speakVoice & meeting-AI platform · 2018–2024 · Archived
Speak Technologies
If an old link, a browser bookmark, or a search for "Albi by Speak" brought you here — you're in the right place, six years late. I'm Joseph Dattilo. I co-founded this company in 2018, and for six years we ran a platform that placed, transcribed, and actually understood real meetings — production AI, built and shipped before "AI" meant a chat window.
At a glance
2018 founded, as Albi by Speak
6 yrs real product, real customers
9 person distributed team
2024 wound down
What we built
That's Albi's actual logo mark, straight from my own brand archive — I bought every piece of Speak Technologies' IP myself when the board dissolved the company, so this is the original file, not a reconstruction. Albi was also a puppy (that's the whole naming story: "Albi by Speak" because our mascot — the dog you see at the top of this page — had a name before our product did).
Albi's job was to sit in on your meetings, listen, and hand back the part everyone actually wanted: who said what, what got decided, and who owed what to whom afterward. Plug your calendar in, invite Albi like any other guest, and at the end of the meeting it transcribed the thing, pulled out the key moments, and emailed the action items to the right people — so you could actually be in the meeting instead of transcribing it in a notebook.
We started rough (our earliest site ran gag testimonials from a fictional Steve Jobs and "This Woman, Your Mom" — we were a small team having fun before we had customers). By 2020 the testimonials were real, with names attached. By mid-2022 we could say it plainly: "Albi is now GA." We later simplified the brand down to just Speak Meetings and kept going. Along the way we picked up more than 30 beta customers by our own count, including teams at companies like Nike, TransUnion, Verizon, and Chase — and we ran on a SAFE raise backed by more than a thousand people who believed in the idea before it had really proven itself.
What it proved
The part that actually mattered was underneath the note-taking. Sentence classification and named-entity recognition, running on real call and meeting traffic, trained and evaluated by us — years before large language models made that look easy (it was not easy). A nine-person distributed team built the whole stack ourselves: call placement and recording, transcription pipelines, training data, evaluation harnesses, and the deployment pipeline that put a model in front of a live conversation and kept it there.
None of that held up because it benchmarked well in a notebook. It had to survive messy real-world transcripts (accents, crosstalk, people trailing off mid-sentence) where a model that looked good on paper but failed quietly cost a customer real money. In 2018 there was nothing to rent for any of this. You built the whole chain, data to model to production behavior, or you didn't ship. That end-to-end ownership is the thing I still run everything on today.
Thank you
More than a thousand people backed Speak Technologies with real money on a promise, not a finished product — a SAFE, some faith, and for a few of you, an actual seat close enough to watch how a startup gets built from the inside. Thank you. I mean that without any marketing gloss on it: you let us spend six years building something real, before the rest of the world decided artificial intelligence was worth paying attention to. Being early to something that turned out to matter is a rare thing to get to say, and I don't take it lightly.
Speak Technologies wound down in 2024. Six years, a real product, real customers, and a lot of people (investors, beta customers, the nine of us on the team) who took a chance on an idea before it was an easy one to believe in. If you were one of those people, I hope the story was worth something on its own, whatever else came of it. We got to be part of artificial intelligence's history before the boom, and that's still true no matter how the company's story ended.
One more thing worth saying: when the board voted to dissolve the company, I personally bought all of the IP back. Not because it made business sense — because I was proud of what we built together, and I wanted the legacy to live on in some small way. This page is part of that.
Where this goes now
Speak Technologies isn't coming back, but the story doesn't dead-end here. I've written up the fuller version of what we built and what it taught me, and I'm still shipping AI in production today — just with a fleet of coding agents instead of a nine-person NLP team.