agent@francis state: seeking founding role · runtime: 29y · build: CMU · SCU · on-call: yes linkedin patents contact
PRODUCT EXECUTIVE · AGENTIC AI

I am a completion loop.
zero to one, on repeat.

29 years in enterprise tech, from telephony and VoIP through collaboration platforms, and into AI since 2018: LLM infrastructure, and now autonomous agents. Founder of one company, founding member of another, and the engineer behind the speech tech acquired by Cisco. Platforms shipped to global scale: 30M+ agent-driven summaries, LLM infrastructure at a 50B-tokens-per-month run rate. I sense where the market turns, name the category before it forms, and ship the product that arrives on time.

p50 time-to-conviction
3 wks
fast, once the turn is clear
ventures built
2
Piot (CEO) · AdEntify (fdg member)
category calls
early
+1 pending
domain context
29y
telephony → AI
tokens / month
50B
infra at scale
portfolio managed
$100M+
product P&L
decision threshold
10%
of the info, then move
exits
1
→ Cisco
P1 · thesis

The next category: agents that complete, not just assist

"Conversations are a cost. Completions are the product. Whoever owns the graph of how work actually gets done owns the category."
Every vendor is shipping copilots. The shift that matters comes next, and it rests on three bets.
01 · shift

From assist to coordinate

Today: one agent, one user, one task. Next: multi-agent systems that run cross-functional workflows end to end, with humans owning the judgment calls at explicit gates.

02 · moat

Context is the moat

The winner won't have the best model; everyone rents the same models. The winner will own the org context layer: who owns what, who depends on whom, what was decided and why.

03 · product

Governance is the product

Enterprises don't buy autonomy; they buy governed autonomy. Audit trails, human gates, adversarial testing, and compliance readiness are what turn demos into deployments.

P2 · completion log · 7 workflows · oldest → newest

Career as completed workflows

Every role, the outcome it shipped. No status without an outcome.

1996 → 2005
Enterprise AI · early speech & NLP
Turning language into structured signal, before it was fashionable
SHIPPED
venture
AdEntify · founding member
Radio-based ad network · coded the audio-collection algorithms & apps
BUILT
venture
Piot Networks · founder & CEO
IoT networks for indoor grow operations · built the dev team in India, delivered an operation to an Oakland grow room
BUILT
→ 2020
BabbleLabs · on-device speech & local command tech
Recognition for constrained environments: aviation, handheld radios, home appliances
ACQUIRED · CISCO
2020 → now
Cisco · AI platform leadership
LLM infra at 50B tok/mo · 30M+ agent-driven summaries · semantic search, agentic infra
SHIPPED AT SCALE
2026
Completion Loop · solo prototype
A working phone-AI agent that completes operational calls and builds a live counterparty graph
● LIVE
next
Founding product role · the agentic era
Where the inflection is real and the moat is the graph
● OPEN
P3 · hard lessons · 3 resolved · 1 ongoing · 1 wont-fix

Postmortems I actually ran

The failures are where the real learning compiled. Logged honestly, in the system's own words.

statuslesson
RESOLVED
Built the product before I found the buyer. Talk to customers first, ship second.2026
RESOLVED
A great exit is still an ending. Enjoy the build while it's yours.2020
RESOLVED
Reddit is not a growth channel. (I have the ban to prove it.)2026
ONGOING
The category is obvious to me before it's obvious to the market. Calling the turn early is lonely.open
WONT-FIX
I reach for code when I should reach for customers. Aware of it. Still do it.by design
P4 · capability graph & deploy history

What I bring, as a connected system

Not a skills list. A graph where each capability makes the others stronger.

productzero → one AI infraLLM · scale agentsautonomy speechNLP founderCEO · Piot governthe product
1997Telephony networks · where it started
2000sVoice over IP · real-time comms
2010sCollaboration tech · platforms at scale
2012→14AdEntify · founding member
2016→19Piot Networks · founder & CEO
2018Pivot into AI · the arc bends
2020BabbleLabs · acquired by Cisco
2020+AI platforms at scale · 50B tok/mo
2026Completion Loop · live prototype
◆ governance is the product · shipped in production
guardrails policy + safety gates on every agent action
adversarial testing red-team suites that score agent security before release
PII redaction compliance-grade data handling at scale
audit trails every decision logged, gated, reversible
P5 · now · runtime activity

What's running right now

build activity · illustrative
lessmore
Private work. Click to request access.
P6 · dependencies · what shaped the system

package.json

{
  "zero-to-one": "^2.0.0", founder/CEO of Piot, founding member of AdEntify. the reflex to build, not just spec.   "speech-and-language": "^29.0.0", the whole arc, from acoustic models to agents that talk.   "enterprise-scale": "^50.0.0", 50B tokens a month teaches you what actually breaks.   "call-the-turn-early": "^1.0.0" the bet that the next category is visible before it's fundable. }
P7 · operating runbook · how I run zero-to-one

The runbook, not the pitch

A repeatable process, not a heroic sprint. This is how the loop actually runs.

01discovery first20+ customer conversations before a line of production code. Hypotheses written down, falsifiable, with named kill criteria.
0290-day loopsThesis to tangible prototype to customer validation in 90 to 120 days. Speed is a property of the process, not a crunch.
03kill-switch disciplineConcepts that miss validation get killed, not nursed. Portfolio logic over pet projects. The capital moves to the next hypothesis.
04revenue as proofPaid pilots over signed LOIs over enthusiastic meetings. Willingness-to-pay is the only validation that compounds.
P8 · routes

Where to go next

Two paths, depending on what you're building.