AI there was Aira - the company CEO Suman Kanuganti built in 2015 to help people who are blind or low-vision read the world around them. That work taught him something the rest of the field is only now relearning: AI is most powerful when it is intimate, personal, and accountable to one human at a time.

He started Personal AI with two co-founders who are not décor on a slide. Sharon Zhang, an MIT CSAIL graduate who built across Nuance and Glint, runs the technology as Chief AI Officer. Kristie Kaiser Sibal, a Forbes 30-under-30 honoree, shapes how the whole thing feels as Chief Design Officer.

The company began life as Luther AI, rebranded to Personal AI in 2021, and has been quietly compounding a single thesis ever since: your knowledge should not retire when you do, and it should not belong to a cloud you don't control.

By 2024 the board had picked up a notable name - Will Hurd, former congressman and one-time OpenAI board member - a useful signal that the "small and personal" argument is being taken seriously in rooms that usually only talk big.

"Not artificial general intelligence - artificial personal intelligence. AI that scales your expertise instead of replacing it."

- The Personal AI thesis, paraphrased from founder talks

Teach it once. It remembers for good.

The product is less "ask a robot" and more "raise a colleague." You feed it what you know; it grounds every answer in that, and it answers in your voice.

01 / MEMORY STACK

Build a memory

An automatic digital library from your everyday text, audio, images, decks and spreadsheets - synced from Gmail, Drive, OneDrive, Dropbox and Outlook.

02 / PLM

Train your model

A Personal Language Model learns your data over time and answers from it - not from the average of the internet.

03 / PERSONAS

Deploy a persona

Multi-modal AI personas with optional voice clones, reachable over chat, SMS and Messenger - your expertise, on call.

04 / ENTERPRISE

Put it to work

Sales automation, lead gen, HR, legal, healthcare and customer service - answers come from company data, kept private.

05 / API

Build on it

A Training Studio and developer API to tune and embed models in your own apps - available via Google Cloud Marketplace.

06 / OWNERSHIP

Keep it yours

Privacy-by-design: your data trains your model, and it stays yours rather than feeding someone else's.

Smaller can be the feature, not the compromise

Personal AI's pitch to enterprises rests on three numbers a CFO cares about - precision, latency and cost. The claim: a model that only has to know one domain can beat a giant generalist on all three. Bars below are illustrative of the positioning, not benchmarked figures.

Answer precision
grounded in you
Latency / speed
runs lean
Cost efficiency
~120M params
Privacy / control
your data, yours

Illustrative positioning based on Personal AI's public materials. Not independent benchmarks.

Sold to the COO before the CIO

Personal AI's enterprise marketing cites a roster of large, recognizable names exploring its personas - the kind of logos that suggest the "personal model" idea has left the lab. Individual plans run around $400/year per persona; enterprise plans around $10,000/year per persona with professional services.

AT&TT-MobileComcast NVIDIABNIWWTAmer Sports

Names referenced in Personal AI's public website materials.

A timeline

2020

Luther AI is born

Suman Kanuganti, Sharon Zhang and Kristie Kaiser Sibal found the company in San Francisco.

2021 - JAN

$3.2M seed

Differential Ventures leads; the company rebrands to Personal AI.

2023 - JAN

$7.8M seed for the PLM

Capital to build the Personal Language Model and prepare a Series A.

2024 - JAN

MODEL-2

Multi-persona, multi-modal release with voice clones, image generation and SMS / Messenger.

2024 - SEP

Google Cloud + a new board member

SLM tech listed on Google Cloud Marketplace; Will Hurd joins the board.

2025

AI personas at work

The founder takes the "personal beats general" argument to podcasts and stages.

The Fact File

  • LEGAL NAME
    Human AI Labs, Inc.
  • HQ
    San Francisco, California
  • FOUNDED
    2020 (as Luther AI)
  • TEAM
    ~53 people (plus a few AI personas)
  • TOTAL RAISED
    ~$12M+ reported
  • CATEGORY
    AI · SaaS · Enterprise · Dev Tools

The staff includes robots

Personal AI's own team page lists AI personas alongside humans - an AI COO named Gates, an AI CFO named Vinny, a customer-service persona named Zoe. The company eats its own cooking.

Interviews & demos

Back to those three small dots

Look again at Personal AI's own picture - the calm cream rectangle, the mono type, the three small dots standing in for a very large idea. It reads like a company that refuses to shout, in a market that mostly does. The dots are the tell: this is a business about small things connected with care, not one enormous brain in the sky.

Whether "personal beats general" wins is still an open question, and the honest version of this story leaves the valuation blank and the Series A unwritten. But the wager is clear, and it is a human one - that the AI worth keeping is the one that remembers what you taught it, answers like you would, and belongs to you. In a year of models that know everything about everyone, Personal AI is quietly building the one that knows something about you.

Find Personal AI

Sources: personal.ai, Wikipedia (Human AI Labs), Crunchbase, BusinessWire, Frontlines.io, Republic, Zapier, public YouTube interviews.